Climate indices play a pivotal role in deciphering and monitoring complex climatic patterns, providing essential insights into long-term trends and variations. This research, spanning multiple seasons, aims to provide comprehensive insights into the dynamics between extreme precipitation indices and the frequency of monthly dry days (MDD) through correlation analysis. By utilizing calculated indices to grasp a range of precipitation patterns, this study places emphasis not just on the intensity but also on how these correlations are spread across different regions. Through rigorous analysis of correlations and long-term anomalies, there are meaningful associations between the precipitation indices and MDD. Such insights are valuable for advancing our understanding of regional climate variations within the seasonal extreme precipitation indices and the dry days and hold significance for applications in the fields of agriculture and water resource management.

Stereo matching aims to perceive the 3D geometric configuration of scenes and facilitates a variety of computer vision in advanced driver assistance systems (ADAS) applications. Recently, deep convolutional neural networks (CNNs) have shown dramatic performance improvements for computing the matching cost in the stereo matching. However, the performance of CNN-based approaches relies heavily on datasets, requiring a large number of ground truth data which needs tremendous works. To overcome this limitation, we present a novel framework to learn CNNs for matching cost computation in an unsupervised manner. Our method leverages an image domain learning combined with stereo epipolar constraints. By exploiting the correspondence consistency between stereo images, our method selects putative positive samples in each training iteration and utilizes them to train the networks. We further propose a positive sample propagation scheme to leverage additional training samples. Our unsupervised learning method is evaluated with two kinds of network architectures, simple and precise CNNs, and shows comparable performance to that of the state-of-the-art methods including both supervised and unsupervised learning approaches on KITTI, Middlebury, HCI, and Yonsei datasets. This extensive evaluation demonstrates that the proposed learning framework can be applied to deal with various real driving conditions.

We address the problem of 3D reconstruction from uncalibrated LiDAR point cloud and stereo images. Since the usage of each sensor alone for 3D reconstruction has weaknesses in terms of density and accuracy, we propose a deep sensor fusion framework for high-precision depth estimation. The proposed architecture consists of calibration network and depth fusion network, where both networks are designed considering the trade-off between accuracy and efficiency for mobile devices. The calibration network first corrects an initial extrinsic parameter to align the input sensor coordinate systems. The accuracy of calibration is markedly improved by formulating the calibration in the depth domain. In the depth fusion network, complementary characteristics of sparse LiDAR and dense stereo depth are then encoded in a boosting manner. Since training data for the LiDAR and stereo depth fusion are rather limited, we introduce a simple but effective approach to generate pseudo ground truth labels from the raw KITTI dataset. The experimental evaluation verifies that the proposed method outperforms current state-of-the-art methods on the KITTI benchmark. We also collect data using our proprietary multi-sensor acquisition platform and verify that the proposed method generalizes across different sensor settings and scenes.

Even though there exist significant advances in recent studies, existing methods for pedestrian detection still have shown limited performances under challenging illumination conditions especially at nighttime. To address this, cross-spectral pedestrian detection methods have been presented using color and thermal, and shown substantial performance gains on the challenging circumstances. However, their paired cross-spectral settings have limited applicability in real-world scenarios. To overcome this, we propose a novel learning framework for cross-spectral pedestrian detection in an unpaired setting. Based on an assumption that features from color and thermal images share their characteristics in a common feature space to benefit their complement information, we design the separate feature embedding networks for color and thermal images followed by sharing detection networks. To further improve the cross-spectral feature representation, we apply an adversarial learning scheme to intermediate features of the color and thermal images. Experiments demonstrate the outstanding performance of the proposed method on the KAIST multi-spectral benchmark in comparison to the state-of-the-art methods.

