In this paper, we use Wireshark packet-level traces to study the performance of the Zoom network application. Our work is motivated by several anecdotal reports of Zoom performance problems on our campus network during the Fall 2021 semester. Through the collection and analysis of Wireshark traces from different vantage points, we are able to pinpoint the root cause of the Zoom performance problems, which is a congested external Internet link for our campus network. We also identify several characteristics of the Zoom application that exacerbate its performance issues on congested and lossy networks, due to multi-layer protocol interactions.

In this paper, we develop a synthetic workload model for the Zoom network application based on empirical Zoom traffic measurements on a campus network. We then use this model in a simulation study of Zoom network traffic at the campus scale. The simulation results show that hybrid learning places a substantial load on the campus network, especially on the wireless network. Additional simulation experiments investigate the potential benefits of locally-hosted Zoom infrastructure, improved load balancing strategies for Zoom servers, and multicast delivery models for Zoom network traffic. The results show that the multicast approach offers the greatest potential benefit.

Shared micromobility is a rapidly growing transportation technology, with several companies establishing e-bike and e-scooter programs in cities all across the globe. In this paper, we use two years of empirical data on e-scooter usage from a pilot project in the City of Calgary to create a synthetic workload model of e-scooter traffic. Using this model, we develop a simulation environment to evaluate the impacts of different e-scooter management policies (e.g., fleet size, battery re-charging strategies, and urban parking infrastructure locations) on the efficacy of the e-scooter system. Our simulation results highlight the importance of proper site selection for parking areas and battery charging infrastructure.

Over the past decade, Instant Messaging (IM) apps have become an extremely popular tool for billions of people to communicate online. In this paper, we use a combination of active and passive measurement techniques to study one week of IM app traffic on a large campus edge network. Despite the challenges of end-to-end encryption, user privacy, NAT, DHCP, and high traffic volumes, we identify the key characteristics of four popular IM apps: Facebook Messenger, Google Hangouts, Snapchat, and WeChat. The main observations from our study indicate a rich ecosystem of IM apps, many of which exhibit strong diurnal patterns, complex user interactions, and heavy-tailed distributions for connection durations and transfer sizes. Collectively, these four IM apps contribute about 650 GB of daily traffic volume on our campus network.

Mobile video streaming is becoming increasingly popular. In this paper, we describe the design and implementation of a cross-platform measurement tool called MoVIE (Mobile Video Information Extraction) for video streaming on mobile devices. MoVIE is a client-side traffic analyzer that studies smartphone video streaming from different viewpoints. It collects information about network-level packet traffic, transport-layer flows, and application-level video player activities. Then it identifies relationships within the collected data to make mobile video streaming activities transparent. MoVIE is an open-source tool with a graphical user interface. In addition to network traffic measurement, MoVIE supports objective Quality of Experience (QoE) evaluation of video streaming. These features make MoVIE a powerful tool for network traffic measurement, multimedia streaming studies, and privacy analysis. We illustrate MoVIE' scapabilities with a small case study of streaming 360° videos.

The email delivery ecosystem in modern enterprise networks is large and complex, featuring several email access protocols (e.g., POP, IMAP, HTTP, and the secure versions of each of these). In this paper, we provide a detailed characterization of IMAPS email traffic on a large campus edge network. The main highlights from our study are: (1) commercial cloud-based email services account for most of the campus email traffic; (2) all email access protocols exhibit heavy-tailed transfer size distributions; and (3) the throughputs achieved by large transfers vary widely based on protocol, transfer size, and time of day.

Instagram is a popular network application for photo sharing, video streaming, and online social media interaction. In this paper, we present results from an initial characterization study of Instagram network traffic, as viewed from a large campus edge network. Despite the challenges of NAT, DHCP, end-to-end encryption, and high traffic volume, we are able to identify key characteristics of Instagram traffic, which exceeds 1 TB per day. The main highlights from our study include classic observations such as diurnal usage patterns, Zipf-like distributions for IP frequency-rank profile, and heavy-tailed transfer size distributions.

Traffic in today's edge networks is diverse, exhibiting many different patterns. This paper focuses on periodic network traffic, which is often used by known network services (e.g., Network Time Protocol, Akamai CDN) as well as by malicious applications (e.g., botnets, vulnerability scanning). We use a simple and flexible SQL-based approach as our computational model for detecting periodic traffic, and apply it to the analysis of seven weeks of Bro connection logs from a campus edge network. Our results show that periodic traffic analysis is effective for detecting P2P, gaming, cloud, scanning, and botnet traffic flows, which often exhibit periodic network communications. We present a classication taxonomy for periodic traffic, and provide an in-depth characterization of this traffic on our campus edge network.

Software-Defined Networking (SDN) can help simplify the management of today's complex networks and data centers. SDN provides a comprehensive view of the network, offering flexibility and easing automation. In SDN, traffic management functionality requires a high-performance and responsive controller. In this paper, we conduct an experimental evaluation of two open-source distributed OpenFlow controllers, namely ONOS and OpenDaylight. Specifically, we construct a testbed and use a standard benchmarking tool called Cbench to evaluate their performance. We benchmark the throughput, latency, and thread scalability of these two controllers in both physical and virtualized (OpenStack) environments. The experimental results show that ONOS provides higher throughput and lower latency than OpenDaylight, which suffers from performance problems on larger network models. Additional experiments demonstrate the effects of thread placement on the performance of these two controllers.

In this paper, we study the dynamics of coupled speed scaling systems, in which service rate is a function of system occupancy. We focus on both Processor Sharing (PS) and Shortest Remaining Processing Time (SRPT) as scheduling disciplines, and study their speed scaling dynamics under heavy load. Using a combination of Markov chain analysis and discrete-event simulation, we identify several important properties of speed scaling systems, which we call the autoscaling effect, the α effect, and the saturation effect. We also identify different overload regimes for PS and SRPT. In particular, SRPT exhibits a starvation effect that differs from the compensation effect of PS. These dynamics lead to different stability, fairness, and robustness properties for PS and SRPT under heavy load.

In this paper, we use experimental measurements to calibrate and validate a discrete-event simulator for dynamic speed scaling systems. The experimental implementation work is carried out in an Ubuntu Linux environment using a quad-core 2.3 GHz Intel i7 processor with the Ivy Bridge micro-architecture. Our implementation provides fine-grain user-level control of process execution, and uses the Running-Average Power Limit (RAPL) Machine Specific Registers (MSRs) to track energy usage. Through careful micro-benchmarking experiments, we determine the power consumption for each of the 12 discrete speeds supported by the processor, while also quantifying the costs of context switches and CPU speed changes. Finally, we use our suitably-parameterized speed scaling simulator to evaluate three different CPU speed scaling algorithms from the literature on simple batch workloads. To the best of our knowledge, our paper provides the first direct comparison of these speed scaling strategies with realistic system costs.

Multimedia streaming services such as YouTube and Netflix consume a staggering amount of Internet bandwidth [1]. Furthermore, traditional mechanisms such as proxy caches, content distribution networks, and redundant traffic elimination are rendered ineffective by copyright concerns, regulatory issues, and the growing prevalence of end-to-end encryption. One possible solution is a peer-to-peer caching system with social relationships at the core of its topology construction. A social topology carries an implicit level of trust, and induces a relatively high degree of correlation between users that can be exploited by the system as a whole. For example, two users with shared interests are more likely to have relevant videos in cache for each other. This short paper discusses the design of a simulator for such a system to provide insight into the performance of different cache management policies.

In speed scaling systems, the execution speed of a processor can be adjusted dynamically under operating system control to provide tradeoffs between response time, fairness, and energy consumption. In this paper, we propose and evaluate an approach called envelope-based turbo charging, applied in conjunction with Fair Sojourn Protocol (FSP) scheduling and job-count-based speed scaling. This approach restores the strict dominance of FSP over Processor Sharing (PS) in speed scaling systems, and preserves fairness. We evaluate our new approach using analysis and simulation. The simulation results show that Turbocharged FSP (T-FSP) outperforms PS in response time, and often in energy consumption as well. Furthermore, the energy consumption of T-FSP is typically within 15% of optimal.

