The demand for high-quality, low-latency video streaming is placing strain on conventional internet infrastructures. This article proposes a hybrid unicast–broadcast video delivery framework designed to address this challenge by integrating advanced 5G broadcast technologies with traditional unicast methods. By offloading popular content to a broadcast network, the approach aims to alleviate congestion and enhance overall streaming efficiency. To ensure reliable video segment delivery over the broadcast network, regardless of the physical layer, we incorporate Packet Recovery (PR) and Forward Error Correction (FEC) mechanisms. Additionally, Temporal Layer Injection (TLI) is employed to further improve video quality while maintaining reduced bandwidth requirements compared to traditional unicast-only approaches. This innovative framework leverages 5G terrestrial broadcasting within Over-the-Top (OTT) streaming environments, enabling seamless delivery of adaptive video content with sub-1-second live latency. Comprehensive experimentation and evaluation through large-scale emulation demonstrate the efficacy of this hybrid approach in meeting the evolving demands of modern multimedia delivery systems. Notably, when broadcasting the top three most commonly watched video streams, 63% of viewers no longer need to request video segments via unicast, as they are efficiently delivered over broadcast channels. This hybrid model offers significant scalability, cost reduction for an ISP, and efficiently delivers content directly to user devices without additional intermediaries, improving viewer experience through low-latency, high-quality streaming.

Current networking solutions are unable to satisfy the low-latency requirements of real-time volumetric video conferencing when faced with heavy congestion scenarios. Traditional congestion controllers use packet loss or the change in round-trip time (RTT) to estimate the bandwidth. Commonly, this method is too slow as congestion has already occurred and the receiving user has already experienced a latency spike. Low latency, low loss and scalable throughput (L4S), recently published as RFC 9330, wants to alleviate this problem by aiming for sub 1 ms queuing delay for low-latency traffic by using accurate explicit congestion notification (AccECN) packet marking to notify applications of early congestion. We propose an L4S-based pipeline for volumetric video delivery, which achieves a more consistent latency under congestion compared to web real-time communication (WebRTC). In addition, L4S bandwidth estimation achieves a 45% faster convergence compared to Google congestion control (GCC) estimation, commonly used in WebRTC. Furthermore, in our detailed evaluation setup the L4S application experiences no packet loss, while the WebRTC-based version suffers from irrecoverable packet loss, resulting in 3% of frames being undecodable.

5G networks offer significant advancements over its predecessor, 4G Long-Term Evolution (LTE). Low latency network access, a key requirement enabling near real-time responsiveness as required by applications such as autonomous driving, factory automation and virtual reality, is one of 5G's key features. In this paper, we present the results of a long-term measurement campaign of the uplink end-to-end (e2e) latency experienced by a 5G-capable device using a commercial sub-6Ghz 5G non-standalone (NSA) network. Our results show an average uplink e2e latency of 12ms, with a 95th percentile of 21ms. This compares favorably with an average uplink e2e latency of 35ms and a 95th percentile of 53ms using 4G LTE to reach the same destination. We also characterize and define, through real-world network parameters in the uplink data transmission process, an unexpected latency pattern that impacts the performance of latency-sensitive applications, even in 5G standalone (SA) networks, such as edge computing or ultra-reliable low-latency communication (URLLC)-a new class of applications targeted in 5G networks.

In response to the growing demand for low-latency video streaming, this paper presents a demonstration of a hybrid unicast-broadcast video delivery system that combines 5G terrestrial broadcasting with over-the-top (OTT) streaming methods. The demonstration features a scalable setup with an interactive dashboard, allowing users to experiment with various configurations and observe key metrics such as bandwidth usage, packet loss, buffer size, and live latency in real-time. Key techniques include Low-Latency DASH (LL-DASH) for HTTP Adaptive Streaming (HAS), packet recovery (PR) and Forward Error Correction (FEC) for reliability, Temporal Layer Injection (TLI) for enhanced quality, and Common Media Application Format (CMAF) with Chunked Transfer Encoding (CTE) for reduced latency. The demonstration shows that this scalable hybrid approach can effectively reduce unicast bandwidth to nearly 0 Mb/s in scenarios without packet loss on the broadcast network, and achieve similar bandwidth reductions in lossy broadcast networks with appropriate Forward Error Correction (FEC) settings, while maintaining a live latency lower than 1 second. These results demonstrate the system's potential for optimizing multimedia delivery, significantly reducing unicast bandwidth while maintaining low-latency streaming.

With the advent of container technology, companies have been developing microservice-based applications, converting the old monolithic software into a group of loosely coupled containers, with the aim of offering greater flexibility and improving operational efficiency. When users access microservices, their initial point of contact is typically a load balancer. This component is responsible for distributing incoming traffic or requests between multiple instances of microservices. Traditional load balancing approaches mainly rely on round-robin, or weighted roundrobin algorithms which are inadequate to maintain the overall performance and scalability of microservice-based applications. Microservices are often deployed in dynamic environments needing a more adaptive and efficient load balancing strategy to optimize resources and reduce the overall latency for end users. This paper presents a dynamic load balancer for Kubernetes (K8s) clusters based on Reinforcement Learning (RL). It aims to minimize the overall latency while promoting fair distribution of requests. To achieve this goal, the load balancer considers both current network delays and processing loads in the cluster. The evaluation shows that our solution is effective even in environments where both the network traffic and the processing loads in the cluster change dynamically over time. In addition, this study highlights the flexibility of DeepSets neural networks in solving the load balancing challenge in diverse setups without retraining. The results show that the DeepSets algorithms can solve the microservice load balancing problem even in scenarios up to 30 times larger than the trained setup.

The increasing demand for high-quality video streaming, coupled with the necessity for low-latency delivery, presents significant challenges in today's multimedia landscape. In response to these challenges, this research explores the optimization of adaptive video streaming by integrating 5G terrestrial broadcasting with over-the-top (OTT) streaming methods. A comprehensive integration of forward error correction (FEC), temporal layer injection (TLI), and broadcast techniques enhance the robustness and efficiency of content delivery over broadcast networks and reduce unicast bandwidth to zero in low loss environments. Multiple strategies are compared through an extensive emulation setup for reducing latency in the end-to-end video delivery chain to sub 3-second live latency, demonstrating the effectiveness of a hybrid unicast-broadcast approach in achieving low-latency while maintaining high-quality video streaming performance with significantly reduced bandwidth. For 62.99% of viewers, unicast bandwidth can be reduced to as low as zero when broadcasting the top 3 TV channels.

This paper introduces RFIMap, a system that aims to inexpensively characterize the spatial and temporal distribution of RF spectrum occupancy of any indoor space at fine granularity (tens of centimeters). RFIMap builds rich wide-band indoor spectrum occupancy maps using low-cost and battery-free commodity RFID tags. RFIMap's spectrum maps have wide-ranging applications such as monitoring ambient interference in smart manufacturing, and smart hospitals. RFIMap relies on the observation that commodity RFID tags naturally reflect ambient transmission at other frequency bands, without any modification. RFIMap uses these reflections to estimate the ambient signal power originally received at these tags. RFIMap further performs a careful modeling of indoor multipath to build a dense spectrum map with fine spatial granularity. Our experiments demonstrate spatial spectrum measurement with 2.15 dB of median error at 2.4 GHz, 4.45 dB of median error at 470-700 MHz TV whitespace band, 2.1 dB of median error at 1.8-1.9 GHz in diverse industrial and university settings.

In the mid-term future, vehicles will generate large amounts of data for both standalone usage (e.g., to recognize road features and external elements such as lanes, signs, and pedestrians) and cooperative usage (e.g., lane merging). However, processing the captured video and image data results comes with significant computational requirements (e.g., GPUs). Computer vision tasks, such as feature extraction, are unfeasible from a business perspective if performed directly in the User Equipment (UE), as automotive manufacturers are unwilling to increase the end-product’s costs. Thus, the logical solution is to collect and upload this data to be processed elsewhere. Nonetheless, processing the data as close to the vehicle is important due to latency constraints, thus calling for the use of Mobile Edge Computing (MEC). An additional benefit of this scenario, in which 5G connectivity enables data to be offloaded to the edge, is that the data from our car is not processed alone. Data from several sources, e.g., multiple vehicles and fixed cameras, can be offloaded to the edge node and processed together, enhancing its quality as more sources of data enhance the prediction output of machine-learning models. This demo showcases a video recording from a vehicle uploaded to an edge node via 5G software-defined-radio FPGA devices. There, a YOLO application to detect objects processes the video and communicates this information to the vehicle, ensuring QoS metrics even when the UE performs handover to a different cell or geographical area.

Recent proposals for reconfigurable data center networks have shown that providing multiple time-varying paths can improve network capacity and lower physical latency. However, existing TCP variants are ill-suited to utilize available capacity because their congestion control cannot react quickly enough to drastic variations in bandwidth and latency.

Sketching algorithms or sketches have emerged as a promising alternative to the traditional packet sampling-based network telemetry solutions. At a high level, they are attractive because of their high resource efficiency and provable accuracy guarantees. While there have been significant recent advances in various aspects of sketching for networking tasks, many fundamental challenges remain unsolved that are likely stumbling blocks for adoption. Our contribution in this paper is in identifying and formulating these research challenges across the ecosystem encompassing network operators, platform vendors/developers, and algorithm designers. We hope that these serve as a necessary fillip for the community to enable the broader adoption of sketch-based telemetry.

The rapid growth in urban population poses significant challenges to moving city dwellers in a fast and convenient manner. This paper contributes to solving the challenges from the viewpoint of passengers by improving their on-vehicle experience. Specifically, we focus on the problem: Given an urban public transit network and a number of passengers, with some of them controllable and the rest uncontrollable, how can we plan for the controllable passengers to improve their experience in terms of their service preference? We formalize this problem as a multi-agent path planning (MAPP) problem with soft collisions, where multiple controllable passengers are allowed to share on-vehicle service resources with one another under certain constraints. We then propose a customized version of the SC-M* algorithm to efficiently solve the MAPP task for bus transit system in complex urban environments, where we have a large passenger size and multiple types of passengers requesting various types of service resources. We demonstrate the use of SC-M* in a case study of the bus transit system in Porto, Portugal. In the case study, we implement a data-driven on-vehicle experience simulator for the bus transit system, which simulates the passenger behaviors and on-vehicle resource dynamics, and evaluate the SC-M* on it. The experimental results show the advantages of the SC-M* in terms of path cost, collision-free constraint, and the scalability in run time and success rate.

Vehicular applications are becoming increasingly complex and resource hungry (e.g. autonomous driving). Today, they run entirely on the vehicle, which is a costly solution that also imposes undesirable resource constraints. This paper uses Simultaneous Localization and Mapping (SLAM) as an example application to explore how these applications can instead leverage edge clouds, utilizing their inexpensive and elastic resource pool. This is challenging as these applications are often latency-sensitive and mission-critical. They also process high-bandwidth sensor data streams and maintain large, complex data structures. As a result, traditional offloading techniques generate too much traffic, incurring high delay. To overcome these challenges, we designed CloudSLAM. It partitions SLAM between the vehicle and the edge. To manage the complex, replicated SLAM state, we propose a new consistency model, Output-driven Consistency, that allows us to maintain a level of consistency that is sufficient for accurate SLAM output while minimizing network traffic. This paper motivates and describes our offloading design and discusses the results of an extensive performance evaluation of a CloudSLAM prototype based on ORB-SLAM.

Over the last two decades, several algorithms have been proposed to infer the type of relationship between Autonomous Systems (ASes). While the recent works have achieved increasingly higher accuracy, there has not been a systematic study on the uncertainty of AS relationship inference. In this paper, we analyze the factors contributing to this uncertainty and introduce a new paradigm to explicitly model the uncertainty and reflect it in the inference result. We also present UNARI, an exemplary algorithm implementing this paradigm, that leverages a novel technique to capture the interdependence of relationship inference across AS links.