We present a deep architecture and learning framework for establishing correspondences across cross-spectral visible and infrared images in an unpaired setting. To overcome the unpaired cross-spectral data problem, we design the unified image translation and feature extraction modules to be learned in a joint and boosting manner. Concretely, the image translation module is learned only with the unpaired cross-spectral data, and the feature extraction module is learned with an input image and its translated image. By learning two modules simultaneously, the image translation module generates the translated image that preserves not only the domain-specific attributes with separate latent spaces but also the domain-agnostic contents with feature consistency constraint. In an inference phase, the cross-spectral feature similarity is augmented by intra-spectral similarities between the features extracted from the translated images. Experimental results show that this model outperforms the state-of-the-art unpaired image translation methods and cross-spectral feature descriptors on various visible and infrared benchmarks.

We present a deep convolutional neural network (CNN) architecture for high-precision depth estimation by jointly utilizing sparse 3D LiDAR and dense stereo depth information. In this network, the complementary characteristics of sparse 3D LiDAR and dense stereo depth are simultaneously encoded in a boosting manner. Tailored to the LiDAR and stereo fusion problem, the proposed network differs from previous CNNs in the incorporation of a compact convolution module, which can be deployed with the constraints of mobile devices. As training data for the LiDAR and stereo fusion is rather limited, we introduce a simple yet effective approach for reproducing the raw KITTI dataset. The raw LiDAR scans are augmented by adapting an off-the-shelf stereo algorithm and a confidence measure. We evaluate the proposed network on the KITTI benchmark and data collected by our multi-sensor acquisition system. Experiments demonstrate that the proposed network generalizes across datasets and is significantly more accurate than various baseline approaches.

We present a novel human body co-segmentation method for unregistered multispectral, color and thermal, images by leveraging CNNs. The main challenges for that tasks are no-alignment between color and thermal images and an absent of ground truth human segmentation labels. To solve these limitations, our key-insight is to formulate the segmentation network for each modality that solve two sub-tasks, correspondence and classification, in a joint and iterative manner. We formulate the learning framework between multispectral images in a way that training labels for one modality are used to learn the network for the other modality. We estimate dense correspondences between multispectral image pairs using intermediate convolutional activations of CNNs and perform human segmentation for each modality through the conditional random fields (CRF) optimization using unary and pairwise fusion. These two steps are formulated as an iterative framework, enables the network to converge on an optimal solution. Experimental results show that our proposed method outperforms conventional state-of-the-art methods on the VAP benchmark consisting of unregistered multispectral color and thermal images.

In this work, we propose an efficient method that generates pedestrian proposals suitable for the autonomous vehicle. Our main intuition is that depth information provides an important cue to assign the scale of pedestrian proposals. Based on the observation that in a 3-D world coordinate the scales of pedestrians are almost similar, we formulate the scales of pedestrian patches by projecting 3-D models to an image plane with its corresponding depth. We also introduce a scale-aware binary description using both color and depth images. By using this descriptor, the regression models are trained to rank the pedestrian proposal candidates and adjust the proposal bounding boxes for an accurate localization. Our algorithm achieves significant performance gains compared to conventional proposal generation methods on the challenging KITTI dataset.

We present a unified framework for the image classification of image sets taken under varying modality conditions. Our method is motivated by a key observation that the image feature distribution is simultaneously influenced by the semantic-class and the modality category label, which limits the performance of conventional methods for that task. With this insight, we introduce modality uniqueness as a discriminative weight that divides each modality cluster from all other clusters. By leveraging the modality uniqueness, our framework is formulated as unsupervised modality clustering and classifier learning based on modality-invariant similarity kernel. Specifically, in the assignment step, each training image is first assigned to the most similar cluster according to its modality. In the update step, based on the current cluster hypothesis, the modality uniqueness and the sparse dictionary are updated. These two steps are formulated in an iterative manner. Based on the final clusters, a modality-invariant marginalized kernel is then computed, where the similarities between the reconstructed features of each modality are aggregated across all clusters. Our framework enables the reliable inference of semantic-class category for an image, even across large photometric variations. Experimental results show that our method outperforms conventional methods on various benchmarks, such as landmark identification under severely varying weather conditions, domain-adapting image classification, and RGB and near-infrared image classification.

