Large language models (LLMs) show remarkable promise for democratizing automated reasoning by generating formal specifications. However, a fundamental tension exists: LLMs are probabilistic, while formal verification demands deterministic guarantees. This paper addresses this epistemological gap by comprehensively investigating failure modes and uncertainty quantification (UQ) in LLM-generated formal artifacts. Our systematic evaluation of five frontier LLMs reveals Satisfiability Modulo Theories (SMT) based autoformalization's domain-specific impact on accuracy (from +34.8\% on logical tasks to -44.5\% on factual ones), with known UQ techniques like the entropy of token probabilities failing to identify these errors. We introduce a probabilistic context-free grammar (PCFG) framework to model LLM outputs, yielding a refined uncertainty taxonomy. We find uncertainty signals are task-dependent (e.g., grammar entropy for logic, AUROC>0.93). Finally, a lightweight fusion of these signals enables selective verification, drastically reducing errors (14-100\%) with minimal abstention, transforming LLM-driven formalization into a reliable engineering discipline.

Energy markets enable matching supply and demand through inter- and intra-region electricity trading. Due to the interconnected nature of the energy markets, the supply-demand constraints in one region can impact prices in another connected region. To incorporate these spatiotemporal relationships, we propose a novel graph neural network architecture incorporating multidimensional time-series features to forecast price (node attribute) and energy flow (edge attribute) between regions simultaneously. To the best of our knowledge, this paper is the first attempt to combine node and edge level forecasting in energy markets. We show that our proposed approach has a mean absolute prediction percentage error of 12.8%, which significantly beats the state-of-the-art baseline techniques.

Accurately predicting the energy consumption of an electric vehicle (EV) under real-world circumstances (such as varying road, traffic, weather conditions, etc.) is critical for a number of decisions like range estimation and route planning. A major concern for electric vehicle owners is the uncertain nature of the battery consumption. This results in the “range anxiety” and reluctance from users for mass adoption of EVs, since they are concerned about untimely drainage of battery. Even at the organizational level, a company running a fleet of electric vehicles must understand the battery consumption profiles accurately for tasks such as route and driver planning, battery sizing, maintenance planning, etc.

Renewable energy is becoming an important power source for data centers, especially with the zero-carbon waste pledges made by big cloud providers. However, one of the main challenges of renewable energy sources is the high variability of power produced. Traditional approaches such as batteries or transmitting to the grid fall short on scale, overhead, or "green-ness". We propose Virtual Battery: instead of adapting the availability of power to match the computation demand we shift computational demand to meet the availability of power. Virtual batteries shift demand by requiring applications to either be flexible and delay-tolerant or proactively migrating to where power is (going to be) available. We show that using multiple virtual battery sites in combination can meet the needs of modern applications. Moreover, we show how an intelligent network and power aware co-scheduler can not only provide availability despite variability but also help mitigate migration related network overhead by over 30% in total and 4.2x at peak.

This paper presents a deep learning approach for a versatile Micro-climate prediction framework (DeepMC). Micro climate predictions are of critical importance across various applications, such as Agriculture, Forestry, Energy, Search & Rescue, etc. To the best of our knowledge, there is no other single framework which can accurately predict various micro-climate entities using Internet of Things (IoT)data. We present a generic framework (DeepMC) which predicts various climatic parameters such as soil moisture, humidity, windspeed, radiation, temperature based on the requirement over a period of 12 hours - 120 hours with a varying resolution of 1 hour - 6hours, respectively. This framework proposes the following new ideas: 1) Localization of weather forecast to IoT sensors by fusing weather station forecasts with the decomposition of IoT data at multiple scales and 2) A multi-scale encoder and two levels of attention mechanisms which learns a latent representation of the interaction between various resolutions of the IoT sensor data and weather station forecasts. We present multiple real-world agricultural and energy scenarios, and report results with uncertainty estimates from the live deployment of DeepMC, which demonstrate that DeepMC outperforms various baseline methods and reports 90%+ accuracy with tight error bounds.

With the increasing adaption of solar energy worldwide, there is a huge interest to develop systems that help drive efficiency during manufacturing and ongoing operations. Due to various real-world conditions and processes, solar panels develop faults during their manufacturing and operations. The objective of this work is to build an End-to-End Fault Detection system to detect and localize faults in solar panels based on their Electroluminescence (EL) Imaging. Today, the majority of fault detection happens through manual inspection of EL images. To this end, we propose the design and implementation of an end-to-end system that firstly divides the solar panel into individual solar cells and then passes these cell images through a classification + detection pipeline for identifying the fault type and localizing the faults inside a cell. We propose a hybrid architecture that contains an ensemble of multiple CNN model architectures for classification and detection. The ensemble is capable of serving both - monocrystalline and polycrystalline solar panels. The proposed system significantly helps in increasing the efficiency of solar panels and reducing warranty and repair costs. We demonstrate the performance of the proposed system using an open EL image dataset with 95% of cell-level fault prediction accuracy and high recall. The proposed algorithms are applicable and can be extended for other solar applications that use RGB, EL, or thermal imaging techniques.

Ahead-of-time forecasting of incident solar-irradiance on a panel is indicative of expected energy yield and is essential for efficient grid distribution and planning. Traditionally, these forecasts are based on meteorological physics models whose parameters are tuned by coarse-grained radiometric tiles sensed from geo-satellites. This research presents a novel application of deep neural network approach to observe and estimate short-term weather effects from videos. Specifically, we use time-lapsed videos (sky-videos) obtained from upward facing wide-lensed cameras (sky-cameras) to directly estimate and forecast solar irradiance. We introduce and present results on two large publicly available datasets obtained from weather stations in two regions of North America using relatively inexpensive optical hardware. These datasets contain over a million images that span for 1 and 12 years respectively, the largest such collection to our knowledge. Compared to satellite based approaches, the proposed deep learning approach significantly reduces the normalized mean-absolute-percentage error for both nowcasting, i.e. prediction of the solar irradiance at the instance the frame is captured, as well as forecasting, ahead-of-time irradiance prediction for a duration for upto 4 hours.

Soil moisture and temperature are key inputs to several precision agricultural applications such as irrigation scheduling, identifying crop health, pest and disease prediction, yield and acreage estimation, etc. The existing remote sensing satellites based soil moisture products such as SMAP are of coarse resolution and physics based land surface model such as NL-DAS, GLDAS are also of coarse resolution as well as not available for real time applications. Keeping this in focus, we have customized high resolution land data assimilation system (HRLDAS) for India. The customization involve: (1) use of Global Data Assimilation System (GDAS) dataset for dynamic forcing fields, (2) ability to ingest local information about the soil characteristics (3) high resolution USGS land-cover and other static datasets, amongst others. In this paper, we present the performance of the customized model against SMAP soil moisture data and local sensors observations. The first results from the comparison shows a significantly reduced errors in model results. The RMSE for LSM generated outputs are less than 2%, whereas SMAP gives 4-5 % error for soil moisture.

Real-time monitoring of agricultural crops is an important exercise because of the huge economic impact. Identification of crop during early stage of the crop cycle can help formulate better agriculture policies and management strategies. In this context, the objective of this article is to evaluate the potential of Sentinel-1 Synthetic Aperture Radar (SAR) and Sentinel-2 optical imagery in crop identification for an Indian region. A multi-class classification algorithm based on random forest is applied to the features extracted from the above mentioned satellite data sets. Initial experimental suggest that the Sentinel-1 SAR data is promising in achieving high classification accuracy (85%).

The impact of soiling on solar panels is an important and well-studied problem in renewable energy sector. In this paper, we present the first convolutional neural network (CNN) based approach for solar panel soiling and defect analysis. Our approach takes an RGB image of solar panel and environmental factors as inputs to predict power loss, soiling localization, and soiling type. In computer vision, localization is a complex task which typically requires manually labeled training data such as bounding boxes or segmentation masks. Our proposed approach consists of specialized four stages which completely avoids localization ground truth and only needs panel images with power loss labels for training. The region of impact area obtained from the predicted localization masks are classified into soiling types using the webly supervised learning. For improving localization capabilities of CNNs, we introduce a novel bi-directional input-aware fusion (BiDIAF) block that reinforces the input at different levels of CNN to learn input-specific feature maps. Our empirical study shows that BiDIAF improves the power loss prediction accuracy by about 3% and localization accuracy by about 4%. Our end-to-end model yields further improvement of about 24% on localization when learned in a weakly supervised manner. Our approach is generalizable and showed promising results on web crawled solar panel images. Our system has a frame rate of 22 fps (including all steps) on a NVIDIA TitanX GPU. Additionally, we collected first of it's kind dataset for solar panel image analysis consisting 45,000+ images.

Managing the Quality-of-Experience (QoE) of video streaming for wireless clients is becoming increasingly important due to the rapid growth of video traffic on wireless networks. The inherent variability of the wireless channel as well as the Variable Bit Rate (VBR) of the compressed video streams make QoE management a challenging problem. In this paper, we investigate scheduling algorithms to transmit multiple video streams from a base station to mobile clients. We present an epoch-by-epoch framework to fairly allocate wireless transmission slots to streaming videos. In each epoch, our scheme reduces the vulnerability to stalling by allocating slots to videos in a way that maximizes the minimum “playout lead” across all videos. We show that the problem of allocating slots fairly is NP-complete even for a constant number of videos. We then present a fast lead-aware greedy scheduling algorithm. Our greedy algorithm is optimal when the channel quality of a user remains unchanged within an epoch. Our experimental results, based on public MPEG-4 video traces and wireless channel traces that we collected from a WiMAX test-bed, show that the lead-aware greedy approach results in a fair distribution of stalls across the clients when compared to other algorithms, while still maintaining similar or fewer average number of stalls per client.