Client mobility is a fundamental challenge when accessing the current Internet, especially in the context of vehicular networking because of its intermittent connectivity nature. Meanwhile, today's network applications are evolving from host-to-host communication to content retrieval, and fostering new designs of Information-centric networking (ICN) protocol and system optimized towards this end. In this paper, we present SoftStage, a client instructed ICN-based network layer function that effectively manages the edge caching to perform reactive content staging to improve vehicular content delivery without any assumption about the client mobility pattern. Experimental results based on an implementation in eXpressive Internet Architecture (XIA) shows that SoftStage achieves up to 10x throughput gain in vehicular networking environments.

TCP is an important factor affecting user-perceived performance of Internet applications. Diagnosing the causes behind TCP performance issues in the wild is essential for better understanding the current shortcomings in TCP. This paper presents a TCP flow performance analysis framework that classifies causes of TCP stalls. The framework forms the basis of a tool that we use to analyze packet-level traces of three services (cloud storage, software download, and web search) deployed by a popular service provider. We find that as many as 20% of the flows are stalled for half of their lifetime. Network-related causes, especially timeout retransmissions, dominate the stalls. A breakdown of the causes for timeout retransmission stalls reveals that double retransmission and tail retransmission are among the top contributors. The importance of these causes depends however on the specific service. Based on these observations, we propose smart-retransmission time out (S-RTO), a mechanism that mitigates timeout retransmission stalls through careful and gentle aggression for retransmission. S-RTO is evaluated in a controlled network and also in a production network. The results consistently show that it is effective at improving TCP performance, especially for short flows.

Automated data collection in urban transportation systems produces a large volume of passenger data. However, quite a few of the data are still incomplete, limiting the insight into passenger mobility. The unavailability of destination information in entry-only passenger data is a very common issue. Traditional approaches for estimating passenger destinations rely on heuristics that can recover only some of the missing destinations. To deal with the remaining incomplete data, this paper, for the first time, proposes a second-order inference methodology to leverage semi-supervised self-training to infer the missing destinations. The methodology involves the design of a base learner to predict the missing destinations based on the statistics of a selected similarity-based "training set", and the design of a selection strategy to select new data with high prediction confidence to update the training set. To further improve the inference, we incorporate personal history priors to modify the base learner. We evaluate our designs using two data sources: a real-data inspired traffic-passenger behavior simulation in the city of Porto, Portugal, and the real bus Automated Fare Collection (AFC) data collected from the same city. The experimental results show that compared to baseline methods that do not use self-training, our approach significantly improves the inference performance and achieves notably high accuracies.

Internet communication today typically involves intermediary middleboxes like caches, compression proxies, or virus scanners. Unfortunately, as encryption becomes more widespread, these middleboxes become blind and we lose their security, functionality, and performance benefits. Despite initial efforts in both industry and academia, we remain unsure how to integrate middleboxes into secure sessions---it is not even clear how to define "secure" in this multi-entity context.

Outsourcing services to third-party providers comes with a high security cost-to fully trust the providers. Using trusted hardware can help, but current trusted execution environments do not adequately support services that process very large scale datasets. We present LASTGT, a system that bridges this gap by supporting the execution of self-contained services over a large state, with a small and generic trusted computing base (TCB). LASTGTuses widely deployed trusted hardware to guarantee integrity and verifiability of the execution on a remote platform, and it securely supplies data to the service through simple techniques based on virtual memory. As a result, LASTGTis general and applicable to many scenarios such as computational genomics and databases, as we show in our experimental evaluation based on an implementation of LAST-GT on a secure hypervisor. We also describe a possible implementation on Intel SGX.

The Internet's inter-domain routing infrastructure, provided today by BGP, is extremely rigid and does not facilitate the introduction of new inter-domain routing protocols. This rigidity has made it incredibly difficult to widely deploy critical fixes to BGP. It has also depressed ASes' ability to sell value-added services or replace BGP entirely with a more sophisticated protocol. Even if operators undertook the significant effort needed to fix or replace BGP, it is likely the next protocol will be just as difficult to change or evolve. To help, this paper identifies two features needed in the routing infrastructure (i.e., within any inter-domain routing protocol) to facilitate evolution to new protocols. To understand their utility, it presents D-BGP, a version of BGP that incorporates them.

Code identity is a fundamental concept for authenticated operations in Trusted Computing. In today's approach, the overhead of assigning an identity to a protected service increases linearly with the service code size. In addition, service code size continues to grow to accommodate richer services. This trend negatively impacts either the security or the efficiency of current protocols for trusted executions. We present an execution protocol that breaks the dependency between the code size of the service and the identification overhead, without affecting security, and that works on different trusted components. This is achieved by computing an identity for each of the code modules that are actually executed, and then building a robust chain of trust that links them together for efficient verification. We implemented and applied our protocol to a widely-deployed database engine, improving query-processing time up to 2× compared to the monolithic execution of the engine.

Networking research benefits from controlled, repeatable experimentation using simulation and emulation systems. Making simulations realistic is a challenge for wireless systems and is especially difficult for vehicular networks. This paper presents a general framework for modeling and reproducing environment-specific channel properties. We show that one can estimate localized environment information, and adding such information to state-of-the art channel models significantly increases accuracy. We describe the proposed framework and validate it for fading and line-of-sight effects in vehicle-to-vehicle channels. While suitable models can provide a close approximation of channel conditions, accuracy is limited by the quality of the input information about the environment being modeled. We present a systematic approach to estimating location-specific scattering properties using aerial photography. Using signal-level channel emulation, we show that the improved fading models produce more accurate results at the packet/link level. The error rates of the improved model are 45 and 22 percent lower than using previous state of the art, for Doppler spectrum similarity and packet delivery ratio, respectively. The proposed models-and their implementation-have been designed to minimize run-time complexity while preserving accuracy. The implementation is efficient gh to allow concurrent real-time simulation of many of channels.

Video streaming on mobile devices is prone to a multitude of faults and although well established video Quality of Experience (QoE) metrics such as stall frequency are a good indicator of the problems perceived by the user, they do not provide any insights about the nature of the problem nor where it has occurred. Quantifying the correlation between the aforementioned faults and the users' experience is a challenging task due the large number of variables and the numerous points-of-failure.

TCP is an important factor affecting user-perceived performance of Internet applications. Diagnosing the causes behind TCP performance issues in the wild is essential for better understanding the current shortcomings in TCP. This paper presents a TCP flow performance analysis framework that classifies causes of TCP stalls. The framework forms the basis of a tool that is publicly available to the research community. We use our tool to analyze packet-level traces of three services (cloud storage, software download and web search) deployed by a popular Chinese service provider. We find that as many as 20% of the flows are stalled for half of their lifetime. Network-related causes, especially timeout retransmission, dominate the stalls. A breakdown of the causes for timeout retransmission stalls reveals that double retransmission and tail retransmission are among the top contributors. The importance of these causes depends however on the specific service. We also propose S-RTO, a mechanism that mitigates timeout retransmission stalls. S-RTO has been deployed on production front-end servers and results show that it is effective at improving TCP performance, especially for short flows.

Online privacy is more important now than ever. Using encryption goes a long way by hiding application data from third parties, but some amount of private information is still exposed in packet headers. Tools like Tor are designed to address this concern, and, more recently, research efforts to replace IP have begun to consider how a network architecture itself can improve privacy.

It is extremely difficult to deploy newinter-domain routing protocols in today's Internet. As a result, the Internet's baseline protocol for connectivity, BGP, has remained largely unchanged, despite known significant flaws. The difficulty of deploying new protocols has also depressed opportunities for (currently commoditized) transit providers to provide value-added routing services. To help, we identify the key deployment models under which new protocols are introduced and the requirements each poses for enabling their usage goals. Based on these requirements, we argue for two modifications to BGP that will greatly improve support for new routing protocols.

A significant fraction of Internet traffic is now encrypted and HTTPS will likely be the default in HTTP/2. However, Transport Layer Security (TLS), the standard protocol for encryption in the Internet, assumes that all functionality resides at the endpoints, making it impossible to use in-network services that optimize network resource usage, improve user experience, and protect clients and servers from security threats. Re-introducing in-network functionality into TLS sessions today is done through hacks, often weakening overall security.

We show how to leverage trusted computing technology to design an efficient fully-passive replicated system tolerant to arbitrary failures. The system dramatically reduces the complexity of a fault-tolerant service, in terms of protocols, messages, data processing and non-deterministic operations. Our replication protocol enables the execution of a single protected service, replicating only its state, while allowing the backup replicas to check the correctness of the results. We implemented our protocol on Trusted Computing (TC) technology and compared it with two recent replication systems.

Increased user concern over security and privacy on the Internet has led to widespread adoption of HTTPS, the secure version of HTTP. HTTPS authenticates the communicating end points and provides confidentiality for the ensuing communication. However, as with any security solution, it does not come for free. HTTPS may introduce overhead in terms of infrastructure costs, communication latency, data usage, and energy consumption. Moreover, given the opaqueness of the encrypted communication, any in-network value added services requiring visibility into application layer content, such as caches and virus scanners, become ineffective.

Parking is a major problem of car transportation, with important implications in traffic congestion and urban landscape. Reducing the space needed to park cars has led to the development of fully automated and mechanical parking systems. These systems are, however, limitedly deployed because of their construction and maintenance costs. Leveraging on semi and fully-autonomous vehicular technology, as well as on the electric propulsion paradigm and in vehicular ad hoc networking, we propose a new parking concept where the mobility of parked vehicles is managed by a parking lot controller to create space for cars entering or exiting the parking lot, in a collaborative manner. We show that the space needed to park such vehicles can be reduced to half the space needed with conventional parking lot designs. We also show that the total travelled distance of vehicles in this new parking lot paradigm can be 30% less than in conventional parking lots. Our proposal can have important consequences in parking costs and in urban landscape.

Though most would agree that accountability and privacy are both valuable, today's Internet provides little support for either. Previous efforts have explored ways to offer stronger guarantees for one of the two, typically at the expense of the other; indeed, at first glance accountability and privacy appear mutually exclusive. At the center of the tussle is the source address: in an accountable Internet, source addresses undeniably link packets and senders so hosts can be punished for bad behavior. In a privacy-preserving Internet, source addresses are hidden as much as possible.

Vehicle-to-Vehicle (V2V) communication is a core technology for enabling safety and non-safety applications in next generation intelligent transportation systems. Due to relatively low heights of the antennas, V2V communication is often influenced by topographic features, man-made structures, and other vehicles located between the communicating vehicles. On highways, it was shown experimentally that vehicles can obstruct the line of sight (LOS) communication up to 50 percent of the time; furthermore, a single obstructing vehicle can reduce the power at the receiver by more than 20 dB. Based on both experimental measurements and simulations performed using a validated channel model, we show that the elevated position of antennas on tall vehicles improves communication performance. Tall vehicles can significantly increase the effective communication range, with an improvement of up to 50 percent in certain scenarios. Using these findings, we propose a new V2V relaying scheme called tall vehicle relaying (TVR) that takes advantage of better channel characteristics provided by tall vehicles. TVR distinguishes between tall and short vehicles and, where appropriate, chooses tall vehicles as next hop relays. We investigate TVR's system-level performance through a combination of link-level experiments and system-level simulations and show that it outperforms existing techniques.

Question answering (Q&A) communities have been gaining popularity in the past few years. The success of such sites depends mainly on the contribution of a small number of expert users who provide a significant portion of the helpful answers, and so identifying users that have the potential of becoming strong contributers is an important task for owners of such communities.