This paper presents a method for detecting a pedestrian by leveraging multi-spectral image pairs. Our approach is based on the observation that a multi-spectral image, especially far-infrared (FIR) image, enables us to overcome inherent limitations for pedestrian detection under challenging circumstances, such as even dark environments. For that task, multi-spectral color-FIR image pairs are used in a synergistic manner for pedestrian detection through deep convolutional neural networks (CNNs) learning and support vector regression (SVR). For inferring the confidence of a pedestrian, we first learn CNNs between color images (or FIR images) and bounding box annotations of pedestrians, respectively. Furthermore, for each object proposal, we extract intermediate activation features from network, and learn the probability of pedestrian using SVR. To improve the detection performance, the learned probability of pedestrian for each proposal is accumulated on the image domain. Based on the pedestrian confidence estimated from each network and accumulated pedestrian probabilities, the most probable pedestrian is finally localized among object proposal candidates. Thanks to its high robustness of multi-spectral imaging in dark environments and its high discriminative power of deep CNNs, our framework is shown to surpass state-of-the-art pedestrian detection methods on multi-spectral pedestrian benchmark.

Recent works in empirical 802.11 wireless LAN performance evaluation have shown that cross-layer interactions in WLANs can be subtle, sometimes leading to unexpected results. Two such instances are: (i) significant throughput degradation resulting from automatic rate fallback (ARF) having difficulty distinguishing collision from channel noise, and (ii) scalable TCP over DCF performance that is able to mitigate the negative performance effect of ARF by curbing multiple access contention even when the number of stations is large. In this paper, we present a framework for analyzing complex cross-layer interactions in 802.11 WLANs, with the aim of providing effective tools for understanding and improving WLAN performance. We focus on cross-layer interactions between ARF, DCF, and TCP, where ARF adjusts coding at the physical layer, DCF mediates link layer multiple access control, and TCP performs end-to-end transport. We advance station-centric Markov chain models of ARF, ARF-DCF with and without RTS/CTS, and TCP over DCF that may be viewed as multi-protocol extensions of Bianchi's IEEE 802.11 model. We show that despite significant increase in complexity the analysis framework leads to tractable and accurate performance predictions. Our results complement empirical and simulation-based findings, demonstrating the versatility and efficacy of station-centric Markov chain analysis for capturing cross-layer WLAN dynamics.

Prior measurement and analysis have shown that the traffic in data networks is better described by heavy-tailed (HT) distributions. For ad hoc routing, it is important to understand the impact of the heavy-tailed nature of network traffic for routing decisions since heavy-tailed workload can lead to severe performance degradation. In this paper, we present our current work on ad hoc routing by exploiting the heavy-tailed nature of the session (flow) lifetime. In particular, long- and shortlived flows are differentiated based on the unique predictability nascent in heavy-tailed distributions, with the long-lived flows distributed across the network as disperse as possible, and the short-lived flows treated as "filler" traffic. We have implemented such ideas in various ad hoc routing protocols, and in this paper we present our results experimenting on top of dynamic source routing (DSR). Simulation results reveal that our modified protocol, called DSR-HT, significantly outperforms the state-of- the-art under heavy-tailed workload.

Extensive measurement and analysis have shown that traffic in data networks is better described by heavy-tailed distributions. We present our current work on ad hoc routing by exploiting the heavy-tailed nature of the flow lifetime. In particular, the long- and short-lived flows are differentiated based on the unique predictability nascent in heavy- tailed distributions, with the long- and short-lived flows being handled separately. A modified AODV has been implemented to enable heavy-tailedness awareness. Simulation results reveal that the new protocol, AODV-HT, significantly outperforms the state-of-the-art under heavy-tailed workload.