The Zipf distribution is widely used to model Web site popularity, video popularity, and file referencing behavior. In recent published work, we proposed and evaluated a Zipf-based policy for probabilistic piece selection in Peer-to-Peer (P2P) media streaming. In this current paper, we revisit this Zipf model in more detail, and identify two fundamentally different modeling approaches, namely regenerative versus degenerative Zipf models. We illustrate the differences between the two models, provide refined analytical models for each, and validate the models with simulations in the context of P2P media streaming. The results show that the regenerative model is more appropriate for P2P streaming, because of its stronger sequential progress.

This paper introduces a novel analytical model for estimating the cache hit ratio as a function of time. The cache may not reach the steady-state hit ratio when the number of Web objects, object popularity, and/or caching resources themselves are subject to change. Hence, the only way to quantify the hit ratio experienced by Web users is to calculate the instantaneous hit ratio. The proposed analysis considers a single Web cache with infinite or finite capacity. For a cache with finite capacity, two replacement policies are considered: Least Recently Used (LRU) and First-In-First-Out (FIFO). Based on the insights from the proposed analytical model, we propose a new replacement policy, called Frequency-Based-FIFO (FB-FIFO). The results show that FB-FIFO outperforms both LRU and FIFO, assuming that the number of Web objects is fixed. Assuming that new popular objects are generated periodically, the results show that FB-FIFO adapts faster than LRU and FIFO to the changes in the popularity of the cached objects when the cache capacity is large relative to the number of newly generated objects.

We apply signal alignment (SA), a wireless communication technique that enables physical layer network coding (PNC) in multi-input multi-output (MIMO) wireless networks. Through calculated precoding, SA contracts the perceived signal space at a node to match its receive capability, and hence facilitates the demodulation of linearly combined data packets. PNC coupled with SA (PNC-SA) has the potential of fully exploiting the precoding space at the senders, and can better utilize the spatial diversity of a MIMO network for higher system degrees-of-freedom (DoF). PNC-SA adopts the idea of `demodulating a linear combination' from PNC. The design of PNC-SA is also inspired by recent advances in IA, though SA aligns signals not interferences. We study the optimal precoding and power allocation problem of PNC-SA, for SNR (singal-to-noise-ratio) maximization at the receiver. The mapping from SNR to BER is then analyzed, revealing that the DoF gain of PNC-SA does not come with a sacrifice in BER. We then design a general PNC-SA algorithm in larger systems, and demonstrate general applications of PNC-SA, and show via network level simulations that it can substantially increase the throughput of unicast and multicast sessions, by opening previously unexplored solution spaces in multi-hop MIMO routing.

We present Fluid Views, a web-based search environment designed to bridge overview and detail by integrating dynamic queries, semantic zooming, and dual layers. The most common form of search results is long ranked and paginated lists, which are seldom examined beyond the top ten items. To support more exploratory forms of information seeking, we bring together the notion of relevance with the power of visual encoding. In Fluid Views, results portray relevance via size and detail in a dynamic top layer and semantic similarity via position on a base map. We designed Fluid Views with temporal, spatial, and content-defined base maps for both textual and visual resources, and tested our prototype system on books, blogs, and photos. Interviews with library professionals indicate the potential of Fluid Views for exploring collections and exciting directions for future research.

The World Wide Web and the services it provides are continually evolving. Even for a single time instant, it is a complex task to methodologically determine the infrastructure over which these services are provided and the corresponding effect on user perceived performance. For such tasks, researchers typically rely on active measurements or large numbers of volunteer users. In this paper, we consider an alternative approach, which we refer to as passive crowd-based monitoring. More specifically, we use passively collected proxy logs from a global enterprise to observe differences in the quality of service (QoS) experienced by users on different continents. We also show how this technique can measure properties of the underlying infrastructures of different Web content providers. While some of these properties have been observed using active measurements, we are the first to show that many of these properties (such as location of servers) can be obtained using passive measurements of actual user activity. Passive crowd-based monitoring has the advantages that it does not add any overhead on Web infrastructure, it does not require any specific software on the clients, but still captures the performance and infrastructure observed by actual Web usage.

Redundant Traffic Elimination (RTE) detects and removes repeated chunks of data across network flows, protocols, and applications, with the purpose of reducing bandwidth usage. In this paper, we explore the effectiveness of RTE in WLAN, compare it to RTE in Ethernet, and investigate specific issues affecting RTE in WLAN. Our results show that applying RTE to WLAN links is promising and can potentially yield high bandwidth savings, although RTE is not as effective in WLAN as in wired networks. However, to exploit the full potential of RTE, it is necessary to deal with specific challenges, such as longer headers, control and management frames, retransmissions, and dropped frames. We find that including parts of MAC headers in RTE can increase overall bandwidth savings by up to 53% in a public WLAN. To handle dropped frames, which can severely compromise the effectiveness of RTE, we make a case for MAC-layer RTE, which detects frame loss at the sender. This preserves 23% more savings than a previous approach. However, frame retransmissions generate additional traffic at MAC layer, which reduces the effectiveness of RTE in general case.

NanoNets are networks of nanomachines at extremely small dimensions, on the order of nanometers or micrometers. Recent advances in physics and engineering have made basic computing and communication feasible on nanomachines, and NanoNets are envisioned as an important emerging technology with broad future applications. Traditional networking solutions require significant modifications for application in NanoNets. In this paper, we focus on routing algorithm design in NanoNets. Based on the salient features of a NanoNet, including low node cost and very low available power, we propose a new routing paradigm for multi-hop data transmission in NanoNets. Our design, termed {\em Buddy Routing (BR)}, is enabled by latest advancements in physical layer network coding, and argues for pair-to-pair data forwarding in place of traditional node-to-node data forwarding. Through both analysis and simulations, we compare BR with point-to-point routing, in terms of raw throughput, error rate, energy efficiency, and protocol overhead, and show the advantages of BR in NanoNets.

Cyberlockers have recently become a very popular means of distributing content. Today, cyberlocker traffic accounts for a non-negligible fraction of the total Internet traffic volume, and is forecasted to grow significantly in the future. The underlying protocol used in cyberlockers is HTTP, and increased usage of these services could drastically alter the characteristics of Web traffic. In light of the evolving nature of Web traffic, updated traffic models are required to capture this change. Despite their popularity, there has been limited work on understanding the characteristics of traffic flows originating from cyberlockers. Using a year-long trace collected from a large campus network, we present a comprehensive characterization study of cyberlocker traffic at the transport layer. We use a combination of flow-level and host-level characteristics to provide insights into the behavior of cyberlockers and their impact on networks. We also develop statistical models that capture the salient features of cyberlocker traffic. Studying the transport-layer interaction is important for analyzing reliability, congestion, flow control, and impact on other layers as well as Internet hosts. Our results can be used in developing improved traffic simulation models that can aid in capacity planning and network traffic management.

We present a comprehensive longitudinal characterization study of the dynamics of content sharing in the global file hosting landscape. We leverage datasets collected from multiple vantage points that allow us to understand how usage of these services evolve over time and how traffic is directed into and out of these sites. We analyze the characteristics of hosted content in the public domain, and investigate the dissemination mechanisms of links. To the best of our knowledge, this is the largest detailed characterization study of the file hosting landscape from a global viewpoint.

Data center networks are characterized by high link speeds, low propagation delays, small switch buffers, and temporally clustered arrivals of many concurrent TCP flows fulfilling data transfer requests. However, the combination of these features can lead to transient buffer overflow and bursty packet losses, which in turn lead to TCP retransmission timeouts that degrade the performance of short-lived flows. This so-called TCP-incast problem can cause TCP throughput collapse. In this paper, we explore an application-level approach for solving this problem. The key idea of our solution is to coordinate the scheduling of short-lived TCP flows so that no data loss happens. We develop a mathematical model of lossless data transmission, and estimate the maximum good put achievable in data center networks. The results indicate non-monotonic good put that is highly sensitive to specific parameter configurations in the data center network. We validate our model using ns-2 network simulations, which show good correspondence with the theoretical results.