Despite the plethora of incremental deployment mechanisms proposed, rapid adoption of new network-layer protocols and architectures remains difficult as reflected by the widespread lack of IPv6 traffic on the Internet. We show that all deployment mechanisms must address four key questions: How to select an egress from the source network, how to select an ingress into the destination network, how to reach that egress, and how to reach that ingress. By creating a design space that maps all existing mechanisms by how they answer these questions, we identify the lack of existing mechanisms in part of this design space and propose two novel approaches: the "4ID" and the "Smart 4ID". The 4ID mechanism utilizes new data plane technology to flexibly decide when to encapsulate packets at forwarding time. The Smart 4ID mechanism additionally adopts an SDN-style control plane to intelligently pick ingress/egress pairs based on a wider view of the local network. We implement these mechanisms along with two widely used IPv6 deployment mechanisms and conduct wide-area deployment experiments over PlanetLab. We conclude that Smart 4ID provide better overall performance and failure semantics, and that innovations in the data plane and control plane enable straightforward incremental deployment.

Vehicular ad hoc networks (VANETs) are seen as an important enabling technology for improving both traffic safety and efficiency. Virtual Traffic Lights (VTLs) are a promising proposal for reducing travel time by efficiently controlling road intersections. VTLs use vehicle-to-vehicle communication to dynamically optimize traffic flow and they display traffic light information on the windshield. However, research so far has assumed that all vehicles are equipped with VTL support and it has ignored the incremental deployment phase, which could last decades. In this paper we present a solution for a VTL partial deployment scenario that is based on the idea of having VTL equipped cars display traffic light information on the outside of the vehicle. This allows drivers in non-equipped vehicles, or even pedestrians, to see the light color and respond accordingly. We show that the benefits of VTLs in terms of intersection throughput and average delay reduction grow as a function of the penetration rate of equipped vehicles.

Early Internet architecture design goals did not put security as a high priority. However, today Internet security is a quickly growing concern. The prevalence of Internet attacks has increased significantly, but still the challenge of detecting such attacks generally falls on the end hosts and service providers, requiring system administrators to detect and block attacks on their own. In particular, as social networks have become central hubs of information and communication, they are increasingly the target of attention and attacks. This creates a challenge of carefully distinguishing malicious connections from normal ones. Previous work has shown that for a variety of Internet attacks, there is a small subset of connection measurements that are good indicators of whether a connection is part of an attack or not. In this paper we look at the effectiveness of using two different co-clustering algorithms to both cluster connections as well as mark which connection measurements are strong indicators of what makes any given cluster anomalous relative to the total data set. We run experiments with these co-clustering algorithms on the KDD 1999 Cup data set. In our experiments we find that soft co-clustering, running on samples of data, finds consistent parameters that are strong indicators of anomalous detections and creates clusters, that are highly pure. When running hard co-clustering on the full data set (over 100 runs), we on average have one cluster with 92.44% attack connections and the other with 75.84% normal connections. These results are on par with the KDD 1999 Cup winning entry, showing that co-clustering is a strong, unsupervised method for separating normal connections from anomalous ones. Finally, we believe that the ideas presented in this work may inspire research for anomaly detection in social networks, such as identifying spammers and fraudsters.

The end-to-end nature of today's transport protocols is increasingly being questioned by the growing heterogeneity of networks and devices, and the need to support in-network services. To address these challenges, we present Tapa, a transport architecture that systematically combines two concepts. First, it unbundles today's transport such that network specific functions (e.g., congestion control) are implemented on a per-segment basis, where a segment spans a part of the end-to-end path that is homogeneous (e.g., wired Internet or an access network) while functions that relate to application semantics (e.g., data ordering) are still implemented end-to-end. Second, it has an explicit notion of in-network services (e.g., caching, opportunistic content retrieval, etc) that can be supported while maintaining precise end-to-end application semantics. In this paper, we present the basic design, implementation and evaluation of Tapa. We also present diverse case studies that show how Tapa can easily support opportunistic content retrieval in online social networks, various mobile and wireless optimizations, and an in-network energy saving service that improves battery life of mobile devices.

This paper presents an efficient opportunistic retransmission protocol (PRO, Protocol for Retransmitting Opportunistically) to improve the performance of IEEE 802.11 WLANs. PRO is a link-layer protocol that allows overhearing nodes to function as relays that retransmit on behalf of a source after they learn about a failed transmission. Relays with better connectivity to the destination have a higher chance of delivering the packet than the source, thereby resulting in a more efficient use of the channel. PRO has four main features. First, channel reciprocity coupled with a runtime calibration process is used to estimate the instantaneous link quality to the destination. Second, a local qualification process filters out poor relays early. Third, a distributed relay selection algorithm chooses the best set of eligible relays among all qualified relays and prioritizes them. Finally, 802.11e Enhanced Distributed Channel Access (EDCA) is leveraged to make sure high-quality relays transmit with higher probability. PRO is designed to coexist with legacy 802.11 stations. Our extensive evaluation on both a controlled testbed and in the real world shows that PRO can improve throughput in diverse wireless environments. PRO helps the most when there is significant contention for the ether, under fading, and with user mobility.

TV white spaces offer an exciting opportunity for increasing spectrum availability, but white space devices (WSDs) cannot interfere with primary users, including TV channels and wireless microphones (mics). Mics are particularly challenging because their use is dynamic and it is hard to avoid interference since mic receivers are receive-only devices. For this reason the FCC and other regulatory agencies have made very conservatives rules that require WSDs to vacate any TV channel that is used by a mic. However, our measurements show that mics typically require only 5% of a channel, wasting as much as 95% of the spectrum.

Motivated by limitations in today's host-based IP network architecture, recent studies have proposed clean-slate network architectures centered around alternative first-class principals, such as content, services, or users. However, much like the host-centric IP design, elevating one principal type above others hinders communication between other principals and inhibits the network's capability to evolve. Our work presents the eXpressive Internet Architecture (XIA), an architecture with native support for multiple principals and the ability to evolve its functionality to accommodate new, as yet unforeseen, principals over time. XIA also provides intrinsic security: communicating entities validate that their underlying intent was satisfied correctly without relying on external databases or configuration.

This paper presents a framework for signal-level emulation of propagation effects over generalized fading channels at the scale of entire networks. Network emulation enables research into network- scale systems - which would otherwise be limited to low-fidelity network simulators and one-off field experiments - to use real radio hardware and realistic channel models. Our hardware and software architecture goes beyond previous work in that it supports real-time emulation of a very general and parametric class of channels, which includes vehicular (broadband mobile-to-mobile) and indoor channels in addition to classical stationary-to- mobile and stationary-to-stationary channels.

The Internet uses a model in which Autonomous Systems (AS) peer bilaterally with each other, resulting in a cascaded connectivity model: the end-to-end service that an end-user sees is the result of this cascade of bilateral traffic agreements. As a result, the routing decisions at each AS beyond the next hop are implicitly delegated, so an AS has limited control over the remaining path. While this may be considered a key feature of the Internet (e.g., helps scalability), the impact of this cascading model on the structure of the Internet is not yet well understood. In this paper, we analyze this cascaded model using concepts of game theory. Although our model cannot capture the full complexity of the real Internet - we actually aim at simplicity and try to only isolate the cascading effect -, our results suggest that cascading bilateral agreements brings order to an arbitrary graph. In particular, it brings forward some well-known properties of the Internet topology, such as the current AS hierarchy and common peering models. Finally, we compare our results with topological and routing data, obtained from CAIDA, looking for experimental evidence of our results, while further exploring the insight obtained from our model.

We discuss the problem of composing Quality-of-Service across several administrative domains considering an evolved DiffServ notion of Per-Domain Behavior (PDB). We discuss two components of the problem: path discovery and the effect of composing PDBs. We assume the most general scenario: domains are free to adopt any PDB, as long there are common semantics and a set of well-known QoS parameters (e.g. one-way delay, peak/sustained bandwidth, etc.). Thus we adopt an abstract representation of PDBs. We obtain three main results. First, we show that a distributed path discovery scheme is feasible and more scalable than a centralized one. Second, we show the outcome of inter-domain QoS composition, for general metrics and for Internet-alike topologies. Finally, and as an important practical result, we show that the Internet may not need a centralized governance model, in terms of the definition of inter-domain PDBs.

Recent work demonstrates that directional antennas have significant potential to improve wireless network capacity in indoor environments. This paper provides a broader exploration of the design space of indoor directional antenna systems along two main dimensions: antenna configuration and antenna control. Studying a number of alternative configurations, we find that directionality on APs and clients can significantly improve performance, even over other configurations with stronger directionality. Moreover, it is sufficient to have a small number of narrow beam antennas to achieve such gains, thus making such a solution practical for actual deployment. Designing systems with directional APs and clients for increased spatial reuse comes, however, with a number of challenges in the way the directional antennas are controlled. Antenna control needs to encompass antenna orientation algorithms, an appropriate MAC layer protocol, and novel client-AP association solutions. To overcome these challenges, we propose Speed, a distributed directional antenna control system that is easy to deploy and significantly improves network capacity over existing solutions.

Energy management is a critical issue for mobile devices, with network activity often consuming a significant portion of the total system energy. In this paper, we propose Catnap, a system that reduces energy consumption of mobile devices by allowing them to sleep during data transfers. Catnap exploits high bandwidth wireless interfaces -- which offer significantly higher bandwidth compared to available bandwidth across the Internet -- by combining small gaps between packets into meaningful sleep intervals, thereby allowing the NIC as well as the device to doze off. Catnap targets data oriented applications, such as web and file transfers, which can afford delay of individual packets as long as the overall transfer times do not increase. Our evaluation shows that for small transfers (128kB to 5MB), Catnap allows the NIC to sleep for up to 70% of the total transfer time and for larger transfers, it allows the whole device to sleep for a significant fraction of the total transfer time. This results in battery life improvement of up to 2-5x for real devices like Nokia N810 and Thinkpad T60.

Bandwidth demands and timing constraints are two major challenges for wireless video streaming applications. In this paper, we present a hybrid spatial/temporal retransmission protocol that tackles both of these challenges. To increase individual throughput as well as overall network capacity, the system uses an opportunistic retransmission protocol (PRO, Protocol for Retransmitting Opportunistically) that relies on overhearing nodes distributed in physical space to function as relays that retransmit failed packets on behalf of the source. Specifically, the best relay out of the set of nodes that currently have the copy of the packet is responsible for retransmitting (relaying) the packet. Relays with stronger connectivity to the destination have a higher chance of delivering packets successfully than the source, thereby resulting in a more efficient use of the channel. To meet timing constraints, a Time-based Adaptive Retransmission strategy (TAR) is applied by both the source and the relays. With TAR, the MAC dynamically determines whether to (re)transmit or discard a packet based on the retransmission deadline of the packet assigned by the video server. This significantly reduces the number of late packet arrivals at the receiver. Our extensive evaluation results both on a testbed and in the real world demonstrate that hybrid temporal/spatial retransmission can boost streaming performance in diverse wireless environments. The benefits are most pronounced for busy networks, under fading conditions, or for mobile users.

Networking researchers have been using tools like wireshark and tcpdump to sniff packets on physical links that use different types of datalink protocols, e.g. Ethernet or 802.11, allowing them to monitor higher level protocols sharing these links. However, monitoring wireless links is more challenging, since the transmission medium is shared by flows using diverse datalink protocols (e.g. 802.11, Bluetooth) and physical layer schemes (e.g. QPSK and GFSK). To this end, we propose RFDump, a software architecture for monitoring packets on heterogeneous wireless networks. The key idea underlying our architecture is the use of a fast detection stage which can tentatively map signals to protocols very efficiently. As a result, RFDump can scale up to a modest number (5-10) of wireless technologies.

When two administratively independent domains need to engage in cooperation of some kind, some initial common known point of contact must exist, both in terms of the application topological position (i.e., IP address, transport protocol and port) and in terms of the interface technology (e.g., the protocol suite of the interaction). In a world of many domains of many types (not necessarily autonomous systems in the sense of interdomain routing), manual or single-directory-based operations are not practical (or even feasible); hence, some form of autonomic discovery and handshaking of domains is needed. We propose and evaluate several strategies to allow autonomic bootstrapping of inter-domain operations, all fit to be deployed as BGP extensions. We show the time/message overhead tradeoff: it's possible to obtain the absolute minimum message complexity of O(N) but at the cost of central administrative burden or, for fully distributed schemes, a tradeoff (message/time complexity) of O(N2)/O(N) or O(N2)/O(log2N).