One of the key challenges in designing a rate adaptation scheme for IEEE 802.11 wireless LANs (WLANs) is to differentiate bit errors from link-layer collisions. Many recent rate adaptation schemes adopt the RTS/CTS mechanism to prevent collision losses from triggering unnecessary rate decrease. However, the RTS/CTS handshake incurs significant overhead and is rarely activated in today's infrastructure WLANs. In this paper we propose a new rate adaptation scheme that mitigates the collision effect on the operation of rate adaptation. In contrast to previous approaches adopting fixed rate-increasing and decreasing thresholds, our scheme varies threshold values based on the measured network status. Using the "retry" information in 802.11 MAC headers as feedback, we enable the transmitter to gauge current network state. The proposed rate adaptation scheme does not require additional probing overhead incurred by RTS/CTS exchanges and can be easily deployed without changes in firmware. We demonstrate the effectiveness of our solution by comparing with existing approaches through extensive simulations.

Abstract — Rate adaptation is one of the basic functionalities in today’s 802.11 wireless LANs (WLANs). Although it is primarily designed to cope with the variability of wireless channels and achieve higher system spectral efficiency, its design needs careful consideration of cross-layer dependencies, in particular, linklayer collisions. Most practical rate adaptations focus on the time-varying characteristics of wireless channels, ignoring the impact of link-layer collisions. As a result, they may lose their effectiveness due to unnecessary rate downshift wrongly triggered by the collisions. Some recently proposed rate adaptations use RTS/CTS to suppress the collision effect by differentiating collisions from channel errors. The RTS/CTS handshake, however, incurs significant overhead and is rarely activated in infrastructure WLANs. In this paper, we introduce a new approach for optimizing the operation of rate adaptations by adjusting the rate-increasing and decreasing parameters based on link-layer measurement. To construct the algorithm, we study the impact of rate-increasing and decreasing thresholds on performance and show that dynamic adjustment of thresholds is an effective way to mitigate the collision effect in multi-user environments. Our method does not require additional probing overhead incurred by RTS/CTS exchanges and may be practically deployed without change in firmware. We demonstrate the effectiveness of our solution, comparing with existing approaches through extensive simulations. I.

Wireless Internet access is facilitated by IEEE 802.11 WLANs that, in addition to realizing a specific form of CSMA/CA-distributed coordination function (DCF)- implement a range of performance enhancement features such as multi-rate adaptation that induce cross-layer protocol coupling. Recent works in empirical WLAN performance evaluation have shown that cross-layer interactions can be subtle, sometimes leading to unexpected outcomes. Two such instances are: significant throughput degradation (a bell-shaped throughput curve) resulting from automatic rate fallback (ARF) having difficulty distinguishing collision from channel noise, and scalable TCP performance over DCF that is able to curtail effective multiple access contention in the presence of many contending stations. The latter also mitigates the negative performance effect of ARF. In this paper, we present station-centric Markov chain models of WLAN cross-layer performance aimed at capturing complex interactions between ARF, DCF, and TCP. Our performance analyses may be viewed as multi-protocol extensions of Bianchi's IEEE 802.11 model that, despite significantly increased complexity, lead to tractable and accurate performance predictions due to modular coupling. Our results complement empirical and simulation-based findings, demonstrating the versatility and efficacy of station-centric Markov chain analysis for capturing cross-layer WLAN dynamics.

Widespread deployment of infrastructure WLANs has made Wi-Fi an integral part of today's Internet access technology. Despite its crucial role in affecting end-to-end performance, past research has focused on MAC protocol enhancement, analysis, and simulation-based performance evaluation without sufficient consideration for modeling inaccuracies stemming from interlayer dependencies, including physical layer diversity, that significantly impact performance. We take a fresh look at IEEE 802.11 WLANs and using experiment, simulation, and analysis demonstrate its surprisingly agile performance traits. Our findings are two-fold. First, contention-based MAC throughput degrades gracefully under congested conditions, enabled by physical layer channel diversity that reduces the effective level of MAC contention. In contrast, fairness degrades and jitter increases significantly at a critical offered load. This duality obviates the need for link layer flow control for throughput improvement. Second, TCP-over-WLAN achieves high throughput commensurate with that of wireline TCP under saturated conditions, challenging the widely held perception that TCP throughput fares poorly over WLANs when subject to heavy contention. We show that TCP-over-WLAN prowess is facilitated by the self-regulating actions of DCF and TCP feedback control that jointly drive the shared channel at an effective load of two to three wireless stations, even when the number of active stations is large. We show that the mitigating influence of TCP extends to unfairness and adverse impact of dynamic rate shifting under multiple access contention. We use experimentation and simulation in a complementary fashion, pointing out performance characteristics where they agree and differ.