TV Whitespaces, recently opened up by the Federal Communications Commission (FCC) for unlicensed use, are seen as a potential cellular offload and/or standalone mechanism, especially in dense metros where the demand for throughput is high. In this paper, we use real data collected from whitespaces databases to empirically demonstrate features unique to whitespaces-power-spectrum tradeoff and spatial variation in spectrum availability. From this study, we conclude the need for whitespaces-specific adaptations to cellular networks so as to be able to extract maximum throughput and guarantee reliability. To tackle the effects of the power-spectrum tradeoff, we propose a novel base-station design that specifically uses low-power transmitters as a means to maximize throughput. This design co-locates and networks together many low-powered mode-I devices to act as a multiple-antenna array. We estimate the size of the array required to meet typical rate targets, and show that the array design significantly outperforms traditional designs in terms of throughput for a given cost. We then turn our attention to spatial variability and study its impact on the problem of locating base stations in a whitespaces network. Here, we propose spectrum-aware placement algorithms for whitespaces, which account for this spatial variability along with key parameters like user density. We show that such algorithms clearly outperform traditional placement algorithms and improve network coverage in this band.

The next-generation mobile devices include smart watches, wristbands, wearables (e.g., Google Glass), etc. In the future such devices will constitute a large fraction of the total devices available in the market [1]. Latest studies confirm that location-based services are the most requested feature by developers with a market share of 13B in 2013 and have expected exponential growth [2]. Future location-based applications/services will use the data generated by the new mobile devices for providing enhanced user experience. This paper presents a vision of such next-generation location-based services, which we call LBS 2.0. We present the challenges and opportunities that LBS 2.0 will pose for mobile data management.

A significant amount of energy is wasted by electrical appliances when they operate inefficiently either due to anomalies and/or incorrect usage. To address this problem, we present SocketWatch - an autonomous appliance monitoring system. SocketWatch is positioned between a wall socket and an appliance. SocketWatch learns the behavioral model of the appliance by analyzing its active and reactive power consumption patterns. It detects appliance malfunctions by observing any marked deviations from these patterns. SocketWatch is inexpensive and is easy to use: it neither requires any enhancement to the appliances nor to the power sockets nor any communication infrastructure. Moreover, the decentralized approach avoids communication latency and costs, and preserves data privacy. Real world experiments with multiple appliances indicate that SocketWatch can be an effective and inexpensive solution for reducing electricity wastage.

Over-the-top estimation of available bandwidth (AB) in a network path has been well studied for wired networks. The AB of a path denotes its slack capacity, i.e., the bandwidth available for use in the path without impacting the existing traffic. This estimation problem has been receiving attention only recently in cellular networks, which are increasingly becoming one of the main modes of access for a large number of applications. In this paper, we discuss the challenges posed by the problem, and why existing techniques developed for wired networks cannot be applied. We show that, interestingly, it may not even be feasible to estimate AB using over-the-top approaches under certain conditions, even when the wireless channel and traffic conditions are non-varying. We then present a novel AB estimation technique for cellular networks, which typically use proportional fair scheduling at base stations. When the wireless channel and traffic conditions are non-varying, our technique can accurately determine AB when it exceeds the “fair share” due to a new flow. We also extend the basic technique for estimation under conditions that are time-varying. The proposed methods can as well be used when one or more bottleneck links in a network path are fair-scheduled using algorithms such as weighted-fair queueing. We evaluate our methods using simulations and over operational networks, and present the results. In simulations, our technique is capable of detecting AB close to 90% of the time under feasible conditions even with bursty traffic.

Packet reordering has now become one of the most significant bottlenecks in next-generation switch designs. A switch practically experiences a reordering delay contagion, such that a few late packets may affect a disproportionate number of other packets. This contagion can have two possible forms. First, since switch designers tend to keep the switch flow order, i.e., the order of packets arriving at the same switch input and departing from the same switch output, a packet may be delayed due to packets of other flows with little or no reason. Further, within a flow, if a single packet is delayed for a long time, then all the other packets of the same flow will have to wait for it and suffer as well. In this paper, we suggest solutions against this reordering contagion. We first suggest several hash-based counter schemes that prevent inter-flow blocking and reduce reordering delay. We further suggest schemes based on network coding to protect against rare events with high queueing delay within a flow. Last, we demonstrate using both analysis and simulations that the use of these solutions can indeed reduce the resequencing delay. For instance, resequencing delays are reduced by up to an order of magnitude using real-life traces and a real hashing function.

We design and implement a novel system called Async, which enables a mobile network operator (MNO) to efficiently manage the growth of mobile data by leveraging the delay-elastic nature of certain applications and the price-sensitive nature of certain users. Specifically, Async introduces an alternate “asynchronous” content-delivery paradigm for heavy content (e.g., videos), and facilitates an MNO to negotiate with users a delay in delivery in exchange for appropriate incentives. The MNO uses the negotiated delays to actively manage Async flows to reduce congestion and improve the quality-of-experience (QoE) of both delayed and regular flows. We show that in comparison to state-of-the-art, Async's network-based flow management enhances QoE for more than 30% of the regular flows, with up to 60% improvement in per-flow QoE metric, while still meeting the negotiated delivery times of 95% of the delayed flows. Async also lowers the delivery times of delayed flows by ∼67% and significantly increases robustness to traffic unpredictability. Our design is robust to disconnections and does not require any modifications to existing network infrastructure and protocols. Our prototype deployment (using Apache's mod_proxy and an Android app) on live networks confirms Async's efficacy in meeting EDTs for diverse deployment scenarios.

In this paper, we study the problem of efficient video delivery over the cellular downlink. The key objective is to maximize the Quality of Experience (QoE) of the user, as measured by application level metrics such as the buffering ratio and low bit rate ratio. We present a two-tiered solution with a standard base-station scheduler that works on a per-packet basis and a Video Management System (VMS) that works at the granularity of thousands of video frames. The video management system uses knowledge of the video playout curves and future channel states to develop a scheduling policy that is feasibility optimal. The algorithms are simple and leverage recent results on real-time scheduling in wireless networks. We evaluate the performance of our algorithms using real video traces and a standard channel model. The VMS ensures that the per-user QoE guarantees are maintained, as compared with a standard PF scheduler that is oblivious to application level QoE requirements.

The Indian electricity sector, despite having the world's fifth largest installed capacity, suffers from a 12.9% peaking shortage. This shortage could be alleviated, if a large number of deferrable loads, particularly the high powered ones, could be moved from on-peak to off-peak times. However, conventional Demand Side Management (DSM) strategies may not be suitable for India as the local conditions usually favor inexpensive solutions with minimal dependence on the pre-existing infrastructure. In this work, we present a completely autonomous DSM controller called the nPlug. nPlug is positioned between the wall socket and deferrable load(s) such as water heaters, washing machines, and electric vehicles. nPlugs combine local sensing and analytics to infer peak periods as well as supply-demand imbalance conditions. They schedule attached appliances in a decentralized manner to alleviate peaks whenever possible without violating the requirements of consumers. nPlugs do not require any manual intervention by the end consumer nor any communication infrastructure nor any enhancements to the appliances or the power grids. Some of nPlug's capabilities are demonstrated using experiments on a combination of synthetic and real data collected from plug-level energy monitors. Our results indicate that nPlug can be an effective and inexpensive technology to address the peaking shortage. This technology could potentially be integrated into millions of future deferrable loads: appliances, electric vehicle (EV) chargers, heat pumps, water heaters, etc.

The increasing demand for high-quality streaming video delivered to mobile clients necessitates efficient bandwidth utilization and allocation at not only the wireless channel but also the wired backhaul of broadband cellular networks. In this context, we propose techniques for increasing the link utilization and enhancing the quality-of-experience (QoE) for end users while multiplexing video streams over a wired link. For increasing the link utilization, we present a generic multi-tiered bandwidth estimation and scheduling scheme that can guarantee lower bounds on loss for flows at lower tiers. This scheme can be used for supporting heterogeneous loss classes, providing differentiated losses for different layers of video streams, or providing per-flow guarantees using lower aggregate bandwidth than schemes proposed in the literature. For enhancing the end-user QoE, we present a scheme for minimizing correlated losses and improving the smoothness of video quality by minimizing the maximum loss suffered by any logical unit of a stream and also the variability in loss across the length of the stream. In extensive simulations performed using video sources encoded in various formats, our multi-tiered scheme could lower the estimated bandwidth and improve statistical multiplexing gains by up to 25% with two and three classes and over 30% in the context of providing per-flow guarantees and differentiated loss for different layers. Our loss-minimization approach could lower the maximum loss by a factor of five and the loss variance by more than an order of magnitude.

Managing the Quality-of-Experience (QoE) of video streaming for wireless clients is becoming increasingly important due to the rapid growth of video traffic on wireless networks. The inherent variability of the wireless channel as well as the Variable Bit Rate (VBR) of the compressed video streams make QoE management a challenging problem. Prior work has studied this problem in the context of transmitting a single video stream. In this paper, we investigate multiplexing schemes to transmit multiple video streams from a base station to mobile clients that use number of playout stalls as a performance metric. In this context, we present an epoch-by-epoch framework to fairly allocate wireless transmission slots to streaming videos. In each epoch our scheme essentially reduces the vulnerability to stalling by allocating slots to videos in a way that maximizes the minimum `playout lead' across all videos. Next, we show that the problem of allocating slots fairly is NP-complete even for a constant number of videos. We then present a fast lead-aware greedy algorithm for the problem. Our choice of greedy algorithm is motivated by the fact that this algorithm is optimal when the channel quality of a user remains unchanged within an epoch (but different users may experience different channel quality). Moreover, our experimental results based on public MPEG-4 video traces and wireless channel traces that we collected from a WiMAX test-bed show that the lead-aware greedy approach performs a fair distribution of stalls across the clients when compared to other algorithms, while still maintaining similar or lower average number of stalls per client.