Forward error correction (FEC) coding and automatic repeat request (ARQ) are two commonly used techniques to tackle packet erasures in wireless video streaming. To preserve flexibility while reducing end-to-end latency, hybrid ARQ (HARQ) has been proposed as an alternative that combines the advantages of FEC and ARQ. This paper presents cooperative hybrid ARQ (CHARQ) that can further boost wireless video streaming quality. CHARQ takes advantage of the fact that, in the wireless environment, broadcast is free (from the sender's point of view) and that errors are location dependent so an intermediate proxy that overhears the source's transmission may transmit on behalf of the source to increase throughput efficiency. Simulation results show improved performance with the aid of cooperative proxies.

Wireless networking has become a popular topic in both undergraduate and graduate courses. However, putting together good assignments in wireless networking is difficult because the behavior of the wireless network depends strongly on the physical environment. We have used a wireless networking testbed based on signal propagation emulation in a number of wireless networking courses. The wireless emulator supports highly realistic experiments, while also offering a high degree of control and repeatability. This combination is very useful in a teaching context. In this paper we give an overview of the wireless emulator and we describe how it was used to support assignments and open-ended projects in several courses.

This paper presents an efficient opportunistic retransmission protocol (PRO, Protocol for Retransmitting Opportunistically) to improve the performance of IEEE 802.11 WLANs. PRO is a link-layer protocol that allows overhearing nodes to function as relays that retransmit on behalf of the source after they learn about a failed transmission. Relays with stronger connectivity to the destination have a higher chance of delivering the packet than the source does, thereby resulting in a more efficient use of the channel. PRO has four main features. First, channel reciprocity coupled with a run-time calibration process is used to estimate the instantaneous link quality to the destination. Second, a local qualification process filters out poor relays early. Third, a distributed relay selection algorithm chooses the best set of eligible relays among all qualified relays and prioritizes them. Finally, 802.11e Enhanced Distributed Channel Access (EDCA) is leveraged to make sure high priority relays transmit with high probability. PRO is designed to coexist with legacy 802.11 stations. We have implemented PRO in the driver of a commodity wireless card. Our extensive evaluation on both a controlled testbed and in the real world shows that PRO boosts throughput in diverse wireless environments, and especially in when there is significant contention for the channels, under fading, and with user mobility.

The demand for wireless bandwidth in indoor environments such as homes and offices continues to increase rapidly. Although wireless technologies such as MIMO can reach link throughputs of 100s of Mbps (802.11n) for a single link, the question of how we can deliver high throughput to a large number of densely-packed devices remains an open problem. Directional antennas have been shown to be an effective way to increase spatial reuse, but past work has focused largely on outdoor environments where the interactions between wireless links can usually be ignored. This assumption is not acceptable in dense indoor wireless networks since indoor deployments need to deal with rich scattering and multipath effects. In this paper we introduce DIRC, a wireless network design whose access points use phased array antennas to achieve high throughput in dense, indoor environments. The core of DIRC is an algorithm that increases spatial reuse and maximizes overall network capacity by optimizing the orientations of a network of directional antennas. We implemented DIRC and evaluated it on a nine node network in an enterprise setting. Our results show that DIRC improves overall network capacity in indoor environments, while being flexible enough to adapt to node mobility and changing traffic workloads.

Wireless channel emulators are important tools for testing radio devices, especially in mobile environments. Wireless network emulators give the same accuracy and control for testing radio network systems that traditional channel emulators give to point to point radio links. Network emulators require many more independent channels than traditional channel emulators. This problem is particularly challenging for the real time channel simulator in the emulator. The challenges of designing a wireless network channel simulator are discussed and a design is presented on a Xilinx Virtex-II Pro FPGA. This channel simulator can model 210 independent channels between 15 nodes with a bandwidth of 90 MHz. The performance of the design was verified by measuring transport-layer throughput between 802.11b radios transmitting through the channel simulator.

Most pervasive computing technologies focus on helping users with computer-oriented tasks. In this NSF-funded project, we instead focus on using computers to support user-centered "activities" that normally do not involve the use of computers. Examples may include everyday tasks around such as answering the doorbell or doing laundry. A focus on activity-based computing brings to the foreground a number of unique challenges. These include activity definition and representation, system design, interfaces for managing activities, and ensuring robust operation. Our project focuses on the first two challenges.

QoS provisioning as a generic telecom service is only useful if deployed end-to-end, so it must cover both the access segment and the interdomain segment of routes. This work focuses on the interdomain component of QoS architectures. Although the research community has been discussing several aspects of the problem, integrated architectures for true end-to-end QoS are still largely conceptual, especially if one considers mesh models and not simpler cascaded ones. Our inter-domain proposal combines offline service discovery and registration with real-time congestion notification, allowing us to significantly reduce the number of packets that were not suitably serviced, in the sense of fulfilling required Service Level Agreement constraints. We qualify our approach of dasiastatelesspsila since our proposal doesnpsilat use additional state in forwarding elements which is one of the most dramatic problems for interdomain scenarios.

Wireless multi-hop access networks are an increasingly popular option to provide cost-efficient last-mile Internet access. However, despite extensive research, performance of even basic communication services, such as TCP, is still problematic. Measurements collected on a wireless testbed indicate that the poor performance of multi-hop access networks is caused by poor interactions between TCP congestion control and link- layer bit-rate adaptation resulting in severely reduced network efficiency even over short wireless paths (< 6 hops). However, bit-rate adaptation improves fairness across TCP flows. The same principal observations hold for hybrid wireless/wireline paths. To investigate approaches to improve TCP performance, we present a simple model that captures the cause for the inefficiency of TCP over autorate links. We then examine several techniques at both the TCP level and the link layer (TCP Vegas, clamping, limiting the buffer size at the wireless routers) to alleviate contention. None of these techniques works for all scenarios, but the simple approach to limit the buffer size is attractive in many settings that include four or more wireless hops.

Testing and evaluating protocols and applications for wireless networks and mobile users is challenging because the physical environment has a significant impact on the behavior and dynamics of the system. It is however important that these physical world effects are considered during system impleme

Increasingly, 802.11 devices are being used by mobile users. This results in very dynamic wireless channels that are difficult to use efficiently. Current rate selection algorithms are dominated by probe-based approaches that search for the best transmission rate using trial-and-error. In mobile environments, probe-based techniques often perform poorly because they inefficiently search for the moving target presented by the constantly changing channel. We have developed a channel-aware rate adaptation algorithm CHARM - that uses signal strength measurements collected by the wireless cards to help select the transmission rate. Moreover, unlike previous approaches CHARM leverages channel reciprocity to obtain channel information, so the information is available to the transmitter without incurring RTS/CTS overhead. This combination of techniques allows CHARM to respond quickly to dynamic channel changes. We implemented CHARM in the Madwifi driver for wireless cards using the Atheros chipset. Our evaluation both in the real world and on a controlled testbed shows that channel-aware rate selection can significantly outperform probe-based rate adaptation, especially over dynamic channels.

Distributed hash tables (DHTs) provide a scalable and robust building block for content discovery in distributed applications such as Peer-to-Peer (P2P) systems. However, the basic DHT put/get API only supports simple exact queries. In this paper, we present a DHT-based system that efficiently supports similarity queries on multidimensional datasets. Our system embeds a logical kd-tree into the DHT's identifier space to form a distributed indexing structure, the distributed kd-tree (DKDT). We avoid creating bottlenecks, which are typical in tree- based systems, by relying on fully distributed protocols for tree management and data registrations and queries. We propose tree compressing and node shrinking techniques to efficiently support applications with high dimensionality datasets. Simulation results using both synthetic and real data show the effectiveness of our system.

Robust video streaming over time-varying, error-prone wireless LANs (WLANs) poses many challenges. In this paper, we present a time-based opportunistic relay (TOR) scheme for high-quality video streaming over WLANs. The proposed scheme exploits path diversity in relaying packets with awareness of the time constraints of video data. Specifically, relay nodes continuously overhear and opportunistically forward packets to enhance end-to-end delivery rates. To relay a packet in a time-aware manner, a relay deadline is computed for each packet and used by relay nodes to determine whether they should relay or discard the packet. Simulation results show that TOR improves video quality by up to 9.76 dB.

Current rate selection algorithms are dominated by probe-based approaches that search for the best transmission rate using trial-and-error. When operating over a dynamic channel, probe-based techniques can perform poorly since they inefficiently search for the moving target presented by the constantly changing channel. We have developed a channel-aware rate adaptation algorithm - CHARM - that responds quickly to dynamic channel changes, and significantly outperforms probe-based algorithms in many instances. Unlike previous approaches, CHARM leverages channel reciprocity to obtain channel information without incurring RTS/CTS overhead. Our work shows that channel-aware rate selection is viable, and can significantly outperform probe-based rate adaptation over both static and dynamic channels.

In this paper we present the design and implementation of a space service that gives pervasive computing applications both a hierarchical and coordinate-based view of physical space. The space service supports both indoor and outdoor campus environments and, besides relational and coordinate space information, it also gives information on how spaces are connected and relevant properties of spaces. We also evaluate the effectiveness of our design using a set of diverse applications, including finding the best "walking" path between two locations, giving walking directions to people, WiFi-based localization, and drawing and annotating maps. We have found that all of our applications depend on both the hierarchical and coordinate views offered by the service, suggesting that the hybrid space model is a very powerful paradigm.

Networking researchers have long faced a fundamental tension between the experimental realism of wireless testbeds on one hand, and the control and repeatability of simulation on the other hand. To overcome the stark tradeoff of these traditional alternatives, we have developed a wireless network emulator that enables both realistic and repeatable experimentation at network scale. A critical component in this emulator is the RF front end that converts RF signals to lower frequencies - where they can be digitized - and vice versa. We discuss the unique requirements that physical layer network emulation demands of an RF front end. We present a design that meets these demands and demonstrate its performance via measurements.

Abstract Video streaming over time‐varying, error‐prone wireless LANs (WLANs) poses many challenges. One problem is that WLANs are designed without awareness of the characteristics of application data, which causes performance degradation in a noisy or congested environment. In this paper, we propose a time‐based adaptive retry (TAR) mechanism for MPEG‐like video streaming over 802.11 wireless networks. TAR dynamically determines whether to send or discard a packet based on itsretransmission deadlineinstead of adopting a static retry limit uniformly over all the packets. Our approach can adapt the retry limit for each individual packet, thus providing indirect unequal error protection over different types of video frames. Analytical and simulation results show that TAR significantly improves video quality and saves channel bandwidth. We also describe a preliminary software‐based implementation of TAR and use it to demonstrate the practicality of the proposed approach. Copyright © 2007 John Wiley & Sons, Ltd.

The increasing availability of information about people's context makes it possible to deploy context-sensitive services, where access to resources provided or managed by a service is limited depending on a person's context. For example, a location-based service can require an individual to be at a particular location in order to let the individual use a printer or learn her friends' location. However, constraining access to a resource based on confidential information about a person's context could result in privacy violations. For instance, if access is constrained based on a person's location, granting or rejecting access will provide information about this person's location and could violate the person's privacy. We introduce an access-control algorithm that avoids privacy violations caused by context-sensitive services. Our algorithm exploits the concepts of access-rights graphs, which represent all the information that needs to be collected in order to make a context-sensitive access decision. Moreover, we introduce hidden constraints, which keep some of this information secret and thus allow for more flexible access control. We present a distributed, certificate-based access-control architecture for context-sensitive services that avoids privacy violations, a sample implementation, and a performance evaluation

With the advances in middleware and Web services technologies, network sendees are evolving from simple client-sender applications to self-configuring services that can compose primitive components distributed in the Internet into a value-added service configuration that provides rich functionalities to users. A resulting research problem is how to continuously adapt such composite service configurations at run time in order to cope with the increasingly dynamic and heterogeneous network environments and computing platforms. In this paper, we propose a self-adaptation architecture that allows service developers to specify their service-specific adaptation knowledge as "externalized" adaptation strategies. These adaptation strategies are used by a general, shared adaptation framework to perform run-time adaptation operations that automatically incorporate service-specific knowledge. In addition to the strategies, we also identify another aspect of adaptation knowledge that is not addressed by previous solutions: adaptation coordination. Our framework provides integrated support for the specification and execution of both aspects of developers' adaptation knowledge.