Wide-spread deployment of infrastructure WLANs has made Wi-Fi an integral part of today's Internet access technology. Despite its crucial role in affecting end-to-end performance, past research has focused on MAC protocol enhancement, analysis and simulation-based performance evaluation without sufficient consideration for modeling inaccuracies stemming from inter-layer dependencies, including physical layer diversity, that significantly impact performance. We take a fresh look at IEEE 802.11 WLANs, and using a combination of experiment, simulation, and analysis demonstrate its surprisingly agile performance traits. Our main findings are two-fold. First, contention-based MAC throughput degrades gracefully under congested conditions, enabled by physical layer channel diversity that reduces the effective level of MAC contention. In contrast, fairness and jitter significantly degrade at a critical offered load. This duality obviates the need for link layer flow control for throughput improvement but necessitates traffic control for fairness and QoS. Second, TCP-over-WLAN achieves high throughput commensurate with that of wireline TCP under saturated conditions, challenging the widely held perception that TCP throughput fares poorly over WLANs when subject to heavy contention. We show that TCP-over-WLAN prowess is facilitated by the self-regulating actions of DCF and TCP congestion control that jointly drive the shared physical channel at an effective load of 2--3 wireless stations, even when the number of active stations is very large. Our results highlight subtle inter-layer dependencies including the mitigating influence of TCP-over-WLAN on dynamic rate shifting.

An important quality-of-service (QoS) issue in wireless multimedia networks is how to control handoff drops. In this paper, we propose admission control algorithms that adaptively control the admission threshold in each cell, in order to keep the handoff dropping probability below a pre-defined level. The admission threshold is dynamically adjusted based on handoff dropping events. We first present a simple admission control scheme which brings out an important performance evaluation criterion - inter-cell fairness - and serves as a reference point. We then investigate the inter-cell unfairness problem and develop two enhanced schemes to overcome this problem. The performance of these protocols is benchmarked and compared with other competitive schemes. The results indicate that our schemes perform very well while, in addition, achieving significantly reduced complexity and signaling load.

Analytical and empirical studies have shown that self-similar traffic can have a detrimental impact on network performance including amplified queuing delay and packet loss ratio. On the flip side, the ubiquity of scale-invariant burstiness observed across diverse networking contexts can be exploited to design better resource control algorithms. We explore the issue of exploiting the self-similar characteristics of network traffic in TCP congestion control. We show that the correlation structure present in long-range dependent traffic can be detected on-line and used to predict future traffic. We then devise an novel scheme, called TCP with traffic prediction (TCP-TP), that exploits the prediction result to infer, in the context of AIMD (additive increase, multiplicative decrease) steady-state dynamics, the optimal operational point for a TCP connection. Through analytical reasoning, we show that the impact of prediction errors on fairness is minimal. We also conduct ns-2 simulation and FreeBSD 4.1-based implementation studies to validate the design and to demonstrate the performance improvement in terms of packet loss ratio and throughput attained by connections.

Recent studies have shown that if the overhead associated with process migration is not high when compared to the estimated lifetime of processes, migrating isolated processes to underutilized workstations can improve system performance. However, if tightly-coupled processes from groups of communicating processes or other existing dependencies are separated, system performance can decrease due to increased communication cost. Current load balancers do not make effective use of communication dependencies in their algorithms. We have implemented a communication-sensitive load balancer for DUNES (Distributed UNix ExtenSion)-a user-level distributed operating system-that uses the run-time communication pattern between processes when balancing load. We examine communication-sensitive load balancing under different workload conditions and compare its performance against a load balancer that is not communication-aware. Our results show that a communication-sensitive load balancer performs substantially better than a load balancer that is not communication-aware under workloads that have a "good mix" of CPU-bound and I/O-bound processes.