TV Whitespaces, recently opened up by the FCC for unlicensed use by wireless devices, are seen as a potential cellular offload solution, especially in dense metros. However, under the new database-driven guidelines, there are typically very few whitespace bands available in such dense metros to a high-powered fixed device, which plays the role of a cellular base station in whitespaces. To address the lack of degrees-of-freedom (DoF) with this traditional architecture of one high-powered serving device, we propose a novel base station design that co-locates and networks together many low-powered devices to act as a multiple-antenna array. Lower-powered whitespace devices have access to more spectral DoF, a property that is unique to whitespaces. In the first part of the paper, we solve an array design problem where we estimate the size of the array required to meet long-term (worst-case) throughput targets. Using extensive simulations, we show that by effectively exploiting both spatial and spectral DoF, the array design outperforms the traditional design in most network conditions. Specifically, the proposed design can support throughputs of the order of a WiMAX cell running applications such as high-definition television. In the second part of the paper, we turn our attention to the operational aspects of such a design. Recognizing that the proposed array can potentially contain hundreds of elements, we propose a dynamic ON-OFF power control algorithm that operates in conjunction with the MaxWeight data scheduling algorithm and responds to the current network state - queues and channels - of the system, thus making the system power-efficient.

Video objects are much larger in size than traditional web objects and tend not to be viewed in entirety. Hence, caching them partially is a promising approach. Also, the projected growth in video traffic over wireless cellular networks calls for resource-efficient caching mechanisms in the wireless edge to lower traffic over the cellular backhaul and peering links and their associated costs. An evaluation of traditional partial caching solutions proposed in the literature shows that known solutions are not robust to video viewing patterns, increasing object pool size, changing object popularity, or limitation in the resources available for caching at the wireless network elements. In this paper, to overcome the limitations, we propose a novel approach that adopts a flexible segmentation policy and generalizes both LRU and LFU when applied to segmented accesses, and in our simulations, is shown to significantly lower wireless backhaul traffic (by around 20-30% and in some cases even higher).

The Internet is witnessing a rapid increase in video traffic. Due to the scalability and the cost-savings offered by cloud-computing, Internet video service providers are increasingly delivering their content from multi-tenant cloud data centers. One of the major challenges faced by such a video service provider is the management of the Quality-of-Experience (QoE) of the end-users in the presence of Variable Bit Rate (VBR) video flows, time varying network conditions in the Internet, and the bounded egress bandwidth provided by the data center. To this end, we present InSite, a light-weight and easy-to-deploy solution for managing the QoE of a set of video flows of a service provider, which are served from a data center. InSite is deployed at the egress of a data center, between the video servers and the clients, and manages the video flows that are transmitted over TCP. The solution uses a novel generalized binary search technique to concurrently search for the appropriate flow rates for a set of flows, with the goal of maximizing the QoE-fairness across the flows, as opposed to TCP-fairness. The search takes into account the total egress bandwidth allocated for the set of video flows at the data center, the unknown and possibly time-varying capacities of any remote bottleneck links, and the playout buffer sizes of the video flows. The solution is also designed to operate with minimal modifications to the video servers and the clients. In our evaluations using extensive ns-3 simulations and a testbed implementation for serving videos over TCP, we observe that deploying InSite achieves several folds reduction in playout stalls over a system without InSite.

Increased commodity use of mobile devices has the potential to enable mission-critical monitoring applications. However, these mobile-enabled monitoring applications have to often work in environments where a delay-tolerant network (DTN) is the only feasible communication paradigm. Detection of complex (composite) events is fundamental to monitoring applications. However, the existing plan-based CED techniques are mostly centralized, and hence are inherently unscalable for DTNs. In this paper, we create Comet â" a decentralized plan-based, efficient and scalable CED for DTNs. Comet shares the task of detecting complex events (CEs) among multiple nodes, with each node detecting a part of the CE by aggregating two or more primitive events or sub-CEs. Comet uses a unique h-function to construct cost and delay efficient CED trees. As finding an optimal CED plan requires exponential-time, Comet finds near-optimal detection plans for individual CEs through a novel multi-level push-pull conversion algorithm. Performance results show that Comet reduces cost by up to 89% compared to pushing all primitive events and over 60% compared to a two-level exhaustive search algorithm.

Most enterprises today run their applications on virtual machines (VMs). VM mobility - both live and offline, can provide enormous flexibility and also bring down OPEX (Operational Expenditure) costs. However, both live and offline migration of VMs is still limited to within a local network because of the complexities associated with cross subnet live and offline migration. These complexities mainly arise from the hierarchical addressing used by various layer 3 routing protocols. For cross data center VM mobility, virtualization vendors require that the network configuration of the new data center where a VM migrates must be similar to that of the old data center. This severely restricts wide spread use of VM migration across data center networks. For offline migration, the above limitations can be overcome by reconfiguring IP addresses for the migrated VMs. However, even this effort is non-trivial and time consuming as these IP addresses are embedded in various configuration files inside these VMs. As enterprises grow and new data centers emerge in different geographic locations, there is a need to interconnect these data centers in a way that allows seamless VM mobility. In this context, we present CrossRoads - a network fabric that provides layer agnostic and seamless live and offline VM mobility across multiple data centers. We leverage software defined networking and implement an OpenFlow based prototype of CrossRoads. CrossRoads extends the idea of location independence based on pseudo addresses proposed in recent research to work with a control plane overlay of OpenFlow network controllers in various data centers. We evaluate CrossRoads on an innovative testbed that leverages nested virtualization to emulate two data centers. Our results confirm that CrossRoads has negligible performance overhead as compared to a Default layer 2 network - its average performance was no worse than 2.3% as compared to Default fabric across all experiments. In some experiments, it even outperformed the Default by up to 30%.

This paper considers the problem of vehicular traffic density estimation, utilizing the information cues present in the cumulative acoustic signal acquired from a roadside-installed single microphone. This cumulative signal comprises several noise signals such as tire noise, engine noise, engine-idling noise, occasional honks, and air turbulence noise of multiple vehicles. The occurrence and mixture weightings of these noise signals are determined by the prevalent traffic density conditions on the road segment. For instance, under a free-flowing traffic condition, the vehicles typically move with medium to high speeds and thereby produce mainly tire noise and air turbulence noise and less engine-idling noise and honks. For slow-moving congested traffic, the cumulative signal will largely be dominated by engine-idling noise and honks; air turbulence and tire noises will be inconspicuous. Furthermore, these various noise signals have spectral content that are very different from each other and, hence, can be used to discriminate between the different traffic density states that lead to them. Therefore, in this work, we extract the short-term spectral envelope features of the cumulative acoustic signals and model their class-conditional probability distributions, conditioned on one of the three broad traffic density states, i.e., Jammed (0-10 km/h), Medium-Flow (10-40 km/h), and Free-Flow (40 km/h and above) traffic. While these states are coarse measures of the average traffic speed, they nevertheless can provide useful traffic density information in the often-chaotic and nonlane-driven traffic conditions of the developing geographies, where other techniques (magnetic loop detectors) are inapplicable. Based on these learned distributions, we use a Bayes' classifier to classify the acoustic signal segments spanning a duration of 5-30 s, which results in a high classification accuracy of ~95%. Using a discriminative classifier such as a support vector machine (SVM) results in further classification accuracy gains over the Bayes' classifier.

Wireless networks (including wireless mesh networks) provide opportunities for using multiple paths. Multihoming of hosts, possibly using different technologies and providers, also makes it attractive for end-to-end transport connections to exploit multiple paths. In this paper, we propose a multipath transport protocol, based on a carefully crafted set of enhancements to TCP, that effectively utilizes the available bandwidth and diversity provided by heterogeneous, lossy wireless paths. Our Multi-Path LOss-Tolerant (MPLOT) transport protocol can be used to obtain significant goodput gains in wireless networks, subject to bursty, correlated losses with average loss rates as high as 50%. MPLOT is built around the principle of separability of reliability and congestion control functions in an end-to-end transport protocol. Congestion control is performed separately on individual paths, and the reliability mechanism works over the aggregate set of paths available for an end-to-end session. MPLOT distinguishes between congestion and link losses through Explicit Congestion Notification (ECN), and uses Forward Error Correction (FEC) coding to recover from data losses. MPLOT uses a dynamic packet mapping based on the current path characteristics to choose a path for a packet. Use of erasure codes and block-level recovery ensures that in MPLOT the receiving transport entity can recover all data as long as a necessary number of packets in the block are received, irrespective of which packets are lost. We present a theoretical analysis of the different design choices of MPLOT and show that MPLOT chooses its policies and parameters such that a desirable tradeoff between goodput with data recovery delay is attained. We evaluate MPLOT, through simulations, under a variety of test scenarios and demonstrate that it effectively exploits path diversity in addition to efficiently aggregating path bandwidths while remaining fair to a conventional TCP flow on each path.