Barrier coverage problems in sensor networks involve detecting intruders that attempt to cross a region of interest. In this paper, we formulate the k-connect barrier count problem for Mixed Sensor Networks (MSNs). The k-connect barrier count problem is to find the maximum number of barriers in an arbitrary MSN where at most k distinct mobile sensors can be used to construct any given virtual edge used in a barrier. We present the solution for the k-connect barrier count problem for k ∈ {0, 1, 2} via Integer Linear Programming. Using simulation results, we show that as k increases, the density of sensors required to achieve barrier coverage decreases. The results quantitatively demonstrate the benefits of mobile sensors.

Most topology control algorithms for wireless ad hoc networks strive to reduce energy consumption by creating a sparse topology with few long-distance links. However, in a sparse topology, the average path length is relatively large (increasing end-to-end delay), and the number of vertex-disjoint paths between source-destination pairs is relatively small (reducing fault-tolerance). Unlike traditional topology control algorithms that generate a single topology with a certain property, we propose a distributed algorithm that generates a family of topologies with a range of characteristics. The network designer can choose a suitable topology by simply tuning a single parameter Δ (power savings threshold), trading off energy savings for other features such as low latency and fault-tolerance. For the topologies generated by the proposed algorithm, we also provide an analytical model to estimate their structural density. The accuracy of the analytical model is validated with extensive simulation results.

This paper develops analytical models that characterize the behavior of on-demand stored media content delivery using BitTorrent-like protocols. The models capture the effects of different piece selection policies, including Rarest-First, two variants of In-Order, and two probabilistic policies (Portion and Zipf). Our models provide insight into system behavior and help explain the sluggishness of the system with In-Order streaming. We use the models to compare different retrieval policies across a wide range of system parameters, including peer arrival rate, upload/download bandwidth, and seed residence time. We also provide quantitative results on the startup delays and retrieval times for streaming media delivery. Our results provide insights into the design tradeoffs for on-demand media streaming in peer-to-peer networks. Finally, the models are validated using simulations.

We propose a new way of navigating the Web using interactive information visualizations, and present encouraging results from a large-scale Web study of a visual exploration system. While the Web has become an immense, diverse information space, it has also evolved into a powerful software platform. We believe that the established interaction techniques of searching and browsing do not sufficiently utilize these advances, since information seekers have to transform their information needs into specific, text-based search queries resulting in mostly text-based lists of resources. In contrast, we foresee a new type of information seeking that is high-level and more engaging, by providing the information seeker with interactive visualizations that give graphical overviews and enable query formulation. Building on recent work on faceted navigation, information visualization, and exploratory search, we conceptualize this type of information navigation as visual exploration and evaluate a prototype Web-based system that implements it. We discuss the results of a large-scale, mixed-method Web study that provides a better understanding of the potential benefits of visual exploration on the Web, and its particular performance challenges.

We introduce the information flaneur as a new human-centred view on information seeking that is grounded in interdisciplinary research. We use the metaphor of the urban flaneur making sense of a city as an inspiring lens that brings together diverse perspectives. These perspectives shift information seeking towards a more optimistic outlook: the information flaneur represents curious, creative, and critical information seeking. The resulting information-seeking model conceptualizes the interrelated nature between information activities and experiences as a continuum between horizontal exploration and vertical immersion. Motivated by enabling technological trends and inspired by the information flaneur, we present explorability as a new guiding principle for design and raise research challenges regarding the representation of information abstractions and details.

Protocol-independent redundant traffic elimination (RTE) is an "on the fly" method for detecting and removing redundant chunks of data from network-layer packets traversing a constrained link or path. Efficient algorithms are needed to sample data chunks and detect redundancy, so that RTE does not hinder network throughput. A recently proposed static algorithm samples chunks based on highly-redundant trigger bytes observed in data content. While this algorithm is fast, it requires pre-computed traffic information for the configuration of its static parameters, and it tends to either under-sample (reducing byte savings) or over-sample (increasing processing cost) on heterogeneous traffic. We propose a dynamic sampling algorithm for redundant content detection. Our algorithm is adaptive and self-configuring, and can precisely match the specified sampling rate. Furthermore, it offers byte savings comparable to the static algorithm, with very low additional processing overhead.

We explored the TCP-incast throughput collapse problem in data center networks from an application-layer perspective. In particular, we presented the model and analyzed the performance of the application-based approach under TCP-incast scenario. The main idea of the approach is to schedule the server responses to data requests so that no packet losses occur at the bottleneck link. The main result we derive is the achievable goodput of a data center application if using lossless scheduling. The simulations confirmed the validity of our model and the results derived through the theoretical analysis. Our future work is to explore practical scenarios of our approach and implement it in a real data center network.

We propose signal alignment (SA), a new wireless communication technique that enables physical layer network coding (PNC) in multi-input multi-output (MIMO) wireless networks. Through calculated preceding, SA contracts the perceived signal space at a node to match its receive diversity, and hence facilitates the demodulation of linearly combined data packets. PNC coupled with SA (PNC-SA) has the potential of fully exploiting the preceding space at the senders, and can better utilize the spatial diversity of a MIMO network for higher transmission rates, outperforming existing techniques including MIMO or PNC alone, interference alignment (IA) and interference alignment and cancellation (IAC). PNC-SA adopts the seminal idea of 'demodulate a linear combination' from PNC. The design of PNC-SA is also inspired by recent advances in IA, though SA aligns signals not interferences. We study the optimal preceding and power allocation problem of PNC-SA, for SNR maximization at the receiver. The mapping from SNR to BER is then analyzed, revealing that the throughput gain of PNC-SA does not come with a sacrifice in BER. We finally demonstrate general applications of PNC-SA, and show via network level simulations that it can substantially increase the throughput of unicast and multicast sessions, by opening previously unexplored solution spaces in multi-hop MIMO routing.

There is a continuous struggle for control of resources at every organization that is connected to the Internet. The local organization wishes to use its resources to achieve strategic goals. Some external entities seek direct control of these resources, for purposes such as spamming or launching denial-of-service attacks. Other external entities seek indirect control of assets (e.g., users, finances), but provide services in exchange for them. Using a year-long trace from an edge network, we examine what various external organizations know about one organization. We compare the types of information exposed by or to external organizations using either active ( reconnaissance ) or passive ( surveillance ) techniques. We also explore the direct and indirect control external entities have on local IT resources.

Today’s Web provides many different functionalities, including communication, entertainment, social networking, and information retrieval. In this article, we analyze traces of HTTP activity from a large enterprise and from a large university to identify and characterize Web-based service usage. Our work provides an initial methodology for the analysis of Web-based services. While it is nontrivial to identify the classes, instances, and providers for each transaction, our results show that most of the traffic comes from a small subset of providers, which can be classified manually. Furthermore, we assess both qualitatively and quantitatively how the Web has evolved over the past decade, and discuss the implications of these changes.

Securing wireless sensor networks against node capture is a challenging task. All well-known random key pre-distribution systems, including the Eschenauer and Gligor's pioneering scheme, its extensions, as well as threshold schemes, become insecure when a large number of nodes are captured. We propose a general technique, called virtual key ring, that can effectively strengthen the resilience of random key pre-distribution systems against node capture attacks by reducing the pre-loaded keying material while maintaining secure connectivity of the network.The technique is general and applicable to many key pre-distribution systems. We however focus on the original EG scheme and propose a virtual key ring system based on this pioneering scheme. We provide detailed mathematical analysis and a security proof for the system, and use extensive simulation to validate the analysis and to compare performance of the new system with the original EG scheme. We also present simulation results for the strengthened resilience when the virtual key ring scheme is combined with the multipath key reinforcement and q-composite techniques, showing that the system resilience is substantially improved against large-scale node capture attack (e.g., 40% of nodes captured).

This paper presents a method for supporting wireless media streaming using a cache that is distributed across the mobile devices in the network. The performance of this scheme is compared to traditional institutional server (IS) caching on a network with a bandwidth constrained wireless backhaul. In addition to traditional caching hit ratio metrics, the paper studies how caching affects the call drop ratio due to limited backhaul bandwidth. These results indicate that the distributed caching method provides better service than IS caching as the number of users is increased. Finally, this paper also presents a scheme for conserving mobile device energy by limiting its participation in the caching scheme. Results show that most of the benefit of the distributed cache can be realized even with relatively few cache assists from each client.