This paper examines how computation can be mapped across the nodes of a distributed search system to effectively utilize available resources. We specifically address computationally intensive search of complex data, such as content-based retrieval of digital images or sounds, where sophisticated algorithms must be evaluated on the objects of interest. Since these problems require significant computation, we distribute the search over a collection of compute nodes, such as active storage devices, intermediate processors and host computers. A key challenge with mapping the desired computation to the available resources is that the most efficient distribution depends on several factors: relative power and number of compute nodes; network bandwidth between the compute nodes; the cost of evaluating query predicates; and the selectivity of the given query. This wide range of variables renders manual partitioning of the computation infeasible, particularly since some of the parameters (e.g., available network bandwidth) can change during the course of a search. This paper proposes several techniques for dynamic partitioning of computation, and demonstrates that they can significantly improve efficiency for distributed search applications.

Robust video streaming over error-prone wireless LANs (WLANs) poses many challenges. In this paper, we propose a timestamp-based content-aware adaptive retry (CAR) mechanism for MPEG video streaming over 802.11 WLANs, where the MAC dynamically determines whether to send or discard a packet based on its retransmission deadline. The retransmission deadline is assigned to each packet according to its temporal relationship and error propagation characteristics with respect to other video packets in the same GOP. The proposed scheme avoids late packets by eliminating the impact of random backoff deference and co-channel interference with proper initial delay introduced at the receiver. Simulation results show CAR significantly improves video quality and saves channel bandwidth.

Recent advances in Internet measurement tools have made it possible to locate bottleneck links that constrain the available bandwidth of Internet paths. In this paper, we provide a detailed study of Internet path bottlenecks. We focus on the following four aspects: the persistence of bottleneck location, the sharing of bottlenecks among destination clusters, the packet loss and queueing delay of bottleneck links, and the relationship with router and link properties, including router CPU load, router memory load, link traffic load, and link capacity. We find that 20% - 30% of the source-destination pairs in our measurement have a persistent bottleneck; fewer than 10% of the destinations in a prefix cluster share a bottleneck more than half of the time; 60% of the bottlenecks on lossy paths can be correlated with a loss point no more than 2 hops away; and bottlenecks can be clearly correlated with link load, while presenting no strong relationship with link capacity, router CPU and memory load.

Over the past few years, wireless networking technologies have made vast forays into our daily lives. Today, one can find 802.11 hardware and other personal wireless technology employed at homes, shopping malls, coffee shops and airports. Present-day wireless network deployments bear two important properties: they are unplanned, with most access points (APs) deployed by users in a spontaneous manner, resulting in highly variable AP densities; and they are unmanaged, since manually configuring and managing a wireless network is very complicated. We refer to such wireless deployments as being chaotic.In this paper, we present a study of the impact of interference in chaotic 802.11 deployments on end-client performance. First, using large-scale measurement data from several cities, we show that it is not uncommon to have tens of APs deployed in close proximity of each other. Moreover, most APs are not configured to minimize interference with their neighbors. We then perform trace-driven simulations to show that the performance of end-clients could suffer significantly in chaotic deployments. We argue that end-client experience could be significantly improved by making chaotic wireless networks self-managing. We design and evaluate automated power control and rate adaptation algorithms to minimize interference among neighboring APs, while ensuring robust end-client performance.

Pervasive computing environments offer a multitude of information services that provide potentially complex types of information. Therefore, when running access control for sensitive information, these environments need to take relationships between information into account. Other approaches to relationship-aware access control (e.g., based on semantic Web rule engines) are often expensive and based on a centralized design. In this paper, we identify three types of information relationships (bundling-based, combination-based, and granularity-based) that are common and important in pervasive computing, and we integrate support for them in distributed, certificate-based access control architecture. In our approach, access control is fully distributed while sophisticated rule engines can still be used to deal with more complex access control cases. To demonstrate the feasibility of our design, we give a complexity analysis of the architecture and a performance analysis of a prototype implementation

Access control to confidential information in pervasive computing environments is challenging for multiple reasons: First, a client requesting access might not know which access rights are necessary in order to be granted access to the requested information. Second, access control must support flexible access rights that include context-sensitive constraints. Third, pervasive computing environments consist of a multitude of information services, which makes simple management of access rights essential. We discuss the shortcomings of existing access-control schemes that rely on either clients presenting a proof of access to a service or services encrypting information before handing the information over to a client. We propose a proofbased access-control architecture that employs hierarchical identity-based encryption in order to enable services to inform clients of the required proof of access in a covert way, without leaking information. Furthermore, we introduce an encryption-based access-control architecture that exploits hierarchical identity-based encryption in order to deal with multiple, hierarchical constraints on access rights. We present an example implementation of our proposed architectures and discuss the performance of this implementation.

Embedded systems encompass a wide range of applications, technologies, and disciplines, necessitating a broad approach to education. We describe embedded system coursework during the first 4 years of university education (the U.S. undergraduate level). Embedded application curriculum areas include: small and single-microcontroller applications, control systems, distributed embedded control, system-on-chip, networking, embedded PCs, critical systems, robotics, computer peripherals, wireless data systems, signal processing, and command and control. Additional cross-cutting skills that are important to embedded system designers include: security, dependability, energy-aware computing, software/systems engineering, real-time computing, and human--computer interaction. We describe lessons learned from teaching courses in many of these areas, as well as general skills taught and approaches used, including a heavy emphasis on course projects to teach system skills.

Ubiquitous computing uses a variety of information for which access needs to be controlled. For instance, a person's current location is a sensitive piece of information that only authorized entities should be able to learn. Several challenges arise in the specification and implementation of policies controlling access to location information. For example, there can be multiple sources of location information. The sources can be within different administrative domains, which might allow different entities to specify policies, and policies need to be flexible. We address these issues in our design of a distributed access control mechanism for a people location system. Our design encodes policies as digital certificates, which enables decentralized storage of policies. We also present an algorithm for the discovery of distributed certificates. Furthermore, we discuss several privacy issues and show how our design addresses them. To show feasibility of our design, we built an example implementation based on SPKI/SDSI certificates. Using measurements, we quantify the influence of access control on query processing time. We also discuss trade-offs between RSA-based and DSA-based signature schemes for digital certificates.

In recent years, distributed hash tables (DHTs) have been proposed as a fundamental building block for large scale distributed applications. Important functionalities such as searching have been added to the DHTs basic lookup capability. However, supporting range queries efficiently remains a difficult problem. We describe an adaptive mechanism that relies on a logical tree data structure, the range search tree (RST), to support range queries efficiently. Nodes in the RST automatically group registrations based on their values. Queries are decomposed into a small number of sub-queries for efficient resolution. The system dynamically optimizes itself to minimize the registration and query cost based on observed load. The system is fully distributed and avoids bottleneck problems encountered in traditional tree-based systems. Extensive simulation results validate the effectiveness of the system.

Ubiquitous computing uses a variety of information for which access needs to be controlled. For instance, a person's current location is a sensitive piece of information, which only authorized entities should be able to learn. Several challenges arise in the specification and implementation of policies controlling access to location information. For example, there can be multiple sources of location information, the sources can be within different administrative domains, different administrative domains might allow different entities to specify policies, and policies need to be flexible. We address these issues in our design of an access control mechanism for a people location system. Our design encodes policies as digital certificates. We present an example implementation based on SPKI/SDSI certificates. Using measurements, we quantify the influence of access control on query processing time. We also discuss trade-offs between RSA-based and DSA-based signature schemes for digital certificates.

The ability to locate network bottlenecks along end-to-end paths on the Internet is of great interest to both network operators and researchers. For example, knowing where bottleneck links are, network operators can apply traffic engineering either at the interdomain or intradomain level to improve routing. Existing tools either fail to identify the location of bottlenecks, or generate a large amount of probing packets. In addition, they often require access to both end points. In this paper we present Pathneck, a tool that allows end users to efficiently and accurately locate the bottleneck link on an Internet path. Pathneck is based on a novel probing technique called Recursive Packet Train (RPT) and does not require access to the destination. We evaluate Pathneck using wide area Internet experiments and trace-driven emulation. In addition, we present the results of an extensive study on bottlenecks in the Internet using carefully selected, geographically diverse probing sources and destinations. We found that Pathneck can successfully detect bottlenecks for almost 80% of the Internet paths we probed. We also report our success in using the bottleneck location and bandwidth bounds provided by Pathneck to infer bottlenecks and to avoid bottlenecks in multihoming and overlay routing.

A common approach to adding self-management capabilities to a system is to provide one or more external control modules, whose responsibility is to monitor system behavior, and adapt the system at run time to achieve various goals (configure the system, improve performance, recover from faults, etc.). An important problem arises when there is more than one such self-management module: how can one make sure that they are composed to provide consistent and complementary benefits? In this paper we describe a solution that introduces a self-management coordination architecture and infrastructure to support such composition. We focus on the problem of coordinating self-configuring and self-healing capabilities, particularly with respect to global configuration and incremental repair. We illustrate the approach in the context of a self-managing video teleconference system that composes two preexisting adaptation modules to achieve synergistic benefits of both.

A content discovery system (CDS) allows nodes in the system to discover contents published by some other nodes in the system. Existing CDS systems have difficulties in achieving both scalability and rich functionality. We present the design and evaluation of a distributed and scalable CDS. Our system uses rendezvous points (RPs) for content registration and query resolution, and can accommodate frequent updates from dynamic contents. Contents stored in our system can be searched via subset matching. We propose a novel mechanism that uses load balancing matrices (LBMs) to balance dynamically both registration and query load across nodes in the system to maintain high system throughput even under skewed load. Our system utilizes existing distributed hash table (DHT) mechanisms for CDS overlay network management and routing. We validate our system's scalability and load balancing properties using extensive simulation.

A network that provides not only connectivity but also computational resources to application flows will enable a new array of network services. For example, applications that require content adaptation can be deployed more easily in a network that provides integrated communication and computational resources. In this paper, we study the problem of finding a path for a flow that has both computational and bandwidth constraints. We present a distributed load-resistive routing algorithm that generates precomputed routing information and optimizes the routing decisions for both applications and computational and communication resource providers. We show through simulations that our distributed approach performs comparably to a centralized algorithm and is more resilient to longer routing update intervals.

A large portion of data exchanged in today's peer-to-peer (P2P) networks consists of music stored as MP3 compressed audio. Existing P2P systems typically are not scalable and only support primitive methods for the searching of music files, e.g., by looking up exact filenames or using simple metadata information such as artist or album name. In this paper, we present the design and evaluation of a scalable P2P system that uses rendezvous points (RPs) for music file registration and query resolution, and supports content-based music information retrieval (MIR) of audio signals.

TCP slow start exponentially increases the congestion window size to detect the proper congestion window for a network path. This often results in significant packet loss, while breaking off slow start using a limited slow start threshold may lead to an overly conservative congestion window size. This problem is especially severe in high speed networks. In this paper we present a new TCP startup algorithm, called paced start, that incorporates an available bandwidth probing technique into the TCP startup algorithm. Paced start is based on the observation that when we view the TCP startup sequence as a sequence of packet trains, the difference between the data packet spacing and the acknowledgement spacing can yield valuable information about the available bandwidth. Slow start ignores this information, while paced start uses it to quickly estimate the proper congestion window for the path. For most flows. Paced Start transitions into congestion avoidance mode faster than Slow Start, has a significantly lower packet loss rate, and avoids the timeout that is often associated with slow start. This paper describes the paced start algorithm and uses simulation and real system experiments to characterize its properties.