The increasing demand for high-quality streaming video delivered to mobile clients necessitates efficient bandwidth utilization and allocation at not only the wireless channel but also the wired backhaul of broadband wireless networks. In this context, we propose techniques for increasing the link utilization and enhancing the quality-of-experience (QoE) for end users while multiplexing video streams over a wired link. For increasing the link utilization, we present a generic multi-tiered bandwidth estimation and scheduling scheme that can guarantee lower bounds on loss for flows at lower tiers. This scheme can be used for supporting heterogeneous loss classes, differentiated losses for different layers of video streams, or per-flow guarantees using lower aggregate bandwidth than schemes proposed in the literature. For enhancing the end-user QoE, we present a scheme for minimizing correlated losses and improving the smoothness of video quality by minimizing the maximum loss suffered by any logical unit of a stream and also the variability in loss across the length of the stream. In simulations performed using MPEG-4 sources, our loss-minimization approach could lower the maximum loss by a factor of five and the loss variance by more than an order of magnitude. Our multi-tiered scheme could lower the estimated bandwidth and improve statistical multiplexing gains by 10-20% with three classes, 5-20% with two classes, and over 30% in the context of providing deterministic per-flow guarantees.

The publish-subscribe (pub-sub) paradigm is maturing and integrating into community-oriented collaborative applications. Because of this, pub-sub systems are faced with an event stream that may potentially contain large numbers of redundant and partial messages. Most pub-sub systems view partial and

The benefit of Class-of-Service (CoS) is an important topic in the "Network Neutrality" debate. Proponents of network neutrality suggest that over-provisioning is a viable alternative to CoS. We quantify the extra capacity requirement for an over-provisioned classless (i.e., best-effort) network compared to a CoS network providing the same delay or loss performance for premium traffic. We first develop a link model that quantifies this Required Extra Capacity (REC). For realistic traffic distributions (e.g., long-range dependent), we find the REC using ns-2 simulations of the CoS and classless links. Our primary contribution is in using these link models to quantify the REC for realistic network topologies under various scenarios including "closed loop" environments with traffic generated by TCP sources that adapt to the available capacity. We show that REC can be significant even when the proportion of premium traffic requiring performance assurances is small, a situation often considered benign for the over-provisioning alternative.

We consider lifetime quality-of-experience (QoE) management for video-on-demand (VoD) users in WiMAX networks. For efficient resource utilization while enhancing user experience, we propose run-time load balancing through joint scheduling among multiple base stations (BSs), referred to as macro scheduling. Macro scheduling employs a utility-maximization approach. To achieve long-term proportional fairness (PF) and manage lifetime QoE, user and flow utilities are modeled as functions of their past service rates, in addition to current channel conditions and bandwidth needs. We show that scheduling flows across multiple BSs jointly to achieve PF is NP-hard in the strong sense. Since approximation algorithms proposed in prior work are computationally expensive for online use, we design efficient heuristics that perform as well as the approximation algorithms. A simulation-based evaluation shows that our overall macro scheduling scheme can improve the number of satisfied users by up to 35% in comparison to other approaches, while only minimally sacrificing on throughput.

Forwarding decisions in routing protocols rely on information about the destination nodes provided by routing table states. When paths to a destination change, corresponding states become invalid and need to be refreshed with control messages for resilient routing. In large and highly dynamic networks, this overhead can crowd out the capacity for data traffic. For such networks, we propose the concept of weak state, which is interpreted as a probabilistic hint, not as absolute truth. Weak state can remain valid without explicit messages by systematically reducing the confidence in its accuracy. Weak State Routing (WSR) is a novel routing protocol that uses weak state along with random directional walks for forwarding packets. When a packet reaches a node that contains a weak state about the destination with higher confidence than that held by the packet, the walk direction is biased. The packet reaches the destination via a sequence of directional walks, punctuated by biasing decisions. WSR also uses random directional walks for disseminating routing state and provides mechanisms for aggregating weak state. Our simulation results show that WSR offers a very high packet delivery ratio ( ≥ 98%). Control traffic overhead scales asO(N), and the state complexity is Θ(N3/2), whereNis the number of nodes. Packets follow longer paths compared to prior protocols (OLSR , GLS-GPSR , ), but the average path length is asymptotically efficient and scales asO(√N). Despite longer paths, WSR's end-to-end packet delivery delay is much smaller due to the dramatic reduction in protocol overhead.

State signaling and maintenance mechanisms play crucial roles in communication network protocols. State is used to facilitate indirections in protocols such as routing. Design approaches for traditional state signaling mechanisms have been categorized into soft and hard state. In both approaches, the state is deterministic. Hence, we call both as having strong state semantics, or more crisply, refer to them as strong state. If the state tracks entities with dynamic nature, strong state rapidly becomes invalidated and needs to be refreshed explicitly through control packets. In this paper, we evaluate the recently proposed weak state. Weak state is a generalization of soft state that is characterized by probabilistic semantics and local updates. It is interpreted as a probabilistic hint and not absolute truth. Weak state also contains the confidence in the state value, which is a measure of the probability that the state remains valid. The confidence or the state semantics is decayed locally without the need for explicit state update traffic traversing the network. The local updates also help the protocol use better estimates for the state value. We define two metrics, pure distortion and informed distortion, to evaluate the consistency of the weak state paradigm and compare it against strong state. Pure distortion measures the average gap between the actual value of the state and the value maintained at a remote node. On the other hand, the use of confidence increases the protocol's ability to cope with even large pure distortion. The resulting effective distortion is captured by the informed distortion metric. Using mathematical analysis, we compare weak with strong state. Local updates reduce the pure distortion because the protocol uses the best estimate of state value. The informed distortion is also significantly less because the probabilistic confidence value hints the protocol if the state is invalid. The weak state mechanism can be used to build protocols (eg: WSR [1]), which systematically interpret the state information. The state itself can be mostly updated locally, with less frequent explicit update messages over the network (i.e. leading to dramatic reductions in control traffic).

This paper develops a distributed, threshold based MAC protocol for cooperative multi input multi output (MIMO) transmissions in distributed wireless systems. The protocol uses a thresholding scheme that is updated dynamically based on the queue length at the sending node to achieve low power transmissions while ensuring stability of the transmission queues at the nodes. Simulation results are provided to evaluate the performance of the proposed protocol and compare it against regular point to point as well as fixed group size cooperative MIMO MAC protocols.

Mobile-enabled social networks applications are becoming increasingly popular. Most of the current social network applications have been designed for high-end mobile devices, and they rely upon features such as GPS, capabilities of the world wide web, and rich media support. However, a significant fraction of mobile user base, especially in the developing world, own low-end devices that are only capable of voice and short text messages (SMS). In this context, a natural question is whether one can design meaningful social network-based applications that can work well with these simple devices, and if so, what the real challenges are. Towards answering these questions, this paper presents a social network-based recommender system that has been explicitly designed to work even with devices that just support phone calls and SMS. Our design of the social network based recommender system incorporates three features that complement each other to derive highly targeted ads. First, we analyze information such as customer's address books to estimate the level of social affinity among various users. This social affinity information is used to identify the recommendations to be sent to an individual user. Second, we combine the social affinity information with the spatio-temporal context of users and historical responses of the user to further refine the set of recommendations and to decide when a recommendation would be sent. Third, social affinity computation and spatio-temporal contextual association are continuously tuned through user feedback. We outline the challenges in building such a system, and outline approaches to deal with such challenges.

The enormous interest for peer-to-peer systems in recent years has prompted research into finding scalable and robust seeding and searching methods to support these overlay networks. Routing and search in these overlay networks have ranged from flooding-based unstructured techniques to structured ones mainly for popular and rate items respectively. In this paper, we propose a new method of establishing a virtual structure and introduce a technique to scalably route packets through an unstructured overlay network. We introduce virtual direction routing (VDR). VDR is a lightweight and scalable overlay network routing protocol that uses the concept of virtual directions to efficiently perform node information seeding and lookup. State information is replicated at nodes along virtual orthogonal lines originating from each node and periodically updated. When a path lookup is initiated, instead of flooding the network, query packets are also forwarded along virtual orthogonal lines until an intersection with the seeded state occurs. We show that VDR achieves high reachability with relatively low seed and search packet TTL even under high network churn. We also show that VDR scales well without imposing DHT-like graph structures (e.g., trees, rings, torus, coordinate-space) and the path stretch compared to random-walk protocols is very good. The tradeoff is added latency by choosing suboptimal paths.

In this paper, we investigate a fine-grain adaptive forward error correction (FGA-FEC) coding scheme for scalable video bitstreams. In our work, both the embedded source bitstream and the error-control codes are granularly adapted at block level in intermediate overlay nodes to satisfy heterogeneous users with both different video frame-rate/spatial resolution/quality preferences and different network connections. The proposed FGA-FEC scheme encodes and adapts the embedded source-coded bitstream in such a way that if part of the video source data is actively dropped, parity bits protecting that piece of data are also removed, yielding an efficient result without any transcoding.

In this paper we propose a distributed system for facilitating cooperative MIMO transmissions in networks without multiple antenna devices. MIMO diversity is achieved by employing groups of nodes in the vicinity of the source and destination to help with the transmission. The distributed sending nodes are assumed to have different carrier frequency offsets (CFO). Space-time block codes (STBC) and code combining are used to utilize spatial diversity. The estimation of multiple CFO and detector for STBC-coded data under multiple CFO are provided. The BER of the proposed system is shown and discussed. We also consider the energy consumption and compare it with other cooperative designs.

We propose a novel transport protocol that effectively utilizes available bandwidth and diversity gains provided by heterogeneous, highly lossy paths. Our Multi-Path LOss-Tolerant (MPLOT) protocol can be used to provide significant gains in the goodput of wireless mesh networks, subject to bursty, correlated losses with average loss-rates as high as 50%, and random outage events. MPLOT makes intelligent use of erasure codes to guard against packets losses, and a Hybrid-ARQ/FEC scheme to reduce packet recovery latency, where the redundancy is adaptively provisioned into both proactive and reactive FECs. MPLOT uses dynamic packet mapping based on current path characteristics, and does not require packets to be delivered in sequence to ensure reliability. We present a theoretical analysis of the different design choices of MPLOT and show that MPLOT makes an optimal trade-off between goodput and delay constraints. We test MPLOT, through simulations, under a variety of test scenarios and show that it effectively exploits path diversity in addition to aggregating path bandwidths. We also show that MPLOT is fair to single-path protocols like TCP-SACK.