BitTorrent-like swarming technologies are very effective for popular content, but less so for the `long tail' of files with disparate popularities, which do not have sufficiently many peers to enable efficient collaboration. Performance degradations are especially pronounced in swarms with reduced file availability. Static bundling groups files into a single data content. It requires no modification to the BitTorrent client, and has been shown to improve availability of unpopular files in BitTorrent swarms. However, as peers are forced to download undesired file pieces, download times increase, especially for peers downloading popular files. We propose to use Stochastic Games and Markov Decision Process (MDP) to model and analyze optimal peer strategies, in a selfish and a cooperative setting respectively, for a BitTorrent-like system with multiple files. Each peer wishes to download a subset of the files, and we allow peers to dynamically decide whether to collaborate with peers targeting a different set of files or not, given the current system state. The Stochastic Game and MPD models take into account both piece availability and average download times, and allow us to study if and when downloading unwanted content can be beneficial. We use dynamic programming to solve the two models, contrast the level of collaboration observed in the selfish and the cooperative settings, and propose an enhanced piece selection mechanism for BitTorrent-like systems with dynamic download decision making. We demonstrate the effectiveness of dynamic file piece selection through both simulations and experiments using a modified BitTorrent client.

While Transmission Control Protocol (TCP) variants with delay-based congestion control (e.g., TCP Vegas) provide low queueing delay and low packet loss, the key problem with their deployment on the Internet is their relative performance when competing with traditional TCP variants with loss-based congestion control (e.g., TCP NewReno). In particular, the more aggressive loss-based flows tend to dominate link buffer usage and degrade the throughput of delay-based flows. In this paper, we study a novel approach for achieving fair sharing of the network resources among TCP variants, using Rate-Delay (RD) Network Services. In particular, loss-based and delay-based flows are isolated from each other and served via different queues. Using extensive ns-2 network simulation experiments, we show that our approach is effective in providing fairness between loss-based NewReno and delay-based Vegas flows.

Mobility models play an important role in the evaluation of wireless ad hoc networks. However, most existing mobility models are limited to 1D or 2D user movement. In this paper, we propose a novel 2.5D mobility model suitable for modeling the movement of backcountry skiers in mountainous terrains. These skiers carry wireless devices for communication in the event of emergencies, such as an avalanche, and thus form a dynamic mobile wireless ad hoc network. Our model represents groups of skiers, with each group having an invisible leader who determines the general direction of movement. We calculate the acceleration, velocity, and position of the group leader based on properties of the terrain, gravitational force, and a randomized sine function. To simulate the coordinated mobility of group members, we use a flocking model, which applies cohesion, separation, and alignment forces to influence group members to follow their leader, while avoiding collisions with each other. To demonstrate the applicability of our model, we use simulation to evaluate the skier rescue probability in the event of an avalanche.

In this paper, we address the problem of low-latency routing in a vehicular highway network. To cover long highways while minimizing the number of required roadside access points, we utilize vehicle-to-vehicle communication to propagate data in the network. Vehicular networks are highly dynamic, and hence routing algorithms that require global network state information or centralized coordination are not suitable for such networks. Instead, we develop a novel distributed routing algorithm that requires minimal coordination among vehicles, while achieving a highly efficient throughput-delay tradeoff. Specifically, we show that the proposed algorithm achieves a throughput that is within a factor of 1/e of the throughput of an algorithm that centrally coordinates vehicle transmissions in a highly dense network, and yet its end-to-end delay is approximately half of that of a widely studied ALOHA-based randomized routing algorithm. We evaluate our algorithm analytically and through simulations and compare its throughput-delay performance against the ALOHA-based randomized routing.

We consider the problem of optimal cluster-based data gathering in Wireless Sensor Networks (WSNs) when nearby readings are spatially correlated. Due to the dense nature of WSNs, data samples taken from nearby locations are statistically similar. We show how this data correlation can be exploited to reduce the amount of data to be transmitted in the network and thus conserve energy. While much attention in recent years has been paid to analyzing and optimizing cluster-based WSNs from various perspectives, the problem of energy-efficient clustering of WSNs in presence of data correlation is not yet fully explored. In this paper, we model a single-cluster network and analytically characterize the optimal cluster size subject to its distance from the sink as well as the degree of correlation. Contrary to existing approaches, our findings show that heterogeneous-sized clusters, where the clusters further from the sink are larger, are more energy-efficient. We also propose a heuristic greedy clustering algorithm to find a near-optimal solution to the problem of energy-efficient clustering. Simulation results confirm the effectiveness of having heterogeneous-sized clusters in WSNs.

In a Wireless Sensor Network (WSN), the sensor nodes rely upon a multi-hop routing protocol to relay their data to the base station. However, most WSN routing protocols are vulnerable to attacks in which a malicious node can disrupt the routes, drop, modify, or divert data away from the base station. In this paper, we use the ns-2 network simulator to demonstrate the vulnerability of the MintRoute protocol to link quality attacks by a malicious node. We then propose a novel "sequence number gap trick" as a lightweight means to test for and detect the presence of a malicious attacker. The simulation results show that judicious use of the sequence number gap trick provides robust detection of malicious nodes, preserving the data delivery capabilities of the WSN.

We introduce the concept of a Visual Backchannel as a novel way of following and exploring online conversations about large-scale events. Microblogging communities, such as Twitter, are increasingly used as digital backchannels for timely exchange of brief comments and impressions during political speeches, sport competitions, natural disasters, and other large events. Currently, shared updates are typically displayed in the form of a simple list, making it difficult to get an overview of the fast-paced discussions as it happens in the moment and how it evolves over time. In contrast, our Visual Backchannel design provides an evolving, interactive, and multi-faceted visual overview of large-scale ongoing conversations on Twitter. To visualize a continuously updating information stream, we include visual saliency for what is happening now and what has just happened, set in the context of the evolving conversation. As part of a fully web-based coordinated-view system we introduce Topic Streams, a temporally adjustable stacked graph visualizing topics over time, a People Spiral representing participants and their activity, and an Image Cloud encoding the popularity of event photos by size. Together with a post listing, these mutually linked views support cross-filtering along topics, participants, and time ranges. We discuss our design considerations, in particular with respect to evolving visualizations of dynamically changing data. Initial feedback indicates significant interest and suggests several unanticipated uses.

Scheduling decisions can have a pronounced impact on the performance of multi-radio wireless systems. In this paper, we study the effects of dispatch policies and queue scheduling strategies on the user-perceived performance for Internet traffic flows in a multi-channel WLAN. Our work is carried out using simulation and an empirical Web workload trace, with mean response time as the primary performance metric. The simulation results demonstrate the good/bad combination of the dispatch policy with queue scheduling strategy, the advantages of deferred dispatch over immediate dispatch, and the sensitivity of dispatch policies to heavy-tailed workload characteristics. The results also highlight the pros and cons of a simple lookahead scheduling policy, particularly in the presence of high variability workloads on a heterogeneous multi-channel system with random losses. Our results provide insights into efficient and robust scheduling policies for multi-channel WLANs.

Recent advances in wireless technology have made it increasingly feasible to equip wireless nodes with multiple radios, thereby allowing each radio to exploit channel diversity in the form of orthogonal, non-overlapping transmission spectrums. Multi-channel operation mitigates interference, but at the same time raises new challenges for network optimization, in terms of judicious channel assignment for efficient bandwidth utilization. While previous research mostly studies optimizing channel assignment for unicast, we focus instead on multicast, which is an efficient mechanism for one-to-many data dissemination. We derive a model for optimal multicast in multi-channel multi-radio wireless networks under the assumption that channel assignment is static. Our model employs network coding as the multicast mechanism of choice, and exploits the broadcast nature of omnidirectional antennas for efficient bandwidth utilization. Based on the model derived, we formulate optimal multicast as a linear integer program. Two accompanying solutions are proposed: a greedy channel assignment scheme and an improved iterative scheme inspired by primal-dual algorithm design. The effectiveness of the two schemes are empirically examined through simulation studies, and are compared to results obtained from solving the integer program as well as its linear programming relaxation. Finally, we present an alternate model for optimal multicast under the assumption that transmission frequencies are not fixed divisions of the usable spectrum.