Pervasive computing applications are increasingly leveraging contextual information from several sources to provide users with behavior appropriate to the environment in which they reside. If these sources of contextual information are used and deployed in an ad hoc manner however they may provide overlapping functionality, fail to provide needed functionality, and require the use of inconsistent interfaces by applications. To overcome these problems, we introduce a contextual information service that provides applications with contextual information via a virtual database. Unlike previous efforts, our service provides applications a consistent, lightweight, and powerful mechanism for obtaining contextual information, and includes explicit support for the on demand computation of contextual information. We show, via example applications and a contextual information service prototype that we have implemented, how this approach can be used to allow proactive applications to adapt their behavior to match a user's current environment.

A difficult problem in networking courses is to find hands-on projects that have the right balance between the level of realism and complexity. This is especially true for projects that focus on the internal functionality of routers and other network devices. We developed a capstone course called "Network Design and Evaluation" that uses a network processor-based platform for networking projects. This platform is more realistic than traditional approaches based on software emulation environments or PC-based routers running Unix, but it is significantly less complex to work with than real commercial routers or even PC-based routers. We are currently teaching this course for the third year, and our experience has been extremely positive. Students enjoy the realism of the platform and not only learn a lot about the internal operation of the network, but also about network configuration and management.

The packet pair mechanism has been shown to be a reliable method to measure the bottleneck link capacity on a network path, but its use for measuring available bandwidth is more challenging. In this paper, we use modeling, measurements, and simulations to better characterize the interaction between probing packets and the competing network traffic. We first construct a simple model to understand how competing traffic changes the probing packet gap for a single-hop network. The gap model shows that the initial probing gap is a critical parameter when using packet pairs to estimate available bandwidth. Based on this insight, we present two available bandwidth measurement techniques, the initial gap increasing (IGI) method and the packet transmission rate (PTR) method. We use extensive Internet measurements to show that these techniques estimate available bandwidth faster than existing techniques such as Pathload, with comparable accuracy. Finally, using both Internet measurements and ns simulations, we explore how the measurement accuracy of active probing is affected by factors such as the probing packet size, the length of probing packet train, and the competing traffic on links other than the tight link.

Grid applications must increasingly self-adapt dynamically to changing environments. In most cases, adaptation has been implemented in an ad hoc fashion, on a per-application basis. This paper describes work which generalizes adaptation so that it can be used across applications by providing an adaptation framework. This framework uses a software architectural model of the system to analyze whether the application requires adaptation, and allows repairs to be written in the context of the architectural model and propagated to the running system. In this paper, we exemplify our framework by applying it to the domain of load-balancing a client-server system. We report on an experiment conducted using our framework, which illustrates that this approach maintains architectural requirements.

Abstract Wireless links differ from traditional ‘wired’ links in two ways that challenge the existing Internet. On wireless links packet loss or corruption due to transmission errors is not rare, which calls into question the standard Internet assumptions that transmission errors should be corrected by transport‐level protocols at end systems and that end‐to‐end packet loss typically indicates network congestion. Also, the severity and location‐ dependent nature of these errors calls into question the meaning of ‘fair’ scheduling, per‐flow quality of service, and even looser notions such as service level agreements, when applied to wireless links. An important question is whether the two unique problems posed by wireless links can be successfully addressed within the standard Internet architecture, as opposed to requiring new transport protocols designed specifically for wireless links or requiring wireless links to ‘fix up’ the operation of specific end‐to‐end protocols. We provide experimental evidence that a combination of protocol‐blind link‐level local error control, which lessens the damage, and error‐sensitive link scheduling, which ensures sensible outcomes in response to link capacity loss, provides a good operating environment while adhering to traditional Internet design practices. Copyright © 2001 John Wiley & Sons, Ltd.

Remos provides resource information to distributed applications. Its design goals of scalability, flexibility, and portability are achieved through an architecture that allows components to be positioned across the network, each collecting information about its local network. To collect information from different types of networks and from hosts on those networks, Remos provides several collectors that use different technologies, such as SNMP or benchmarking. By matching the appropriate collector to each particular network environment and by providing an architecture for distributing the output of these collectors across all querying environments, Remos collects appropriately detailed information at each site and distributes this information where needed in a scalable manner. Prediction services are integrated at the user-level, allowing history-based data collected across the network to be used to generate the predictions needed by a particular user. Remos has been implemented and tested in a variety of networks and is in use in a number of different environments.

Increased network bandwidth is making desktop video conferencing an attractive application for an increasing number of computer users. Unfortunately, two competing standards for video conferencing signaling are in use, H.323 and SIP. In this paper we look at the interoperability between these two standards by developing a conferencing gateway that supports conferences involving both SIP and H.323 clients. By appropriately translating between H.323 and SIP operations, our prototype gateway supports basic multi-party video conferencing between NetMeeting (an H.323 client) and VIC (a SIP client) without modifications to the clients. However, our experiments also show that seamless interoperation would require changes to the client implementations and the standards.

The Internet is rapidly evolving from a network that provides basic best-effort communication service to an infrastructure capable of supporting complex value-added services. These services typically have multiple fluffs with interdependent resource requirements. These dependencies provide opportunities to share the same set of resources among related flows over time leading to significant resource gains. We call this type of sharing temporal resource sharing. Exploiting temporal sharing requires support in the signaling protocol that performs resource allocation for the related flows. We examine the problem of supporting temporal sharing in a signaling protocol. This paper makes the case that temporal sharing support must be designed to be extensible, so that service providers can define and implement new sharing behaviors without having to modify the signaling protocol. We motivate the need for an extensible design by showing that the range of possible temporal sharing behaviors is large and supporting the most general forms of temporal sharing is computationally expensive. We then present a design for extensible signaling support for temporal sharing. We have implemented the temporal sharing design presented in this paper in the Beagle signaling protocol. We present an evaluation of the Beagle design and contrast it with other signaling protocols like RSVP and Tenet-2.

While packet scheduling for wired links is a maturing area, scheduling of wireless links is less mature. A fundamental difference between wired and wireless links is that wireless media can exhibit substantial rates of link errors, resulting in significant and unpredictable loss of link capacity. This capacity loss results in a special challenge for wireless schedulers. For example, a weighted fair queue (WFQ) scheduler assumes an error free link and specifies how flows should share the link capacity. However, this specification is not sufficient to determine the correct outcome when link capacity is sharply reduced, because flows that have been allocated the same weights may differ greatly in their ability to tolerate throughput loss. In this paper, we first describe the wireless scheduling challenge in terms of an effort-outcome disconnection. Next we propose a novel notion of fairness for wireless links, effort-limited fairness (ELF), which extends WFQ via dynamic weight adjustments. ELF guarantees that all flows experiencing an error rate below a per flow threshold receive their expected service, defined as a specified rate for reserved flows or a specified share of best-effort capacity for best effort flows. After motivating and defining ELF, we present a practical approximation algorithm, which we evaluate through both trace-driven simulation and measurement of a prototype wireless radio network based on the WaveLAN physical layer.

Network-aware applications, i.e., applications that adapt to network conditions in an application-specific way, need both static and dynamic information about the network to be able to adapt intelligently to network conditions. The CMU Remos interface gives applications access to a wide range of information in a network-independent fashion. Remos uses a logical topology to capture the network information that is relevant to applications in a concise way. However, collecting this information efficiently is challenging for several reasons: networks use diverse technologies and can be very large (Internet); applications need diverse information; and network managers might have concerns about leaking confidential information. In this paper we present an architecture for a hierarchical collector of network information. The decentralized architecture relies on data collectors that collect information on individual subnets; data collectors can collect information in manner that is appropriate for that subnet and can control the distribution of the information. For application queries that involve multiple subnets, we use a set of master collectors to partition requests and distribute subrequests to individual data collectors and to combine the results. Collectors cache recent network information to improve efficiency and responsiveness. This paper presents and justifies the collector architecture, describes a prototype implementation, and presents preliminary measurements characterizing its operation.

There is an emerging class of multi-party, multimedia, multi-flow applications that have a high-level structure that imposes dependencies between resource allocations for flows within the application. These applications are also capable of making intelligent decisions on how resource allocation should be controlled within the application. The development of such applications places new requirements on signaling protocols. This paper outlines these new requirements, discusses ways in which they can be supported and presents the design and implementation of an experimental signaling protocol that supports these requirements. This paper makes the case that for these structured applications, there is an advantage to allocating the resources in an integrated fashion, i.e. computation, storage and communication resources for all the flows are allocated at the same time in a coordinated fashion. The concept of a virtual mesh is introduced as a key abstraction that encapsulates the set of resources that are allocated and managed in an integrated fashion to meet the needs of applications. The paper presents two mesh setup algorithms and a performance evaluation comparing them. Temporal resource sharing within the virtual mesh is discussed in detail and signaling support for temporal sharing at setup and runtime is examined. It is important to characterize temporal sharing since it can significantly reduce the resource requirements for applications. We have implemented the Beagle signaling protocol that supports this integrated resource management model. Beagle representations and mechanisms for mesh setup and temporal sharing are described and a prototype implementation is presented.

In an integrated services network, resources are shared by multiple traffic classes. Service classes that deliver quality-of-service (QoS) to applications have priority over others that do not. In such a multi-service network, routing decisions for high priority QoS traffic will affect what resources are available for lower priority traffic: poor route selection can result in congestion for, or even starvation of, lower priority traffic. Whereas many studies have focused on routing algorithms that optimize the network throughput for individual service classes, little effort has been devoted to routing algorithms that address inter-class resource sharing. In this paper, we propose a routing algorithm that allows dynamic sharing of link resources among multiple traffic classes. The algorithm is based on the concept of "virtual residual bandwidth", which is derived from the link residual bandwidth by taking the congestion condition of low priority traffic into account. By using the virtual residual bandwidth in the link cost function for QoS sessions, we discourage QoS sessions from using links that are heavily loaded with low priority traffic. Our approach is simple in the sense that besides changing the link cost function, no other changes to the routing algorithms for individual service classes are required. An extensive simulation study shows that when the traffic load is unevenly distributed, significant performance improvements can be achieved for low priority traffic without sacrificing performance for high priority traffic. The result demonstrates that QoS routing is important even when the QoS traffic load is light and call blocking is not an issue.

Development of portable network-aware applications demands an interface to the network that allows an application to obtain information about its execution environment. The paper motivates and describes the design of Remos, an API that allows network-aware applications to obtain relevant information. The major challenges in defining a uniform interface are network heterogeneity, diversity in traffic requirements, variability of the information, and resource sharing in the network. Remos addresses these issues with two abstraction levels, explicit management of resource sharing, and statistical measurements. The flows abstraction captures the communication between nodes, and the topologies abstraction provides a logical view of network connectivity. Remos measurements are made at network level, and therefore information to manage sharing of resources is available. Remos is designed to deliver best effort information to applications, and it explicitly adds statistical reliability and variability measures to the core information. The paper also presents preliminary results and experience with a prototype Remos implementation for a high speed IP based network testbed.

The Internet is rapidly changing from a set of wires and switches that carry packets into a sophisticated infrastructure that delivers a set of complex value-added services to end users. Services can range from bit transport all the way up to distributed value-added services like video teleconferencing, data mining, and distributed interactive simulations. Before such services can be supported in a general and dynamic manner we have to develop appropriate resource management mechanisms. These resource management mechanisms must make it possible to identify and allocate resources that meet service or application requirements, support both isolation and controlled dynamic sharing of resources across organizations sharing physical resources, and be customizable so services and applications can tailor resource usage to optimize their performance. The Darwin project is developing a set of customizable resource management mechanisms that support value-added services, In this paper we present these mechanisms, describe their implementation in a prototype system, and describe the results of a series of proof-of-concept experiments.