The increased usage of directional methods of communications has prompted research into leveraging directionality in every layer of the network stack. In this paper, we explore the use of directionality in layer 3 to facilitate routing in highly mobile environments. We introduce mobile orthogonal rendezvous routing protocol (MORRP), a lightweight, but scalable routing protocol utilizing directional communications (such as directional antennas or free-space-optical transceivers) to relax information requirements such as coordinate space embedding, node localization, and mobility. This relaxation is done by introducing a novel concept called the directional routing table (DRT) which maps a set-of-IDs to each directional interface to provide probabilistic routing information based on interface direction. We show that MORRP achieves connectivity with high probability even in highly mobile environments while maintaining only probabilistic information about destinations. We also compare MORRP with various proactive, reactive, and position-based routing protocols using single omni-directional interfaces and 8 directional interfaces and show that MORRP gains over 10-14X additional goodput vs. traditional protocols and 15-20% additional goodput vs. traditional protocols using multiple interfaces.

The benefit of Class-of-Service (CoS) is an important topic in the "Network Neutrality" debate. Proponents of network neutrality suggest that over-provisioning is a viable alternative to CoS. We quantify the extra capacity requirement for an over-provisioned classless (i.e., best-effort) network compared to a CoS network providing the same delay or loss performance for premium traffic. We first develop a link model that quantifies this Required Extra Capacity (REC). For bursty and realistic traffic distributions, we find the REC using ns-2 simulation comparisons of the CoS and classless link cases. We use these link models to quantify the REC for realistic network topologies. We show that REC can be significant even when the proportion of premium traffic is small, a situation often considered benign for the over-provisioning alternative.

The increased usage of directional methods of communications to improve medium reuse, network capacity, and bandwidth has prompted research into leveraging directionality in every layer of the network stack. Recently, there has been work on bringing the apparent capacity gains on layer 2 using directional communications methods to layer 3 by using directionality to route packets scalably in unstructured, flat networks. In their protocol, Orthogonal Rendezvous Routing Protocol, Cheng et al. [1] showed that by ”drawing” two lines orthogonal to each other at each node, it is possible to provide over 98% connectivity while maintaining only O(N3/2) evenly distributed states at a cost of only 1.2 path stretch. In this paper, we seek to provide more in-depth performance analysis by tuning additional factors such as the number of directions to transmit, the number of interfaces per node, among others, to understand its affect on varying network densities, topologies, connections, and traffic patterns. We show that by sending packets out in more directions, increased connectivity, smaller average path length, better goodput results only up to a point as compared to other routing protocols. The trade-off, however, is added state information maintained at each node and additional control packets received. We also show that the addition of more interfaces generally yields better packet delivery success, average path length, and goodput.

In this paper, we extend our proposed FGA-FEC coding scheme, a generalized MD-FEC method, to wireless networks. To protect the encoded scalable video bitstream over a lossy channel and facilitate content adaptation at intermediate nodes, we use product codes based on BCH/CRC codes as row codes and RS codes as column codes. We give a fast algorithm to optimize the product codes within several iterations from a near optimal point. Simulations show good performance in both content adaptation and protection.

The desire or ability of an ISP to provide differentiated service is a current hotly debated topic. In this paper, we quantify the value of having differentiated service (i.e., class-of-service (CoS)) support in an IP backbone. We compare the capacity requirements of a Diffserv environment providing service for applications that require delay or loss assurances in comparison to a network that provides classless (i.e., best-effort) service and still has to meet the same performance assurances. Our modeling framework first develops a link model that quantifies the required extra capacity (REC) in order for a classless link to provide the same level of performance as experienced by premium class traffic passing through a fixed capacity CoS link. We develop the REC calculations for the cases when average delay or the average loss probability is the target performance goal with Poisson or Markov modulated Poisson process (MMPP) input traffic. Our primary contribution is in quantifying the value of the CoS support in a network setting.

Free-Space Optical communication technology is a potential solution to the last mile or broadband access problem. Conventional free-space optical (FSO) communication is over a single link between two nodes. We explore multi-channel FSO communication system using compact (a maximum of a Sq.Ft) 2-dimensional antennas with multiple communication links between them to achieve very high aggregate bandwidths (100's of Gbps). But, close packaging of optical channels on the arrays causes inter-channel interference, reducing per channel capacity. We model the error due to inter-channel interference for such arrays and estimate the channel capacity. We address the multi-channel interference issue by both array design and by employing optical orthogonal codes (OOCs) for free-space optical communications and show that we can achieve multi Gbps bandwidths using such arrays. Possible applications for such multi-channel FSO systems can be in multi-hop broadband access networks or mesh networks and in back haul, connecting wireless base stations.

Routing in multi-hop wireless networks involves the indirection from a persistent name (or ID) to a locator. One of the biggest issues in routing is providing adequate connectivity while scaling the network. Recently, Bow-Nan Cheng et al., (2006) has attempted to mitigate this issue by using directional communication methods to find intersections between source-rendezvous and rendezvous-destination paths, providing effective routing in unstructured, flat networks. Cheng et al. showed that by "drawing" two lines orthogonal to each other at each node, it is possible to provide over 98% connectivity while maintaining only order O(N3/2) states. It is interesting, however to investigate what happens when additional lines are "drawn" and how that affects connectivity, path length and state complexity. In this paper, we examine how transmitting along one, two, three, and four lines affects routing and provide both analytical bounds for connectivity as well as packetized simulations on how these methods stack up in a more realistic environment. We show that by sending packets out in more directions, increased connectivity and smaller average path length results only up to a point. The trade-off, however, is added state information maintained at each node. We also show that in mobile environments, adding additional lines increases the chances for successful packet delivery only marginally.

Broadband technologies have made multi-hop wireless communications a reality. Loss-prone multi-hop networks pose challenges to link and transport layer protocols. Wireless links need to export low link-latencies, high goodputs and low residual loss rates to effectively enable interactive applications. Current link protocols with high ARQ persistence incur high latencies that impair such applications. We propose LL-HARQ, a link protocol that meets these goals. However, under high and bursty loss rates even LL-HARQ exports a small residual loss rate that could accumulate over multiple hops. Since TCP-SACK cannot handle error rates greater than 5%, a transport protocol (LT-TCP) designed for loss tolerance can be used under such cases. We provide insights into the structuring of the building blocks and balance between error-protection functions at the two layers and examine the case for cross-layer co-operation. Finally, we demonstrate that the combination achieves improved end-end performance (delay, loss and goodput) over traditional approaches.

Routing in communication networks involves the indirection from a persistent name (or ID) to a locator and delivering packets based upon the locator. In a large-scale, highly dynamic network, the ID-to-locator mappings are both large in number, and change often. Traditional routing protocols require high overhead to keep these in directions up-to-date. In this paper, we propose Weak State Routing (WSR), a routing mechanism for large-scale highly dynamic networks. WSR's novelty is that it uses random directional walks biased occasionally by weak indirection state information in intermediate nodes. The indirection state information is weak, i.e. interpreted not as absolute truth, but as probabilistic hints. Nodes only have partial information about the region a destination node is likely to be. This method allows us to aggregate information about a number of remote locations in a geographic region. In other words, the state information maps a set-of-IDs to a it geographical region. The intermediate nodes receiving the random walk use a method similar to longest-prefix-match in order to prioritize their mappings to decide how to bias and forward the random walk. WSR can also be viewed as an unstructured distributed hashing technique. WSR displays good rare-object recall with scalability properties similar to structured DHTs, albeit with more tolerance to dynamism and without constraining the degree distribution of the underlying network.Through simulations, we show that WSR offers a high packet delivery ratio, more than 98%. The control packet overhead incurred in the network scales as O(N) for N-node networks. The number of mappings stored in the network appears to scale as Θ(N(3/2)). We compare WSR with Dynamic Source Routing (DSR) and geographic forwarding (GPSR) combined with Grid Location Service (GLS). Our results indicate that WSR delivers more packets with less overhead at the cost of increased path length.

In a wireless LAN, the throughput is essentially important for multimedia applications due to their high bit rate requirement. In this paper, we propose two link layer error protection schemes that improve the effective throughout of wireless networks. Error control is applied to the packet header and packet payload separately. The network intermediate nodes either use header FEC or header CRC checksum to successfully transport the packets from the source to the destination. Only at the destination, error in payload may be corrected. We compare proposed schemes with 802.11 protocol and show that header error protection strategy can effectively increase the throughput, reduce the delay, and improve the video performance, via both theoretical analysis and simulation results.

Routing in multi-hop wireless networks involves the indirection from a persistent name (or ID) to a locator. Concepts such as coordinate space embedding help reduce the number and dynamism complexity of bindings and state needed for this indirection. Routing protocols which do not use such concepts often tend to flood packets during route discovery or dissemination, and hence have limited scalability. In this paper, we introduce orthogonal rendezvous routing protocol (ORRP) for meshed wireless networks. ORRP is a lightweight, but scalable routing protocol utilizing directional communications (such as directional antennas or free-space-optical transceivers) to relax information requirements such as coordinate space embedding and node localization. The ORRP source and ORRP destination send route discovery and route dissemination packets respectively in locally-chosen orthogonal directions. Connectivity happens when these paths intersect (i.e. rendezvous). We show that ORRP achieves connectivity with high probability even in sparse networks with voids. ORRP scales well without imposing DHT-like graph structures (eg: trees, rings, torus etc). The total state information required is O(N3/2) for N-node networks, and the state is uniformly distributed. ORRP does not resort to flooding either in route discovery or dissemination. The price paid by ORRP is suboptimality in terms of path stretch compared to the shortest path; however we characterize the average penalty and find that it is not severe.