In multi-rate cellular transmission systems, users with different Quality of Service (QoS) requirements share the same wireless channel. In this paper, we investigate the problem of efficient resource allocation for scheduling with differentiated QoS support in a multi-rate system. We propose Dynamic Global Proportional Fairness (DGPF) scheduling on the downlink. We investigate the performance of the scheduling algorithm and model the proposed scheme in a High Speed Downlink Packet Access (HSDPA) simulation environment. The simulation results show that our approach can achieve suitable QoS for different classes of users without compromising aggregate network throughput. The results also show that TCP dynamics affect overall system performance.

In this paper, we consider the design of caching infrastructure to enhance the client-perceived performance of mobile wireless clients retrieving multimedia objects from the Internet. We consider three primary issues: location of the cache, size of the cache, and management policy for the cache. We consider both infrastructure-oriented caching at the Access Point (AP), as well as peer-assisted caching at the mobile clients. Simulation is used as the methodology for evaluation and comparison of caching strategies. The simulation results show that AP caching is generally more effective than client-side caching, that adequate performance is achievable with a mix of rather modest AP and client-side caches, and that Least Frequently Used (LFU) is the most effective cache replacement policy. Additional simulation experiments show that our results are robust across different request generation rates and client turnover rates.

We study multicast in a noncooperative environment where information flows selfishly route themselves through the cheapest paths available. The main challenge is to enforce such selfish multicast flows to stabilize at a socially optimal operating point incurring minimum total edge cost, through appropriate cost allocation and other economic measures, with replicable and encodable properties of information flows considered. We show that known cost allocation schemes are not sufficient. We provide a shadow-price-based cost allocation for networks without capacity limits and show that it enforces minimum-cost multicast. This improves previous result where a 2-approximate multicast flow is enforced. For capacitated networks, computing cost allocation by ignoring edge capacities will not yield correct results. We show that an edge tax scheme can be combined with a cost allocation to strictly enforce optimal multicast flows in this more realistic case. If taxes are not desirable, they can be returned to flows while maintaining weak enforcement of the optimal flow. We relate the taxes to VCG payment schemes and discuss an efficient primal-dual algorithm that simultaneously computes the taxes, the cost allocation, and the optimal multicast flow, with potential of fully distributed implementations.

Current 3G cellular networks, such as those based on 1xEVDO Revision A, have several mechanisms to provide quality of service (QoS) to network applications. In this paper, we investigate the performance for mixed Web/VoIP traffic on the downlink in 1xEVDO revision a networks. The impacts from the mixed traffic are studied by using link-level simulations. Our simulator is built based on a realistic system model that includes physical channel parameters, transmission formats, and channel scheduling algorithms, as well as hierarchical traffic models. User-perceived throughput and packet delay are explored. We present our observations and determine the Erlang capacity of the network. We carry out performance analysis using an equivalent queue approach, and discuss appropriate scheduling algorithms and call admission control mechanisms based on the simulation results.

This paper develops analytic models that characterize the behavior of on-demand stored media content delivery using BitTorrent-like protocols. The models capture the effects of different piece selection policies, including Rarest-First and two variants of In-Order. Our models provide insight into transient and steady-state system behavior, and help explain the sluggishness of the system with strict In-Order streaming. We use the models to compare different retrieval policies across a wide range of system parameters, including peer arrival rate, upload/download bandwidth, and seed residence time. We also provide quantitative results on the startup delays and retrieval times for streaming media delivery. Our results provide insights into the optimal design of peer-to-peer networks for on-demand media streaming.

Peer-to-Peer (P2P) applications continue to grow in popularity, and have reportedly overtaken Web applications as the single largest contributor to Internet traffic. Using traces collected from a large edge network, we conduct an extensive analysis of P2P traffic, compare P2P traffic with Web traffic, and discuss the implications of increased P2P traffic. In addition to studying the aggregate P2P traffic, we also analyze and compare the two main constituents of P2P traffic in our data, namely BitTorrent and Gnutella. The results presented in the paper may be used for generating synthetic workloads, gaining insights into the functioning of P2P applications, and developing network management strategies. For example, our results suggest that new models are necessary for Internet traffic. As a first step, we present flow-level distributional models for Web and P2P traffic that may be used in network simulation and emulation experiments.

In common Web-based search interfaces, it can be difficult to formulate queries that simultaneously combine temporal, spatial, and topical data filters. We investigate how coordinated visualizations can enhance search and exploration of information on the World Wide Web by easing the formulation of these types of queries. Drawing from visual information seeking and exploratory search, we introduce VisGets--interactive query visualizations of Web-based information that operate with online information within a Web browser. VisGets provide the information seeker with visual overviews of Web resources and offer a way to visually filter the data. Our goal is to facilitate the construction of dynamic search queries that combine filters from more than one data dimension. We present a prototype information exploration system featuring three linked VisGets (temporal, spatial, and topical), and used it to visually explore news items from online RSS feeds.

In this paper we study the use of TCP Vegas as a transport protocol for streaming media. We also consider TCP NewReno as a transport protocol for streaming media. We find that: 1) TCP is able to transport streaming media with good performance in a wide variety of scenarios; 2) TCP Vegas performs better than TCP NewReno in many cases; and 3) for viable media streams, both TCP variants need to achieve mean throughputs that are at least 1.5 times the encoding rate of the media objects being carried.

Wireless local area networks (WLANs) have become omnipresent: WLANs are available at airports, coffee shops, university campuses, corporate environments, and homes. This surge in the popularity of WLANs motivates the study of how these networks are used. Characterizing WLANs, however, is complicated by a number of factors including the geographic diversity of WLAN deployments and the need for capturing activity in the wireless environment instead of the wired environment. In this paper, we describe our experiences with the deployment and use of a remote passive wireless-side measurement infrastructure for monitoring usage of WLANs, and compare our results with a commonly used wired-side measurement technique.

No abstract available.

Analyzing traces of wireless network activity has many pragmatic purposes, from capacity planning to network design. Unfortunately, capturing complete traces of wireless traffic is difficult, and using incomplete traces can degrade the quality of the aforementioned analyses. In this paper we examine three different methods for estimating the completeness of wireless traces. We find that a method that examines MAC-layer sequence numbers provides the most accurate results. We also examine the effect of the placement of wireless sensors on the completeness of wireless-side traces. We determine that locating sensors such that the signal strengths between clients and access points is over 40% results in low miss rates at the sensor, and few CRC errors.

Traffic classification is the ability to identify and categorize network traffic by application type. In this paper, we consider the problem of traffic classification in the network core.Classification at the core is challenging because only partial information about the flows and their contributors is available. We address this problem by developing a framework that can classify a flow using only unidirectional flow information. We evaluated this approach using recent packet traces that we collected and pre-classified to establish a "base truth". From our evaluation, we find that flow statistics for the server-to-client direction of a TCP connection provide greater classification accuracy than the flow statistics for the client-to-server direction. Because collection of the server-to-client flow statistics may not always be feasible, we developed and validated an algorithm that can estimate the missing statistics froma unidirectional packet trace.

This paper studies the Web proxy cache placement problem, in which m caching proxies are to be placed in a network so as to minimize the average response time for users accessing Web content. We compare an evolutionary approach to this network optimization problem with two classical approaches, namely dynamic programming and packet-level simulation. The results show that the evolutionary approach produces results as good as or better than the other approaches. Furthermore, the evolutionary approach is computationally faster, enabling the study of larger network scenarios than possible with the other approaches.

In this paper, we compare the throughputs of two different TCP NewReno variants, namely Slow-but-Steady and Impatient. We develop analytic throughput models of these variants as a function of round-trip time, loss event rate, and the burstiness of packet drops within a loss event. Our models build upon prior work on TCP Reno throughput modeling, but extend this work to provide an analytical characterization of the NewReno fast recovery algorithms. We validated our models using the ns-2 simulator. Our models accurately predict the steady-state NewReno throughput for a wide range of loss rates. Based on these models, we analytically determine the preferred operating regions for each TCP variant. Our results show that the Slow-but-Steady variant is comparable to or superior to the Impatient variant in all but the most extreme scenarios for network packet loss.