Wireless links can exhibit high error rates due to attenuation, fading, or interfering active radiation sources. To make matters worse, error rates can be highly variable due to changes in the wireless environment. Researchers and developers have explored a wide range of solutions to optimize communication in this difficult error environment, including traditional end-to-end solutions, link-layer solutions, and solutions involving layer four processing inside the network. A significant challenge is ensuring that systems with multiple layers of error control avoid compromising performance by duplication of effort. We argue and demonstrate that protocol-independent link-level local error control can achieve high communication efficiency even in a highly variable error environment, that adaptation is important to achieve this efficiency, and that inter-layer coexistence is achievable. The logical link control layer of our WaveLAN-based experimental LAN includes three error control mechanisms: local retransmission, adaptive packet shrinking, and adaptive error coding. Measurements generated on a variety of network topologies and trace-based error environments demonstrate the TCP performance improvements and good coexistence with TCP's end-to-end retransmission strategy.

Compared to kernel-level servers, user-level ones can be debugged and maintained more easily and safely, but traditionally have had much worse performance. We describe a novel I/O-oriented inter-proces communication (IPC) facility that combines the emulated copy data passing scheme for monolithic systems with new copy avoidance techniques for microkernel systems. Unlike previous optimizations, I/O-oriented IPC does not require changes in existing user applications or complex restructuring of servers; it offers an API with copy semantics and allows the same servers to be installed at kernel or user level. In end-to-end experiments on an ATM network at 512 Mbps, I/O-oriented IPC gave user-level protocol servers performance approaching that of kernel-level ones. Performance differences scaled roughly inversely to the processor's SPECint95 rating, projecting fast further improvement.

Data copying and checksumming are the most expensive operations on hosts performing high-bandwidth network I/O over a high-speed network. Under some conditions, outboard buffering and checksumming can eliminate accesses to the data, thus making communication less expensive and faster. One of the scenarios in which outboard buffering and checksumming pays off is the common case of applications accessing the network using the Berkeley sockets interface and the Internet protocol stack. In this paper, we describe the host software for a host interface with outboard buffering and checksumming support. The platform used is DEC Alpha workstations with a Turbochannel I/O bus and running the DEC OSF/1 operating system. Our implementation does not only achieve ‘single copy’ communication for applications that use sockets, but it also interoperates efficiently with in-kernel applications and other network devices. Measurements show that for large reads and writes the single-copy path through the stack is five to seven times more efficient than the traditional implementation. We also present a detailed analysis of the measurements using a simple I/O model. © 1998 John Wiley & Sons, Ltd.

Transmission of multimedia streams imposes a minimum-bandwidth requirement on the path being used to ensure end-to-end Quality-of-Service (QoS) guarantees. While any shortest-path algorithm can be used to select a feasible path, additional constraints that limit resource consumption and balance the network load are needed to achieve efficient resource utilization. We present a systematic evaluation of four routing algorithms that offer different tradeoffs between limiting the path hop count and balancing the network load. Our evaluation considers not only the call blocking rate but also the fairness to requests for different bandwidths, robustness to inaccurate routing information, and sensitivity to the routing information update frequency. It evaluates not only the performance of these algorithms for the sessions with bandwidth guarantees, but also their impact on the lower priority best-effort sessions. Our results show that a routing algorithm that gives preference to limiting the hop count performs better when the network load is heavy, while an algorithm that gives preference to balancing the network load performs slightly better when the network load is light. We also show that the performance of using pre-computed paths with a few discrete bandwidth requests is comparable to that of computing paths on-demand, which implies feasibility of class-based routing. We observe that the routing information update interval can be set reasonably large to reduce routing overhead without sacrificing the overall performance, although an increased number of sessions can be misrouted.

Data copying can be a bottleneck in end-to-end communication over high-speed networks. Emulated copy is an alternative I/O data passing scheme that preserves the API and integrity guarantees of copying but avoids the latter using virtual memory manipulations - transient output copy-on-write (TCOW), input alignment, and page swapping. We characterize and evaluate the support necessary in network adapters for emulated copy in checksummed, multiple-packet communication. Our experiments on an ATM network show that: (1) emulated copy gives performance better than that of copying even without hardware checksumming support; (2) TCOW improves multiple-packet output performance without any hardware support or changes in applications; (3) page swapping provides additional similar improvements on multiple-packet input if there is input alignment, which requires either hardware support (early-demultiplexed/system-aligned buffering) or changes in applications (pooled/application-aligned buffering); and (4) The performance of application-aligned buffering is largely unaffected by header/data splitting, a common optimization. We propose a new optimization, buffer snap-off, that extends system-aligned buffering to the general case of arbitrary, unmatched data transfer and application input buffer lengths.

A critical issue in the design of fast packet-switch-based networks is the avoidance of data loss due to congestion. In the context of ATM networks, many link-level congestion control mechanisms for ABR traffic have been proposed and simulated, and a few have been implemented. This paper presents, for the first time, an experimental comparison of several such mechanisms. To drive this comparison, we have developed a set of benchmarks that provide a quantitative characterization of congestion control performance. We also describe a simple but novel technique, the "virtual port card", for implementing both non-native switch behavior and long link delays in ATM networks. This technique allows us to compare a wide range of mechanisms on a single flexible hardware platform. Our measurements show that while several of the ATM flow control mechanisms can eliminate cell loss, there are still several unresolved problems. First, for some traffic scenarios we see very poor application-level performance, especially in the presence of bursty traffic. Second, the performance is very sensitive to the flow control parameters and identifying an appropriate set of parameters is difficult since it depends heavily on the traffic conditions.

Clusters of workstations are emerging as an important architecture. Programming tools that aid in distributing applications on workstation clusters must address problems of mapping the application, heterogeneity and maximizing system utilization in the presence of varying resource availability. Both computation and communication capabilities may vary with time due to other applications competing for resources, so dynamic load balancing is a key requirement. For greatest benefit, the tool must support a relatively wide class of applications running on clusters with a range of computation and communication capabilities. We have developed a system that supports dynamic load balancing of distributed applications consisting of parallelized DOALL and DOACROSS loops. The focus of the paper is on how the system automatically determines key load balancing parameters using run-time information and information provided by programming tools such as a parallelizing compiler. The parameters discussed are the grain size of the application, the frequency of load balancing, and the parameters that control work movement. Our results are supported by measurements on an implementation for the Nectar system at Carnegie Mellon University and by simulation. © 1997 by John Wiley & Sons, Ltd.

Distributed-memory systems have traditionally had great difficulty performing network I/O at rates proportional to their computational power. The problem is that the network interface has to support network I/O for a supercomputer, using computational and memory bandwidth resources similar to those of a workstation. As a result, the network interface becomes a bottleneck. In this article we present an I/O architecture that addresses these problems and supports high-speed network I/O on distributed-memory systems. The key to good performance is to partition the work appropriately between the system and the network interface. Some communication tasks are performed on the distributed-memory parallel system, since it is more powerful and less likely to become a bottleneck than the network interface. Tasks that do not parallelize well are performed on the network interface, and hardware support is provided for the most time-critical operations. This architecture has been implemented for the iWarp distributed-memory system and has been used by a number of applications. We describe this implementaiton, present performance results, and use application examples to validated the main features of the I/O architecture.

Commodityworkstations connected by commodity networks are increasingly viewed as an economically viable alternative to tightly coupled multiprocessors.In recent years, many scientific computing applications have been able to make effective use of various types of workstation clusters.The main difference between workstation clusters and the more traditional, tightly coupled distributedmemory systems is communication performance.In this paper we present a host-network interface architecture that supports efficient remote memory writes across standard ATM networks, bringing performance closer to that of special-purpose, tightly coupled systems for a large class of applications.We show that minimal hardware support is required on the adaptor, and that the required features are very similar to those already needed on adaptors for high-speed networks.We also describe an implementation of this architecture, and present measurements of communication performance indicating its effect on the breadth of applications that can use a general purpose network m an effective way.

No abstract available.

There is general belief that networks based on wireless technologies have much higher error rates than those based on more traditional technologies such as optical fiber, coaxial cable, or twisted pair wiring. This difference has motivated research on new protocol suites specifically for wireless networks. While the error characteristics of wired networks have been well documented, less experimental data is available for wireless LANs.In this paper we report the results of a study characterizing the error environment provided by AT&T WaveLAN, a commercial product designed for constructing 2 Mb/s in-building wireless networks. We evaluated the effects of interfering radiation sources, and of attenuation due to distance and obstacles, on the packet loss rate and bit error rate. We found that under many conditions the error rate of this physical layer is comparable to that of wired links. We analyze the implications of our results on today's CSMA/CA based wireless LANs and on future pico-cellular shared-medium reservation-based wireless networks.

We study how to improve the throughput of high-bandwidth traffic such as large file transfers in a network where resources are fairly shared among connections. While it is possible to devise priority or reservation-based schemes that give high-bandwidth traffic preferential treatment at the expense of other connections, we focus on the use of routing algorithms that improve resource allocation while maintaining max-min fair share semantics. In our approach, routing is closely coupled with congestion control in the sense that congestion information, such as the rates allocated to existing connections, is used by the routing algorithm. To reduce the amount of routing information that must be distributed, an abstraction of the congestion information is introduced. Using an extensive set of simulation, we identify a link-cost or cost metric for "shortest-path" routing that performs uniformly better than the minimal-hop routing and shortest-widest path routing algorithms. To further improve throughput without reducing the fair share of single-path connections, we propose a novel prioritized multi-path routing algorithm in which low priority paths share the bandwidth left unused by higher priority paths. This leads to a conservative extension of max-min fairness called prioritized multi-level max-min fairness. Simulation results confirm the advantages of our multi-path routing algorithm.

Multicomputers built around a general network are an attractive architecture for a wide class of applications. The architecture provides many benefits compared with special-purpose approaches, including heterogeneity, reuse of application and system code, and sharing of resources. The architecture also poses new challenges to both computer system implementers and users. First, traditional local-area networks do not have enough bandwidth and create a communication bottleneck, thus seriously limiting the set of applications that can be run effectively. Second, programmers have to deal with large bodies of code distributed over a variety of architectures, and work in an environment where both the network and nodes are shared with other users. Our experience in the Nectar project shows that it is possible to overcome these problems. We show how networks based on high-speed crossbar switches and efficient protocol implementations can support high bandwidth and low latency communication while still enjoying the flexibility of general networks, and we use three applications to demonstrate that network-based multicomputers are a practical architecture. We also show how the network traffic generated by this new class of applications poses severe requirements for networks.

Among the many benefits of ATM networking are the potential for connections with negotiated quality-of-service (QoS) guarantees and application-specific data management at network endpoints. We describe the architecture of a PCI bus host adapter for OC-3 and OC-12 ATM, focusing on challenges in the areas of buffer management and flow control, since these are vital to realizing the bandwidth and QoS potential of ATM endpoint hosts.

We evaluate the impact of a gigabit network on the implementation of a distributed chemical process optimization application. The optimization problem is formulated as a stochastic Linear Assignment Problem and was solved using the Thinking Machines CM-2 (SIMD) and the Cray C-90 (vector) computers at PSC, and the Intel iWarp (MIMD) system at CMU, connected by the Gigabit Nectar testbed. We report our experience distributing the application across this heterogeneous set of systems and present measurements that show how the communication requirements of the application depend on the structure of the application. We use detailed traces to build an application performance model that can be used to estimate the elapsed time of the application for different computer system and network combinations. Our results show that the application benefits from the high-speed network, and that the need for high network throughput is increasing as computer systems get faster. We also observed that supporting high burst rates is critical, although structuring the application so that communication is overlapped with computation relaxes the bandwidth requirements.