We develop a spot pricing framework for intra-domain expected bandwidth contracts with loss based QoS guarantees. The framework accounts for both costs and risks associated with QoS delivery. A nonlinear pricing scheme is used for cost recovery and a utility based options pricing approach is developed for the risk related pricing. Application of options pricing techniques in Internet services provides a mechanism for fair risk sharing between the provider and the customer, and may be extended to price other uncertainties in QoS guarantees.

In this paper we analyze the error performance of free-space optical (FSO) communication over multiple hops. We first develop an error model for a single hop based on visibility, atmospheric attenuation, and geometric spread of the light beam. We model atmospheric visibility by Gaussian distributions with mean and variance values to reflect clear and adverse weather conditions. Based on this, we find the end-to-end bit error distribution of the FSO link for single hop and multi-hop scenarios. We present simulation results for decoded relaying, where each hop decodes the signal before retransmitting. We demonstrate that multi-hop FSO communication achieves a significant reduction in the mean bit error rate and also reduces the variance of the bit error rate. We argue that by lowering mean error and error variance, multi-hop operation facilitates an efficient system design and improves the reliability of the FSO link by application of specific coding schemes (such as forward error correction techniques).

In this paper, we introduce a novel link layer cooperation technique in noisy wireless networks to improve overall system throughput and reliability, and to reduce the cost of retransmission and energy consumption. Under a cluster-based network design, code combining is used together with FEC to improve the link layer reliability. This approach is different from how code combining is used in the conventional hybrid ARQ, which is in a sequential way. The analytical results and the simulations show that with the cooperation of nodes in a clustering network, the link reliability is greatly improved with the same power consumption. Equivalently, this can be viewed as the same link performance with a lower transmission power and lower interference.

In this paper, we propose a joint source-network error control coding (JSNC) scheme which efficiently integrates scalable video coding, error control coding and overlay infrastructure to stream video to heterogeneous users. The distributed overlay nodes adapt both the video bitstream and error control coding based on both user requirements and available bandwidth. A novel fine granular adaptive FEC (FGA-FECcheme, a generalization of MD-FEC, is proposed for error recovery during video transmission to heterogeneous users. Encoding once, the FGA-FEC can satisfy multiple heterogeneous users simultaneously without decoding/re-encoding FEC at intermediate nodes.

Virtual private networks (VPNs) feature notable characteristics in structure and traffic patterns that can be exploited by the service provider to achieve significant capacity savings. Efficient provisioning of point-to-point connections using statistical admission control is well understood. However, provisioning a VPN involves provisioning a set of point-to-multipoint connections and features an additional dimension in the form of a traffic matrix. Consequently we have multiple network mechanisms that are important for efficient operation: a) admission control, b) signaling-based per-link reservations, c) traffic matrix estimation. In this paper we examine the relative importance of mechanisms that positively affect the operational efficiency in the context of VPN provisioning. Using insights from our extensive measurement based study on the structural properties usually observed in VPNs, we build a simulation framework to quantify the trade-offs in opting for one mechanism over the other. We arrive at our conclusions with the help of simulations featuring a variety of VPN structures and network topologies. We find that the structural characteristics of VPNs cause traffic matrix estimation to be a dominant factor in determining the utilization gains. Consequently, we find that deploying statistical techniques might not be worth the effort if the traffic matrix is not incorporated. While signaling-based reservation mechanisms lead to higher utilization, edge-based techniques prove to be lot more scalable and simpler to realize. We explore the means to reduce the performance penalty associated with such simpler techniques.

Achieving efficient and fair bandwidth allocation while minimizing packet loss in high bandwidth-delay product networks has long been a daunting challenge. Existing end-to-end congestion control (eg TCP) and traditional congestion notification schemes (eg TCP+AQM/ECN) have significant limitations in achieving this goal. While the recently proposed XCP protocol addresses this challenge, XCP requires multiple bits to encode the congestion-related information exchanged between routers and end-hosts. Unfortunately, there is no space in the IP header for these bits, and solving this problem involves a non-trivial and time-consuming standardization process.In this paper, we design and implement a simple, low-complexity protocol, called Variable-structure congestion Control Protocol (VCP), that leverages only the existing two ECN bits for network congestion feedback, and yet achieves comparable performance to XCP, ie high utilization, low persistent queue length, negligible packet loss rate, and reasonable fairness. On the downside, VCP converges significantly slower to a fair allocation than XCP. We evaluate the performance of VCP using extensive ns2 simulations over a wide range of network scenarios. To gain insight into the behavior of VCP, we analyze a simple fluid model, and prove a global stability result for the case of a single bottleneck link shared by flows with identical round-trip times.

An automated method is presented for selecting optimal parameter settings for vessel/neurite segmentation algorithms using the minimum description length principle and a recursive random search algorithm. It trades off a probabilistic measure of image-content coverage against its conciseness. It enables nonexpert users to select parameter settings objectively, without knowledge of underlying algorithms, broadening the applicability of the segmentation algorithm, and delivering higher morphometric accuracy. It enables adaptation of parameters across batches of images. It simplifies the user interface to just one optional parameter and reduces the cost of technical support. Finally, the method is modular, extensible, and amenable to parallel computation. The method is applied to 223 images of human retinas and cultured neurons, from four different sources, using a single segmentation algorithm with eight parameters. Improvements in segmentation quality compared to default settings using 1000 iterations ranged from 4.7%-21%. Paired t-tests showed that improvements are statistically significant (p < 0.0005). Most of the improvement occurred in the first 44 iterations. Improvements in description lengths and agreement with the ground truth were strongly correlated (p = 0.78).

This paper generalizes the TCP Vegas congestion avoidance mechanism and uses accumulation , buffered packets of a flow inside network routers, as a congestion measure based on which a family of congestion control schemes can be derived. We call this model Accumulation-based Congestion Control (ACC), which fits into the nonlinear optimization framework proposed by Kelly. The ACC model serves as a reference for packet-switching network implementations. We show that TCP Vegas is one possible scheme under this model. It is well known that Vegas suffers from round trip propagation delay estimation error and reverse path congestion. We therefore design a new Monaco scheme that solves these problems by employing an out-of-band, receiver-based accumulation estimator, with the support of two FIFO priority queues from the (congested) routers. Comparisons between these two schemes demonstrate that Monaco does not suffer from the problems mentioned above and achieves better performance than Vegas. We use ns-2 simulations and Linux implementation experiments to show that the static and dynamic performance of Monaco matches the theoretic results. One key issue regarding the ACC model in general, i.e., the scalability of bottleneck buffer requirement, and a solution using a virtual queueing algorithm are discussed and evaluated.

Routing in ad-hoc networks is a complicated task because of many reasons. The nodes are low-memory, low-powered, and they cannot maintain routing tables large enough for well-known routing protocols. Because of that, greedy forwarding at intermediate nodes is desirable in ad-hoc networks. Also, for traffic engineering, multi-path capabilities are important. So, it is desirable to define routes at the source like in source based routing (SBR) while performing greedy forwarding at intermediate nodes. In this paper, we investigate trajectory-based routing (TBR) which was proposed as a middle-ground between SBR and greedy forwarding techniques. We address various issues regarding implementation of TBR. We also provide techniques to efficiently forward packets along a trajectory defined as a parametric curve.

Overlay networks offer promising capabilities for video streaming, due to their support for application-layer processing at the overlay forwarding nodes. In this paper we propose a novel overlay multi-hop FEC (OM-FEC) scheme that provides FEC encoding/decoding capabilities at intermediate nodes in an overlay path. Based on the current network conditions, the end-to-end overlay path is partitioned into segments, and appropriate FEC codes are applied over those segments. We evaluate our work in a real-world scenario and illustrate that the proposed OM-FEC can outperform a pure end-to-end strategy by 10-15 dB in terms of video PSNR.

This paper presents an efficient multiplexing and error control system to improve streaming video performance over path aggregates. While providing the application with increased aggregate bandwidth, the scheme reduces performance degradation due to high path latencies and loss rates. The reduction in effective loss and delay is achieved by smart multiplexing and exploiting the high latency paths to the user's advantage. A novel out-of-order transmission algorithm utilizes the higher latency paths to transfer suitable frames from within the transmit buffer. We present an FEC strategy for our scheme that decouples the transmission of error correction frames from the associated data. This provides protection against correlated losses. Our scheme, while not completely optimized, can provide close to optimal performance at a considerably lower complexity. We verify the performance of our scheme using the ns-2 simulator.

Virtual Private Networks provide secure and reliable communication between customer sites. With increase in number and size of VPNs, providers need efficient provisioning techniques that adapt to customer demand by leveraging a good understanding of VPN properties.In this paper we analyze two important properties of VPNs that impact provisioning - (a) structure of customer endpoint (CE) interactions and (b) temporal characteristics of CE-CE traffic. We deduce these properties by computing traffic matrices from SNMP measurements. We find that existing traffic matrix estimation techniques are not readily applicable to the VPN scenario due to the scale of the problem and limited measurement information. We begin by formulating a scalable technique that makes the most out of existing measurement information and provides good estimates for common VPN structures.We then use this technique to analyze SNMP measurement from a large IP VPN service provider. We find that even with limited measurement information we can realize adaptive provisioning for a significant fraction of VPNs, namely, those constituting the Hub-and-Spoke category. In addition, the ability to infer the structure of VPNs holds special significance for provisioning tasks arising from topology changes, link failures and maintenance. We are able to provide a classification of VPNs by structure and identify CEs that act as hubs of communication and hence require prioritized treatment during restoration and provisioning.