In this paper, a comprehensive system model is built to evaluate system performance for data services in CDMA networks. Unlike traditional analyses, we model both system capacity and traffic demands using stochastic processes. Call-level simulation is conducted based on the proposed analytical model. The simulation results show that stochastic traffic characteristics can affect the system performance significantly. Traffic correlation is beneficial for the same system load level. Moreover, the simulation results illustrate that using a simple activity factor to model the traffic process can cause capacity over estimation

This paper proposes a scalable architecture for multimedia streaming in wireless LANs. Current IEEE 802.11 WLANs can support tens of media streaming users. We propose a hierarchical approach that can support over 1000 concurrent users. Our architecture exploits several existing technologies, including multi-channel WLANs, power control, and caching. The paper first describes the system design and operation, as well as assumptions and constraints. Performance issues are then explored via simulation, using synthetically-generated media streaming workloads. The simulation results show that: 1) our design can support up to 1600 concurrent media streaming clients using current WLAN technology; 2) the system architecture is quite robust to the user-level characteristics of the media streaming workload; and 3) proper cache management can make the system operate effectively even with limited cache sizes.

This paper focuses on network traffic measurement of Peer-to- Peer (P2P) applications on the Internet. P2P applications supposedly constitute a substantial proportion of today's Internet traffic. However, current P2P applications use several obfuscation techniques, including dynamic port numbers, port hopping, HTTP masquerading, chunked file transfers, and encrypted payloads. As P2P applications continue to evolve, robust and effective methods are needed for P2P traffic identification. The paper compares three methods to classify P2P applications: port-based classification, application-layer signatures, and transport-layer analysis. The study uses empirical network traces collected from the University of Calgary Internet connection for the past 2 years. The results show that port-based analysis is ineffective, being unable to identify 30%-70% of today's Internet traffic. Application signatures are accurate, but may not be possible for legal or technical reasons. The transport-layer method seems promising, providing a robust means to assess aggregate P2P traffic. The latter method suggests that 30%-70% of the campus Internet traffic for the past year was P2P.

Multi-channel medium access control (MAC) protocols have recently been proposed to improve the performance of the transmission control protocol (TCP) in IEEE 802.11 wireless ad hoc networks. This paper uses ns-2 network simulations to study the impact of channel selection techniques on multi-channel MAC protocol performance, particularly for the bi-directional multi-channel MAC protocol. Three channel selection strategies are studied: random, lowest channel first, and soft channel reservation. The simulation results identify four distinct scenarios in which data channel frame losses can occur. Among the channel selection strategies evaluated, the soft channel reservation technique is the most effective for the missed reservation problem. This channel selection strategy reduces link-layer data frame losses and provides higher TCP throughput compared to the other channel selection approaches

This paper studies a Bad Apple phenomenon caused by Head-of-Line blocking in IEEE 802.11 networks. This problem can adversely affect the performance of wireless multimedia streaming. We summarize the characteristics of the Bad Apple problem, and then propose a Station-Based Adaptation (SBA) algorithm to solve it. SBA maintains perstation information, operating at the MAC layer as an extension to 802.11 DCF. SBA deactivates flows destined to temporarily unreachable stations, and later reactivates flow transmission when the stations are available. Simulation results show that SBA can effectively improve the performance and robustness of the 802.11 system when stations experience unpredictable temporary failures that are several seconds in duration. 1

This paper describes the analysis of low-level measurements from a CDMA2000 1x cellular data network. The network traces record detailed information about wireless Internet packet data call activity on the network, including mobile station identity, call initiation, burst behaviour, supplementary channel usage, soft handoffs, and call termination. The analysis in this paper focuses on one continuous week-long trace data set, representative of cellular data network activity. The results from the analysis illustrate the burstiness of the packet call arrival process and the diurnal patterns of cellular data users. The results also characterize the activity per cell site, activity per user, data burst activity, user mobility, and the density of cellular network coverage. Several observations reinforce known results about heavytailed properties in wired Internet traffic, while others show interesting differences in wireless versus wireline traffic

This work studies the user-level and network-level performance for a mobile client browsing the Web using a wireless personal digital assistant (PDA). A wireless network analyzer is used to collect traces of the TCP/IP network traffic generated to and from the wireless PDA. Multi-layer protocol analysis is used to assess the Web browsing performance achieved, and identify protocol performance problems. The analyses focus on server response time, document size, HTTP transfer time, TCP performance, and wireless channel quality. The results from our experiments provide insight into strategies that could be used by Web content providers, Web servers, HTTP and TCP to improve Web browsing performance for mobile users.

The TCP protocol often suffers from performance problems in conventional single-channel multihop wireless ad hoc networks. The problems arise from hidden node and exposed node issues, which can lead to channel contention in the forward direction between TCP DATA packets that are part of the same TCP flow control window, as well as contention between TCP DATA and TCP ACK packets flowing in opposite directions. In this paper, we propose and evaluate a novel bidirectional multi-channel MAC protocol designed to improve TCP performance over a multihop wireless network. The protocol uses multiple transmission channels at the physical layer to reduce TCP DATA-DATA contention, and bidirectional RTS/CTS channel reservations to reduce TCP DATA-ACK collisions. With four channels, simulation results for static multihop networks show TCP throughput gains of 50% to 180%, compared to a conventional IEEE 802.11 MAC protocol. Fairness is also improved with our protocol, since contention is confined to a short handshake period on the control channel.

Abstract This paper proposes a methodological approach to the evaluation of Web server performance in a simple local area network test environment. The paper examines how different system and application configuration parameters can, over a range of workloads, impact the performance of a Web server. Our approach relies on relatively fine‐grain reporting of performance data for a broad set of system‐level metrics. Graphical visualization of these performance indices helps to identify the primary system bottleneck in each configuration studied. The Apache Web server is used as a case study to demonstrate the methodology. Our experiments quantify the performance implications of several configuration decisions common to any Web server implementation, and also serve to illustrate several performance anomalies specific to the Apache Web server (if misconfigured). Copyright © 2004 John Wiley & Sons, Ltd.

This paper presents a simulation study of an IEEE 802.11b wireless LAN (WLAN) used as a classroom area network. The simulation is conducted using OPNET modeler 9.1. The first part of the paper discusses parameterization and validation of the simulation model, based on empirical measurements in a wireless classroom environment. The second part of the paper presents a simulation study designed to estimate the number of clients that can be supported in the WLAN, as well as the user-perceived Web response time as a function of network load. The simulation results show that an IEEE 802.11b WLAN can easily support up to 100 clients with modest Web browsing activities. The results also show that protocol features such as persistent connections provide a significant performance advantage in a WLAN environment.

This paper discusses the use of a parallel discrete-event network emulator called the Internet Protocol Traffic and Network Emulator (IP-TNE) for Web server benchmarking. The experiments in this paper demonstrate the feasibility of high-performance WAN emulation using parallel discrete-event simulation techniques on shared-memory multiprocessors. Our experiments with the Apache Web server achieve 3400 HTTP transactions per second for simple Web workloads, and 1000 HTTP transactions per second for realistic Web workloads, for static document retrieval across emulated WAN topologies of up to 4096 concurrent Web/TCP clients. The results show that WAN characteristics, including round-trip delays, link speeds, packet losses, packet sizes, and bandwidth asymmetry, all have significant impacts on Web server performance. WAN emulation enables stress testing and benchmarking of Web server performance in ways that may not be possible in simple LAN test scenarios.

This paper discusses the design and evaluation of CATNIP, a Context-Aware Transport/Network Internet Protocol for the Web. This integrated protocol uses application-layer knowledge (i.e., Web document size) to provide explicit context information to the TCP and IP protocols. While this approach violates the traditional layered Internet protocol architecture, it enables informed decision-making, both at network endpoints and at network routers, regarding flow control, congestion control, and packet discard decisions.The ns-2 network simulator is used to evaluate the performance of the context-aware TCP/IP approach, using a simple network topology and a synthetic Web workload. Simulation results indicate a 10-20% reduction in TCP packet loss using simple endpoint control mechanisms. More importantly, using CATNIP context information at IP routers can produce 20-80% reductions in the mean Web page retrieval times, and 60-90% reductions in the standard deviation of retrieval times.