Distributed-memory systems have traditionally had great difficulty performing network I/O at rates proportional to their computational power. The problem is that the network interface has to support network I/O for a supercomputer, using computational and memory bandwidth resources similar to those of a workstation. As a result, the network interface becomes a bottleneck. We implemented an architecture for network I/O for the iWarp system with the following two key characteristics: first, application-specific tasks are off-loaded from the network interface to the distributed-memory system, and second, these tasks are performed in close cooperation with the application. The network interface has been used by several applications for over a year. In this paper we describe the network interface software that manages the communication between the iWarp distributed-memory system and the network interface, we validate the main features of our network interface architecture based on application experience, and we discuss how this architecture can be used by other distributed-memory systems.

An important challenge in the area of distributed computing is to automate the selection of the parameters that control the distributed computation. A performance-critical parameter is the grain size of the computation, i.e., the interval between successive synchronization points in the application. This parameter is hard to select since it depends both on compile time (loop structure and data dependences, computational complexity) and run time components (speed of compute nodes and network). On networks of workstations that are shared with other users, the run-time parameters can change over time. As a result, it is also necessary to consider the interactions with dynamic load balancing, which is needed to achieve good performance in this environment. In this paper we present a method for automatically selecting the grain size of the computation consisting of nested DO loops. The method is based on close cooperation between the compiler and the runtime system. We evaluate the method using both simulation and measurements for an implementation on the Nectar multicomputer.

Data copying and checksumming are the most expensive operations when doing high-bandwidth network IO over a high-speed network. Under some conditions, outboard buffering and checksumming can eliminate accesses to the data, thus making communication less expensive and faster. One of the scenarios in which outboard buffering pays off is the common case of applications accessing the network using the Berkeley sockets interface and the Internet protocol stack. In this paper we describe the changes that were made to a BSD protocol stack to make use of a network adaptor that supports outboard buffering and checksumming. Our goal is not only to achieve "single copy" communication for application that use sockets, but to also have efficient communication for in-kernel applications and for applications using other networks. Performance measurements show that for large reads and writes the single-copy path through the stack is significantly more efficient than the original implementation.

Achieving high-speed network I/O on distributed memory systems is a hard problem because their architectures are, in general, ill-suited for communication with the external world One of the problems is that messages are distributed over the private memories of the distributed memory system. This can result in poor performance since communication includes a complex scatter/gather operation. This paper presents a strategy in which the task of creating large contiguous messages is performed on the distributed-memory system, thus minimizing the overhead on the network interface. The performance results for an implementation of this strategy for an iWarp system with a HIPPI interface board are presented.>

Existing parallelizing compilers are targeted towards parallel architectures where all processors are dedicated to a single application. However a new type of parallel system has become available in the form of high performance workstations connected by high speed networks. Such systems pose new problems for compilers because the available processing power on each workstation may change with time due to other tasks competing for resources. We argue that it is possible for a parallelizing compiler to generate code that can dynamically shift portions of the application's workload between processors to improve performance. We have implemented a run-time system that supports automatically generated programs with dynamic load balancing. We describe this system and present performance measurements. We also describe the compiler functionality needed to generate parallel programs with dynamic load balancing.>

We consider the problem of high-speed I/O for a single application running on multiple nodes of a distributed-memory parallel computer. Our model is that the parallel system is connected to an I/O system that provides the interface between the internal connections of the parallel system and one or more external connections, such as HIPPI links. We identify two primary operations for this I/O system: scattering data from a high speed link across several lower speed links and gathering data from multiple links onto a single high speed link. We show that these core operations are the basis of the I/O system, independent of the relative speeds of the internal and external connections.

Achieving high-speed network I/O on distributed-memory systems is difficult because their architecture is in general ill-suited for communication processing. Some of the common problems are: inability to do protocol processing, inefficient handling of data distribution, and poor management of the I/O. The authors present an I/O architecture that addresses these problems and supports high-speed network I/O on distributed-memory systems. The key to good performance is to partition the work appropriately between the system and the network interface. The authors perform some communication tasks on the distributed-memory parallel system since it is more powerful, and less likely to become a bottleneck than the network interface. Tasks that do not parallelize well are performed on the network interface and hardware support is provided for the most time-critical operations. They emphasize the use of simple I/O mechanisms that can be used by programming tools that map applications on the distributed-memory system to implement efficient I/O for the class of applications they support. This architecture has been implemented for the iWarp distributed-memory system. The authors describe this implementation and present performance results.>

Switched LANs are become more widely used because they can provide a higher bandwidth than LANs based on shared media. Examples of packet switched LANs include HIPPI, and switched FDDI and Ethernet. A number of studies have evaluated the performance of HIPPI networks, making simplifying assumptions about both the network and the traffic load. In this paper we present the results of a simulation study of circuit-switched LANs such as HIPPI using more realistic models for the system and the traffic. We observe changes in throughput as high as a factor of ten when we change the system and traffic parameters. We also show how packet scheduling can be used to improve performance in some cases.>

The goal of load balancing is to assign to each node a number of tasks proportional to its performance. On distributed-memory machines, it is important to take data dependencies into account when distributing tasks, since they have a big impact on the communication requirements of the distributed application. The authors present a load balancing architecture that can deal with applications with heterogeneous tasks. The idea is to provide a set of load balancers that are effective for different types of homogeneous tasks, and to allow users to combine these load balancers for applications with heterogeneous tasks. This architecture was implemented on the Nectar multicomputer and performance results are presented for several applications with homogeneous and heterogeneous tasks.>

Abstract Reducing latency has traditionally been the biggest For multicomputer applications, the most important challenge. Dedicated multicomputers with special-purpose performance parameters of a network is the latency for short messages. In this paper we present an analysis of communication latency using measurement of the Nectar system. Nectar is a high-performance interconnects such as the Intel Touchstone system have latencies below 100 microseconds [4], while latencies between Unix workstations communicating over general multicomputer built around a high-bandwidth networks are typically one order of magnitude higher. The crosspoint network. Nodes are connected to the Nectar latency between two Sun4/330 running Sun OS 4.1 is for network using network coprocessors that are primarily responsible the protocol processing, but that can also execute application code. This architecture allows us to analyze message latency both between workstations example about 800 microseconds. One reason for the higher latency is the difference in communication medium. The interconnection networks used by dedicated with an outboard protocol engine and between multicomputers only have to cover a few meters and lightweight nodes with a minimal runtime system and a guarantee data integrity in hardware, while general fast, simple network interface (the coprocessors). We study how much context switching, buffer management and protocol processing contribute to the communication latency and we discuss how the latency is influenced by the protocol implementation. networks have to cover much larger distances and introduce errors in the data stream with non-zero probability. The communication protocols that recover from these errors introduce overhead, thus adding to the latency. This overhead however is, or should be, of the 1.

Programmers who want to do network computing face several challenges: the network and attached systems are shared resources with an unpredictable behavior and network communication primitives are often hard to use. The programming environment developed for the Nectar system addresses both problems. It provides simple and efficient communication primitives, and an efficient monitoring kernel that allows both programmers and programming tools to monitor the behavior of the program in the dynamic network environment. Experience shows that monitoring the progress of applications interactively is both desirable and practical.>

Several large applicationshave been paralleli,zed on Nectar, a network-based multicomputer recently developed by Carnegie Mellon.These applications were previously either too large or too complex to be easily implemented on distributed memory parallel systems.Parallelizing these applications was made possible by the cooperative use of many existing general-purpose computers over high-speed networks, and by an implementation methodology based on a clean separation between applicatiionspecific and system-specific code.We illustrate these points using our experience with parallelizing three real-world applications.The success in these applications clearly points out a new direction in parallel processing.The Nectar system [1] developed by Carnegie Mellon is intended to provide general support for parallelizing large applications.The system is a multicomputer built around a high-speed network.The use of existing general-purpose computers as its nodes and the highbandwidth and low-latency network makes the system inherently suited for large applications.The system has allowed us to parallelize applications that were previously either tm complex or too communicationintensive to be suited for parallel processing.This paper describes the Nectar implementation of three applications: (1) COSMOS [4], a switch-level circuit simulator developed by Randy Bryant and his associates at Carnegie Mellon; (2) NOODLES [7], a solid modeling package developed by Professor

Multicomputersbuiltaround a general network are now a viable alternative to multicomputersbased on a system-speci~c interconnect because of architectural improvements in two areas.First, the host-network interface overhead can be minimized by reducing copy operations and host interrupts.Second, the network can provide high bandwidth and low latency by using high-speed crossbar switches and efficient protocol implementations.While still enjoying thejexibility of general networks, the resulting network-based multicomputers achieve high performance for typical multicomputer applications that use system-specijic interconnects.We have developed a network-based mtdticomputer called Nectar that supports these claims.

We have built a high-speed local-area network called Nectar that uses programmable communication processors as host interfaces. In contrast to most protocol engines, our communication processors have a flexible runtime system that supports multiple transport protocols as well as application-specific activities. In particular, we have implemented the TCP/IP protocol suite and Nectar-specific communication protocols on the communication processor. The Nectar network currently has 25 hosts and has been in use for over a year.

Abstract We extend the well‐known interval analysis method so that it can be used to gather global flow information for individual array elements. Data dependences between all array accesses in different basic blocks, different iterations of the same loop, and across different loops are computed and represented as labelled arcs in a program flow graph. This approach results in a uniform treatment of scalars and arrays in the compiler and builds a systematic basis from which the compiler can perform numerous global optimizations. This global dataflow analysis is performed as a separate phase in the compiler. This phase only gathers the global relationships between different accesses to a variable, yet the use of this information is left to the code generator. This organization substantially simplifies the engineering of an optimizing compiler and separates the back end of the compiler (e.g. code generator and register allocator) from the flow analysis part. The global dataflow analysis algorithm described in this paper has been implemented and used in an optimizing compiler for a processor with deep pipelines. This paper describes the algorithm and its compact implementation and evaluates it, both with respect to the accuracy of the information and to the compile‐time cost of obtaining and using it.

Nectar is a “network backplane” for use in heterogeneous multicomputers. The initial system consists of a star-shaped fiber-optic network with an aggregate bandwidth of 1.6 gigabits/second and a switching latency of 700 nanoseconds. The system can be scaled up by connecting hundreds of these networks together.

The widespread use of reduced-instruction-set computers has generated a lot of interest in the tradeoff between the density of an instruction set and the size of the instruction cache. In this paper we present and justify a method that predicts the cache performance for a wide range of architectures, based on the miss rate for a single architecture. When we apply the method to a number of cache organizations we find that changes in code density can have a dramatic impact on memory traffic, but that modest improvements in code density do not reduce program execution time significantly in a well-balanced system.

Register allocation is an important optimization in many compilers, but with per-procedure register allocation, it is often not possible to make good use of a large register set. Procedure calls limit the improvement from global register allocation, since they force variables allocated to registers to be saved and restored. This limitation is more pronounced in LISP programs due to the higher frequency of procedure calls. An interprocedural register allocation algorithm is developed by simplifying a version of interprocedural graph coloring. The simplification corresponds to a bottom-up coloring of the interference graph. The scheme is evaluated using a number of LISP programs. The evaluation considers the scheme's limitations and compares these “software register windows” against the hardware register windows used in the Berkeley RISC and SPUR processors.

One of the major factors that distinguishes LISP from many other languages (Pascal, C, Fortran, etc.) is the need for run-time type checking. Run-time type checking is implemented by adding to each data object a tag that encodes type information. Tags must be compared for type compatibility, removed when using the data, and inserted when new data items are created. This tag manipulation, together with other work related to dynamic type checking and generic operations, constitutes a significant component of the execution time of LISP programs. This has led both to the development of LISP machines that support tag checking in hardware and to the avoidance of type checking by users running on stock hardware. To understand the role and necessity of special-purpose hardware for tag handling, we first measure the cost of type checking operations for a group of LISP programs. We then examine hardware and software implementations of tag operations and estimate the cost of tag handling with the different tag implementation schemes. The data shows that minimal levels of support provide most of the benefits, and that tag operations can be relatively inexpensive, even when no special hardware support is present.