Traditionally uncooperative rate control schemes have implied open loop protocols such as UDP, CBR, etc. In this paper we show that closed loop uncooperative rate control schemes also exist and that the current AQM proposals cannot efficiently control their mis-behavior. Moreover, these proposals require that AQM be installed at all routers in the Internet which is not only expensive but requires significant network upgrade.In this paper we show that management of uncooperative flows need not be coupled with AQM design but can be viewed as edge based policing question. In this paper we propose an analytical model for managing uncooperative flows in the Internet by re-mapping their utility function to a target range of utility functions. This mapping can be achieved by transparently manipulating congestion penalties conveyed to the uncooperative users.The most interesting aspect of this research is that this task can be performed at the edge of the network with little state information about uncooperative flows. The proposed solution is independent of the buffer management algorithm deployed on the network. As such it works with Drop-Tail queues as well as any AQM scheme. We have analyzed the framework and evaluated it on various single and multi-bottleneck topologies with both Drop-Tail and RED. Our results show that the framework is robust and works well even in presence of background traffic and reverse path congestion.

Virtual Private Networks (VPNs) feature notable characteristics in structure and traffic patterns that allow for efficient resource allocation. A strategy that exploits the underlying characteristics of a VPN can result in significant capacity savings to the service provider. There are a number of admission control and bandwidth provisioning strategies to choose from. We examine tradeoffs in design choices in the context of distinctive characteristics of VPNs. We examine the value of signaling-based mechanisms, traffic matrix information and structural characteristics of VPNs in the way they impact resource utilization and service quality. We arrive at important conclusions which could have an impact on the way VPNs are architected. We show that the structure of VPNs profoundly influences achievable resource utilization gains with various admission control and provisioning schemes.

In this paper we evaluate rate distributions between competing flows in a network of droptail queues. Specifically we look at the case when some of the flows are non-conformant or mis-behaving and it's effect on conformant flows. Our results show in a network of droptail queues mis-behaving flows can have significantly higher bandwidth allocations at the cost of conformant flows. Further this unequal sharing worsens in a multi-bottleneck scenario where conformant flows may consistently time-out. However the distribution of rates improves if RED is used at the bottleneck thus suggesting deployment of RED. In this paper we also look at the fairness from the network's perspective rather then end-user's. As such we propose an analytical model for managing non-conformant or mis-behaving flows by manipulating congestion penalties conveyed to them. We show that this penalty transformation can map a user's utility function, U/sub s/, to any objective utility function, U/sub obj/. These penalty transformation modules can be completely implemented at the edge and can also work with droptail queues. We have analyzed the framework and evaluated it for both single and multi bottleneck scenarios.

This paper generalizes the TCP Vegas congestion avoidance mechanism and proposes a model to use accumulation, buffered packets of a flow inside network routers, as a congestion measure on which a family of congestion control schemes can be derived. We call this model accumulation-based congestion control (ACC). We use a bit-by-bit fluid model to define the accumulation concept and develop a general control algorithm, which includes a set of control policies. The new prove its proportional fairness and global stability. The ACC model serves as a reference for packet network implementations. We show that TCP Vegas is one possible scheme, which fits into the ACC model. It is well known that Vegas suffer from round trip propagation delay estimation error and reverse path queuing delay. We therefore design a new scheme called Monaco, which solves these problems by employing an out-of-band receiver-based accumulation estimator, with minimal support from network routers. Analysis and simulations comparisons between Vegas and Monaco demonstrate the effectiveness of the Monaco accumulation estimator. We use ns-2 simulations to show that the static and dynamic performance of Monaco matches the theoretic results. One key issue regarding the ACC model in general, i.e., the scalability of router buffer requirement, is discussed.

One of the biggest obstacles for implementing congestion pricing is the pricing-time scale. The Internet traffic is highly variant and hard to control without a mechanism that operates on very low time-scales, i.e. on the order of round-trip-times (RTTs). However, pricing naturally operates on very large time-scales because of human involvement. So, in order to put tight control on congestion through pricing, new implementation methods and architectures are needed for congestion pricing. In order to solve this problem, we propose a novel approach pricing over congestion control (POCC). The essence of POCC is to overlay congestion pricing on top of an underlying congestion control scheme which enforces a much tighter control than pricing. This way congestion in the interior network is controlled very tightly, while pricing is done at time-scales large enough to incorporate human involvement.

In this paper we present a video communication system that integrates end-to-end buffer management and congestion control at the source with the playout adjustment mechanism at the receiver. While each component of the system has been considered independently in the literature, our focus in this work is their integration. The proposed system exploits the fact that when congestion control is implemented at the source, most of the loss occurs at the source and not within the network. Based on this observation, we design the buffer management to trade off random loss for controlled loss of visually less important data. Frame rate is adjusted at the receiver to maximize the visual quality of the displayed video based on the overall loss. We tested our system with both H.26L and a subband/wavelet video coder, and found that it significantly improves the received video quality in both cases.

Peer-to-peer based multimedia delivery is becoming increasingly more important in today's networks. Using a peer- to-peer network to assist video streaming is a topic of considerable interest. In this paper, we propose a novel hybrid video downloading/streaming scheme (HDS) that efficiently integrates traditional client/server based video streaming and peer-to-peer based media distribution. Furthermore, we propose a receiver-driven algorithm to coordinate the downloading and streaming modes; and control the state transitions between these modes. We have performed real-world experiments and simulations to validate our concept. These results show that our proposed scheme greatly increases the availability of video content on the receiver side and simultaneously reduces the server load significantly.

Since optimization of networks pricing has attracted significant attention over the last decade. These studies assumed concave utility functions for users and derived optimal pricing strategies for the network provider. In this paper, we consider effect of user's elasticity to price and bandwidth on optimality of pricing. We first derive optimal pricing strategy for the case logarithmic user utilities. Then, we investigate two types of elasticity for users: demand-price elasticity and utility-bandwidth elasticity. By incorporating these two elasticities, we develop a non-logarithmic utility function for users. Finally, we derive an optimal pricing strategy for the non-logarithmic user utilities and illustrate that pricing strategy should be more conservative when the elasticities increase.

One of the key issues for implementing congestion pricing is the pricing granularity (i.e. pricing interval or time scale). The Internet traffic is highly variant and hard to control without a mechanism that operates on very low time-scales, i.e. on the order of round-trip-times (RTTs). However, pricing naturally operates on very large times-scales because of human involvement. Moreover, structure of wide-area networks does not allow frequent price updates for many reasons, such as RTTs are very large for some cases. In this paper, we investigate the issue of pricing granularity, identify problems, and propose solutions.

The capacity of ad hoc wireless networks is constrained by the interference between concurrent transmissions from neighboring nodes. Gupta and Kumar have shown that the capacity of an ad hoc network does not scale well with the increasing number of nodes in the system when using omnidirectional antennas [6]. We investigate the capacity of ad hoc wireless networks using directional antennas. In this work, we consider arbitrary networks and random networks where nodes are assumed to be static.In arbitrary networks, due to the reduction of the interference area, the capacity gain is proven to be √2π/α when using directional transmission and omni reception. Because of the reduced probability of two neighbors pointing to each other, the capacity gain is √2π/β when omni transmission and directional reception are used. Although these two expressions look similar, the proof technique is different. By taking advantage of the above two approaches, the capacity gain is 2π/√αβ when both transmission and reception are directional.For random networks, interfering neighbors are reduced due to the decrease of interference area when directional antennas are used for transmission and/or reception. The throughput improvement factor is 2π/α, 2π/β and 4π2/αβ for directional transmission/omni reception, omni transmission/direc-tional reception, and directional transmission/directional reception, respectively.We have also analyzed hybrid beamform patterns that are a mix of omnidirectional/directional and a better model of real directional antennas.

Parameter configuration is a common procedure used in large-scale network protocols to support multiple operational goals. It can be formulated as a black-box optimization problem and solved with an efficient search algorithm. This paper proposes a new heuristic search algorithm, Recursive Random Search(RRS), for large-scale network parameter optimization. The RRS algorithm is based on the initial high-efficiency feature of random sampling and it attempts to maintain this high efficiency by constantly "restarting" random sampling with adjusted sample spaces. Besides the high efficiency, the RRS algorithm is robust to the effect of random noise and trivial parameters in the objective function because of its root in random sampling. These features are very important for the efficient optimization of network protocol configuration. The performance of RRS is demonstrated with the tests on a suite of benchmark functions. The algorithm has been applied to the configuration of several network protocols, such as RED, OSPF and BGP. One example application in OSPF routing algorithm is presented.

Continuing the process of improvements made to TCP through the addition of new algorithms in Tahoe and Reno, TCP SACK aims to provide robustness to TCP in the presence of multiple losses from the same window. In this paper we present analytic models to estimate the latency and steady-state throughput of TCP Tahoe, Reno, and SACK and validate our models using both simulations and TCP traces collected from the Internet. In addition to being the first models for the latency of finite Tahoe and SACK flows, our model for the latency of TCP Reno gives a more accurate estimation of the transfer times than existing models. The improved accuracy is partly due to a more accurate modeling of the timeouts, evolution of cwnd during slow start and the delayed ACK timer. Our models also show that, under the losses introduced by the droptail queues which dominate most routers in the Internet, current implementations of SACK can fail to provide adequate protection against timeouts and a loss of roughly more than half the packets in a round will lead to timeouts. We also show that with independent losses SACK performs better than Tahoe and Reno and, as losses become correlated, Tahoe can outperform both Reno and SACK.