This article studies the "filter effects" that occur in Web proxy caching hierarchies due to the presence of multiple levels of caches. That is, the presence of one level of cache changes the structural characteristics of the workload presented to the next level of cache, since only the requests that miss in one cache are forwarded to the next cache.Trace-driven simulations, with empirical and synthetic traces, are used to demonstrate the presence and magnitude of the filter effects in a multilevel Web proxy caching hierarchy. Experiments focus on the effects of cache size, cache replacement policy, Zipf slope, and the depth of the Web proxy caching hierarchy.Finally, the article considers novel cache management techniques that can better exploit the changing workload characteristics across a multilevel Web proxy caching hierarchy. Trace-driven simulations are used to evaluate the performance of these approaches. The simulation results demonstrate that size-based partitioning and heterogeneous cache replacement policies each offer improvements in overall caching performance. The sensitivity of the results to the degree of workload overlap among child-level proxy caches is also studied.

This paper describes the design and use of a synthetic Web proxy workload generator (ProWGen) to investigate the sensitivity of proxy cache replacement policies to selected Web workload characteristics. Trace-driven simulations with synthetic workloads from ProWGen show the relative sensitivity of three popular cache replacement algorithms-LRU, LFU-aging and GD-size-to Zipf slope, temporal locality, and correlation (if any) between file size and popularity, and the relative insensitivity of these algorithms to one-timers and heavy tail index. Performance differences between the three policies are also highlighted.

This paper studies the performance of the transmission control protocol (TCP) over asynchronous transfer mode (ATM) when asymmetric digital subscriber line (ADSL) technology is used in the local loop. TCP can experience performance degradation in this network architecture because of protocol conversion overhead and data losses due to transmission errors. A simulation model is developed to simulate ADSL network components and noisy local loops. The simulation experiments are designed to study the impact of channel errors on TCP performance for unidirectional bulk data transfers, using both an independent error model and a burst error model. The primary performance metrics are cell loss ratio, packet loss ratio, and TCP effective throughput. The simulation results illustrate how TCP is affected by channel errors, as well as the impacts of the TCP maximum segment size (MSS), switch buffer size, bandwidth asymmetry, and the percentage of noisy lines.

Focuses on the differences between monofractal and multifractal network traffic, both in terms of modeling approaches and in terms of practical impacts on network performance. Empirical traffic traces are used in the parameterization and evaluation of traffic models. Simulation is used to evaluate the performance differences between monofractal and multifractal traffic, both for individual traffic streams and for aggregations of traffic streams. The simulation results indicate that multifractal traffic offers greater potential for multiplexing gains, both within and across sources. The sensitivity of these results to source characteristics and switch buffer size is also explored.

Describes three visually interactive tools for the analysis, modeling and generation of long-range dependent (LRD) network traffic: (1) the synTraff toolkit uses a three-step modeling approach based on F-ARIMA (fractal autoregressive integrated moving average) processes to generate monofractal traffic; (2) the WsynTraff toolkit implements the wavelet-domain independent Gaussian (WIG) model (S. Ma et al., 1998) for representing multifractal traffic; and (3) the MsynTraff toolkit models multifractal traffic using the multifractal wavelet model (MWM) (V. Ribeiro et al., 1999). All three tools can be used to generate synthetic LRD traffic streams for use in network simulations.

TCP's sliding window flow control mechanism often leads to bursty packet traffic in the Internet. In the presence of network congestion, multiple packet losses can occur degrading the performance of TCP and the user-perceived performance of the Web. This paper presents a rate-based pacing (RBP) mechanism for TCP flow control. This mechanism can reduce the burstiness of TCP packet transmission and improve document transfer performance. Simulation is used to evaluate the performance of this mechanism. Our RBP TCP protocol is found to provide similar performance to Reno TCP in terms of throughput and packet loss for bulk data transfers, but lower packet loss and 10-20% lower latency than Reno TCP for Web-like document transfers.

This paper investigates, through simulation, the performance of five connection admission control (CAC) algorithms used in ATM networks, namely: PCR CAC, SCR CAC, AVG CAC, GCAC, and Norros CAC, when presented with synthetically generated homogeneous self-similar traffic sources. Various traffic and system parameters have been used. The simulation results show that statistical multiplexing gains both within a source and across sources should be exploited to improve network resource utilization. Source granularity, and variability have been shown to have the most impact on the CAC performance. The limited impact of the Hurst parameter shows that the long-range correlation structure of the traffic sources can be neglected in evaluating CAC performance, at least when the buffer size is small. The CAC algorithms should be more conservative when link capacity is low and more aggressive when link capacity is high. While none of the CAC algorithms performs satisfactorily in all scenarios, the Norrors CAC and AVG CAC perform better than the others.

This paper compares and evaluates three ABR flow control strategies (ERICA, ERICA+, and DEBRA) for ATM networks, using simulation. A set of benchmark network scenarios is used to illustrate the properties of each algorithm, such as efficiency, fairness, responsiveness, and scalability, as well as robustness in the presence of uncooperative sources. The simulation results show that ERICA+ and DEBRA perform similarly on most scenarios, and slightly better than ERICA, while DEBRA is more robust than ERICA and ERICA+. Overall, the study shows that DEBRA, a new explicit-rate flow control strategy, shows good potential for ABR traffic control in ATM networks.

Achieving statistical gains when multiplexing video streams, as in a video-on-demand (VOD) scenario, is difficult because of the stringent QOS demands and the self-similar nature of the traffic. This paper explores, through empirical simulation, the QOS, network utilization, and statistical characteristics of the aggregate traffic resulting from multiple independent MPEG video streams. In addition, the simulation results are compared against several recently-derived theoretical results for self-similar network traffic. Three main results are evident from our experiments. First, moderate statistical multiplexing gain can be achieved when multiplexing multiple self-similar streams. Second, video multiplexing is extremely sensitive to traffic phasing effects and to heavy-tailed frame size distributions. Finally, the theoretical approach considered (Norros (see IEEE Journal on Selected Areas in Communications, vol.13, no.6, p.953-62, 1995) effective bandwidth formulation) appears promising but requires fine-tuning to be practical for call admission and network dimensioning.

This paper presents a new approach to TCP congestion control. The new scheme includes two parts: (I) the smooth-start algorithm, which replaces the slow-start algorithm at the start of a TCP connection or after a retransmission timeout, and (2) the dynamic recovery algorithm, which replaces the fast recovery algorithm to recover packet losses when a TCP connection is congested. Both algorithms require modifications only to the sender side of the TCP implementation. Simulation is used to evaluate the performance of the algorithms. The simulation experiments are conducted using the ns simulator to facilitate comparisons with Tahoe, Reno, New-Reno, SACK, and FACK TCP. The simulation results show that the new scheme performs at least as well as SACK and FACK TCP, which in turn consistently outperform TCP Tahoe and Reno. Furthermore, the implementation of the new scheme is simpler than that of SACK and FACK.

Thii paper describes a new protocol to support IP multicast for mobile hosts in an IP internetwork.It uses the basic unicast routing capability of IETF Mobile IP as a foundation, and leverages existing IP multicast to provide multicast services for mobile hosts as well.We believe that the resulting scheme is simple, scalable, transparent, and to the extent possible, independent of the underlying multicast routing facility.Discrete-event simulation was used as the vehicle for a LLproof of concept" debugging of the protocol, as well as to determine its performance characteristics.A key feature of the new protocol is the use of designated multicast service providers (DMSPs) to address the scalability issues of mobile multicast.Our simulation results suggest distinct performance advantages of our protocol using DMSPs over two other approaches proposed for the mobile multicast problem, namely remote subscription and bi-directional tunnelling, particularly as the number of mobile group members increases.

This paper presents a workload characterization study for Internet Web servers. Six different data sets are used in the study: three from academic environments, two from scientific research organizations, and one from a commercial Internet provider. These data sets represent three different orders of magnitude in server activity, and two different orders of magnitude in time duration, ranging from one week of activity to one year. The workload characterization focuses on the document type distribution, the document size distribution, the document referencing behavior, and the geographic distribution of server requests. Throughout the study, emphasis is placed on finding workload characteristics that are common to all the data sets studied. Ten such characteristics are identified. The paper concludes with a discussion of caching and performance issues, using the observed workload characteristics to suggest performance enhancements that seem promising for Internet Web servers.

This paper presents a simulation study of two proposed usage-based pricing strategies (one static, one dynamic) for computer networks. In the static pricing strategy evaluated, a price per packet is assigned and held constant over time. Results from the static strategy simulations show that network utilization decreases as the price per packet increases. Revenue, on the other hand, first increases and then falls as the price per packet increases. In the dynamic pricing strategy simulations, the price varies over time as a result of the user demand for bandwidth. As the number of users on the network increases, the price per unit of bandwidth increases. While the behaviour of the dynamic pricing scheme makes it appear promising as a pricing framework for packet-switched networks, further work is clearly required to better address the tradeoffs between network utilization, revenue, and network efficiency.

This paper addresses the problem of multicast routing in ATM networks. Formal experimental methods are used to evaluate the relative performance of three simple multicast routing algorithms on simple mesh-based networks, using call-level simulation. The performance study focuses on the scalability of multicast routing algorithms with respect to network size, the level of multicast traffic on the network, and the size of multicast groups. The results show that the choice of a multicast routing algorithm has a major impact on network performance. Shared link algorithms, which route multicast calls along common link channels whenever possible, improve the overall call blocking performance, reduce the call blocking bias against large multicast groups, and scale much better with network size. However, the improved call blocking performance for multicast calls comes at the expense of increased call blocking for unicast calls.

The phenomenal growth in popularity of the World Wide Web (WWW, or the Web) has made WWW traffic the largest contributor to packet and byte traffic on the NSFNET backbone. This growth has triggered recent research aimed at reducing the volume of network traffic produced by Web clients and servers, by using caching, and reducing the latency for WWW users, by using improved protocols for Web interaction.Fundamental to the goal of improving WWW performance is an understanding of WWW workloads. This paper presents a workload characterization study for Internet Web servers. Six different data sets are used in this study: three from academic (i.e., university) environments, two from scientific research organizations, and one from a commercial Internet provider. These data sets represent three different orders of magnitude in server activity, and two different orders of magnitude in time duration, ranging from one week of activity to one year of activity.Throughout the study, emphasis is placed on finding workload invariants: observations that apply across all the data sets studied. Ten invariants are identified. These invariants are deemed important since they (potentially) represent universal truths for all Internet Web servers. The paper concludes with a discussion of caching and performance issues, using the invariants to suggest performance enhancements that seem most promising for Internet Web servers.

Loss-load curves are a feedback mechanism for rate-based congestion control in packet-switched networks. The novel feature of loss-load curves is that they provide a direct mathematical relationship between offered load and the level of packet loss within the network. With this feedback information, traffic sources can dynamically compete for and share the available bandwidth on a network. Sources can use loss-load curve information to maximize raw throughput, to optimize file transfer response time, or to control end-to-end packet loss within the network. The focus of this paper is on the mathematical underpinnings of the loss-load model. The paper presents three key theoretical results for the loss-load model, and then discusses the advantages and disadvantages of the loss-load approach to dynamic bandwidth allocation. Formal proofs of the theoretical results are given.

Adaptive video coding algorithms are digital video compression algorithms that can adapt the encoding of a video stream dynamically based on the amount of bandwidth available on the network. While such algorithms are more complicated than traditional video coding algorithms, they are attractive because of their inherent robustness to changes in network load (i.e. network congestion). Adaptive video coding algorithms seem particularly suitable for high speed network environments, such as B-ISDN/ATM, that offer Available Bit Rate (ABR) services. The goal of this paper is to assess the role that adaptive video coding algorithms will play in future high speed networks. The paper presents a simple mathematical model and analysis of several hypothetical video coding algorithms for high speed networks, and a simulation study of one such adaptive video coding algorithm that we have implemented in a local area network environment. The results show that adaptive video coding algorithms are indeed robust across a wide range of network loads. More importantly, however, the results suggest that the domain of adaptive video coding algorithm is quite narrow: moderately to heavily loaded networks with speeds on the order of 10 Mbps and 100 Mbps. As a result, adaptive video coding algorithms will likely play only a limited role in future high speed networks.

Network traffic measurement and workload characterization are key steps in the workload modeling process. Much has been learned through network measurement and workload modeling in the last ten years, but new challenges are now at the forefront: measuring network traffic in the Internet environment, understanding the implications of network traffic structure (e.g., self-similarity, autocorrelation, long range dependence), and accurate modeling of network traffic workloads for high speed network environments.This "hot topic" session brings together three prominent speakers to address each of these topics, in turn.

Loss-load curves are a recently proposed feedback mechanism for rate-based congestion control in datagram computer networks. In the loss-load model, packet loss inside the network is a direct function of sender transmission rates, and senders have the responsibility of choosing their own transmission rate based on the loss-load tradeoff curve provided by the network. Earlier work has provided the mathematical basis for the loss-load model and provided preliminary simulation results demonstrating its responsiveness, fairness, and stability. The loss-load model works well for simple network environments where each source has many packets to transmit, and wishes to maximize raw throughput.In this paper, the applicability of loss-load curves is extended to senders with more general network traffic requirements, such as file transfer activities. In particular, the paper considers senders wishing to minimize the response time for a transfer of N packets. When N is small, a sender chooses a low transmission rate, to minimize the probability of packet loss. When N is large, a higher transmission rate is chosen to increase raw throughput. The loss-load curve mechanism provides necessary and sufficient information for senders at the edges of the network to choose a proper transmission rate based on their traffic requirements and the current network load.Simulation results show that the loss-load mechanism is effective in achieving high network utilization and minimizing response time for file transfers. The loss-load model also does a good job of providing equitable service to senders with different traffic workloads and different round trip times.

Congestioncontrol is an important problem in high-speed computer networks.Networks must limit the packet load that they accept to avoid excessive packet loss and delay.Hosts must limit their transmission rates to match what the network can handle at any given time or else suffer high sender cooperation, to provide protection from misbehaving senders, and to keep offered load close to network capacity at all times.Our analytic and simulation results show that the algorithm converges, without oscillation, to a small and stable overload, and that it provides a bounded and predictable level of packet loss to cooperating senders.

A cached state approach to connection management is described. The approach, based on the request-response model of communication, uses implicit connection setup, and dynamic caching of communication state records beyond the time needed by the transport layer. The caching of connection records can reduce callbacks on isolated requests if secure or authenticated communication is required. The cached state approach can provide low latency for request-response interaction by eliminating unnecessary round-trip times from the cost of communication, and does so with little added cost. The effectiveness of the approach depends on locality in communication. Measurements in the V distributed system show the cached-state approach fits well with the observed communication behavior.>

Communication in modern distributed systems demands low-latency transaction-oriented communication rather than stream-oriented communication as in the past. The performance and functionality of current standard transport protocols has become a major limitation in the move to higher-speed networks and larger-scale, more sophisticated distributed systems. An overview is presented of the versatile message transaction protocol (VMTP) developed to address these limitations. The authors then present measurements of VMTP performance in actual use in the V distributed system, showing that its performance matches their objectives.>

Communication systems are undergoing a change in use from stream to request-response or transaction communication. In addition, communication systems are becoming increasingly based on high-speed, low delay, low error rate channels. These changes call for a new generation of networks, network interfaces, and transport protocol design. The performance characteristics of request-response protocols on these high-performance networks should guide the design of this new generation, yet relatively little data of this nature is available.

The scheduler used in an operating system is an important factor in the performance of the system under heavy load. This paper describes the scheduling philosophy employed in the UNIX operating system and outlines the standard scheduling strategies. Modified strategies which address deficiencies in the standard strategies are described. The effectiveness of these modified strategies is assessed by means of performance experiments.