While it is well known that second order scaling in network traffic can lead to larger queueing delays, higher drop rates and extended periods of congestion, reducing the scaling exponents has remained an open problem. In this paper we evaluate some techniques to reduce the degree of scaling in TCP traffic, specifically by reducing two related causes: (1) timeouts and exponential backoffs; and (2) burstiness and ACK compression. We propose a simple modification to the RED algorithm, and show that it can lead to significant reductions in both multi and mono fractal properties of TCP traffic as compared to the currently implemented active and passive buffer management policies. We then evaluate TCP pacing and show that it too can reduce the multi and mono fractal scaling of traffic. We also show that though our techniques are aimed at small time-scale TCP related causes of scaling, they are also effective in reducing the degree of self-similarity in traffic even when application and user level causes are also present, as long as TCP is used as the underlying transport protocol.

Random early detection (RED) is an active queue management mechanism designed to provide better performance than traditional DropTail. However, its parameter setting has proved to be very sensitive to network scenarios and needs constant tuning to achieve ideal performance under varying network conditions. In view of the fact that RED has not been understood well enough for an analytical approach, this paper takes advantage of network simulation techniques and formulates the optimal configuration of RED as a black-box optimization problem. An optimization objective is designed to effectively reflect the tradeoff between utilization and queueing delay. Based on the proposed RED optimization scheme, a general automatic network management system, i.e., on-line simulation system, has been used for the on-line tuning of RED under changing network conditions. The proposed approach is empirically validated with simulations and real network experiments. The simulation results show that RED controlled with on-line simulation system is able to stabilize around the expected equilibrium status under varying conditions and maintain high utilization.

We present a baseline implementation strategy for the well-known Smart Market pricing scheme on DiffServ. Our strategy models Smart Market's theoretically defined properties as much as possible. In order to suit the DiffServ framework, we propose ways of focusing Smart Market's complex operations at the edges while keeping the interior simple. Based on the proposed implementation strategy, we develop a packet-based simulation of Smart Market. By simulation, we then investigate Smart Market's performance in terms of stability, fairness, and service differentiation on UDP and TCP traffic. We also look at the importance of packet sorting (i.e. sorting of packets at routers according to their bids as proposed in Smart Market) in Smart Market's performance. By several simulations, we find that packet sorting does not really improve the performance for all three metrics (stability, fairness, and service differentiation). So, it is not necessary to implement packet sorting for Smart Market's possible deployment, thus significantly reducing the necessity for router upgrades.

We propose MCA, a rate-based end-to-end multicast congestion avoidance scheme. Congestion avoidance [6] is different from congestion control in the sense that our scheme detects and responds to network congestion without necessarily inducing packet loss. Our scheme is a single-rate scheme and operates end-to-end, i.e., it goes at the rate allowed by the worst congested receiver and does not expect packet marking or other support from intermediate bottlenecks. Congestion is detected autonomously at receivers using the concept of "accumulation " and simple thresholding techniques proposed in our recent unicast work [8]. Congestion feedback to senders can be in the form of single-bit congestion indication (CIs) or as a multi-bit output rate measure. The feedback is sparse in the sense that at most one feedback is generated per measurement period (unlike multiple loss indications generated during packet loss). The source implements two key blocks: a filtering block to discriminate between competing feedback from receivers, and a congestion response block which implements a rate-increase/decrease policy. The two different feedback models (bitbased or explicit rate-based) leads to two different schemes: bit-based and explicit rate-based schemes. Simulation results show that both schemes avoid the drop-to-zero problem and is fair with unicast congestion avoidance schemes.

In this paper we propose an edge-based quality of service (QoS) architecture aimed at site-to-site private networks over the Internet. We extend the traditional point-to-point service model to a point-to-set service model, assuming a finite, bounded set of destination sites. A point-to-set service allows a user to have a pool of premium tokens, which could be flexibly assigned to traffic going towards any destination within the set. The proposed point-to-set service provides statistical assurances and flexibility to users while allowing providers to obtain multiplexing gains. To realize the point-to-set service model, we introduce edge-based dynamic bandwidth tracking and provisioning schemes. The tracking algorithm predicts demand towards a given destination edge. This information is used to efficiently allocate bandwidth towards the destinations in the set. Simulation results are presented to demonstrate the merits of the proposed architecture in terms of cost savings to the customer and efficient resource utilization to the provider.

In this paper we present analytic models to estimate the latency and steady-state throughput of TCP Tahoe, Reno and SACK and validate our models using both simulations and TCP traces collected from the Internet. We also conduct a study comparing the performance of these versions of TCP under different loss scenarios. In addition to being the first models for the latency of finite Tahoe and SACK flows, our model for the latency of TCP Reno gives a more accurate estimation of the transfer times than existing models. Our models show that under the losses introduced by the droptail queues which dominate most routers in the Internet, current implementations of SACK fail to provide adequate protection against timeouts and a loss of roughly more than half the packets in a round will lead to timeouts. We also show that with independent losses, SACK performs better than Tahoe and Reno and as losses become correlated, Tahoe can outperform both Reno and SACK.

The complexity and dynamics of the Internet is driving the demand for scalable and effective network control. This paper proposes a collaborative on-line simulation architecture to provide pro-active and automated control functions for networks. The general model includes autonomous on-line simulators which continuously monitor/model the network conditions and execute a search in the parameter state space for better settings of the protocol parameters. The protocol parameters are then tuned by the on-line simulation system. We describe the building blocks of this architecture and investigate the implementation challenges in the areas of network modeling, on-line simulation and parameter search. We also discuss the applicability of this system and present the simulation and test results of a preliminary implementation.

This paper describes the "explicit rate indication for congestion avoidance" (ERICA) scheme for rate-based feedback from asynchronous transfer mode (ATM) switches. In ERICA, the switches monitor their load on each link and determine a load factor, the available capacity, and the number of currently active virtual channels. This information is used to advise the sources about the rates at which they should transmit. The algorithm is designed to achieve high link utilization with low delays and fast transient response. It is also fair and robust to measurement errors caused by the variations in ABR demand and capacity. We present performance analysis of the scheme using both analytical arguments and simulation results. The scheme is being considered for implementation by several ATM switch manufacturers.

The available bit rate (ABR) service is designed to fairly allocate the bandwidth unused by higher priority services. The network indicates to the ABR sources the rates at which they should transmit to minimize their cell loss. Switches must constantly measure the demand and available capacity, and divide the capacity fairly among the contending connections. In order to compute the fair and efficient allocation for each connection, a switch needs to determine the effective number of active connections. We propose a method for determining the number of active connections and the fair bandwidth share for each. We prove the efficiency and fairness of the proposed method analytically, and simulate it for a number of configurations.

ABR traffic management for point-to-multipoint connections controls the source rate to be the minimum rate supported by all the branches of the multicast tree. A number of algorithms have been developed for extending ABR congestion avoidance algorithms to perform feedback consolidation at the branch points. This paper discusses various design options and implementation alternatives for the consolidation algorithms, and proposes a number of new algorithms. The performance of the proposed algorithms and the previous algorithms is compared under a variety of conditions. Results indicate that the algorithms we propose eliminate the consolidation noise (caused if the feedback is returned before all branches respond), while exhibiting a fast transient response.

Compressed video is well known to be self-similar in nature. We model VBR carrying long-range dependent, multiplexed MPEG-2 video sources traffic. The actual traffic for the model is generated using fast-Fourier transform of the fractional Gaussian noise sequence. Our model of compressed video sources bears similarity to an MPEG-2 transport stream carrying video, i.e., it is long-range dependent and generates traffic in a piecewise constant bit rate manner. We study, the effect of such VBR traffic on ABR carrying TCP traffic. The effect of such VBR traffic is that the ABR capacity is highly variant. We find that a switch algorithm like ERICA+ can tolerate this variance in ABR capacity while maintaining high throughput and low delay. We present simulation results for terrestrial and satellite configurations.

ATM-UBR service responds to congestion by dropping cells when switch buffers become full. TCP connections running over UBR experience low throughput and high unfairness. For 100% TCP throughput, each switch needs buffers equal to the sum of the window sizes of all the TCP connections. Intelligent drop policies can improve the performance of TCP over UBR with limited buffers. The UBR+ service proposes enhancements to UBR for intelligent drop. The early packet discard scheme improves throughput but does not attempt to improve fairness. The selective packet drop scheme based on per-connection buffer occupancy improves fairness. The fair buffer allocation scheme further improves both throughput and fairness.

We extend our earlier studies of buffer requirements of TCP over ABR in two directions. First, we study the performance of TCP over ABR in an ATM backbone. We find that the TCP queues are at the edge router and not inside the ATM network. The edge router requires buffering equal to the sum of the receiver window sizes of the participating TCP connections. Second, we study the performance when ABR capacity is variable due to the effect of various patterns of VBR background traffic. The key factors in this study are the VBR traffic pattern, ABR feedback delays and the sensitivity of the ABR switch scheme to variance. We present our experiences in refining the ERICA+ switch scheme to handle these conditions.

We study the performance of selective acknowledgements (SACK) with TCP over the ATM-UBR service category. We examine various unspecified bit rate (UBR) drop policies, TCP mechanisms and network configurations to recommend optimal parameters for TCP over UBR. We discuss various TCP congestion control mechanisms compare their performance for LANs and WANs. We describe the effect of satellite delays on TCP performance over UBR and present simulation results for LANs, WANs and satellite networks. SACK TCP improves the performance of TCP over UBR, especially for large delay networks. Intelligent drop policies at the switches are an important factor for good performance in local area networks.