Effective mitigation of denial of service (DoS) attack is a pressing problem on the Internet. In many instances, DoS attacks can be prevented if the spoofed source IP address is traced back to its origin which allows assigning penalties to the offending party or isolating the compromised hosts and domains from the rest of the network. IP traceback mechanisms based on probabilistic packet marking (PPM) have been proposed for achieving traceback of DoS attacks. We show that probabilistic packet marking-of interest due to its efficiency and implementability vis-a-vis deterministic packet marking and logging or messaging based schemes-suffers under spoofing of the marking field in the IP header by the attacker which can impede traceback by the victim. We show that there is a trade-off between the ability of the victim to localize the attacker and the severity of the DoS attack, which is represented as a function of the marking probability, path length, and traffic volume. The optimal decision problem-the victim can choose the marking probability whereas the attacker can choose the spoofed marking value, source address, and attack volume-can be expressed as a constrained minimax optimization problem, where the victim chooses the marking probability such that the number of forgeable attack paths is minimized. We show that the attacker's ability to hide his location is curtailed by increasing the marking probability, however, the latter is upper-bounded due to sampling constraints. In typical IP internets, the attacker's address can be localized to within 2-5 equally likely sites which renders PPM effective against single source attacks. Under distributed DoS attacks, the uncertainty achievable by the attacker can be amplified, which diminishes the effectiveness of PPM.

Denial of service (DoS) attack on the Internet has become a pressing problem. In this paper, we describe and evaluate route-based distributed packet filtering (DPF), a novel approach to distributed DoS (DDoS) attack prevention. We show that DPF achieves proactiveness and scalability, and we show that there is an intimate relationship between the effectiveness of DPF at mitigating DDoS attack and power-law network topology.The salient features of this work are two-fold. First, we show that DPF is able to proactively filter out a significant fraction of spoofed packet flows and prevent attack packets from reaching their targets in the first place. The IP flows that cannot be proactively curtailed are extremely sparse so that their origin can be localized---i.e., IP traceback---to within a small, constant number of candidate sites. We show that the two proactive and reactive performance effects can be achieved by implementing route-based filtering on less than 20% of Internet autonomous system (AS) sites. Second, we show that the two complementary performance measures are dependent on the properties of the underlying AS graph. In particular, we show that the power-law structure of Internet AS topology leads to connectivity properties which are crucial in facilitating the observed performance effects.

End-to-end QoS control over best-effort and differentiated service networks which exhibit variability in their exported service properties looms as an important challenge. In previous work, we have shown how packet-level adaptive FEC can be used in dynamic networks to facilitate invariant user-specified QoS in an end-to-end manner. This paper addresses two important problems-self-similar burstiness and performance degradation of reactive controls subject to long feedback loops-complementing the stability/optimality considerations studied earlier. First, for adaptive redundancy control to be effective, its susceptibility to correlated packet drops and queueing delays stemming from self-similar burstiness must be fortified. Second, to preserve FEC's viability over ARQ when transporting real-time traffic in WAN, preactivity must be injected to offset the performance degradation of reactive feedback controls when subject to long RTT. In this paper, we use the recently advanced multiple-time-scale congestion control framework-first investigated in the throughput maximization context-to endow adaptive redundancy control with both selective protection against self-similar burstiness as well as preactivity to feedback redundancy control. We analyze, implement, and benchmark our protocol-AFEC-MT-in the context of transporting periodic real-time traffic, in particular, MPEG video.

With the proliferation of workstation clusters connected by high-speed networks, providing efficient system support for concurrent applications engaging in nontrivial interaction has become an important problem. Two principal barriers to harnessing parallelism are: efficient mechanisms that achieve transparent dependency maintenance while preserving semantic correctness; and scheduling algorithms that match coupled processes to distributed resources while explicitly incorporating their communication costs. This paper describes a set of performance features, their properties, and implementation in a system support environment called DUNES that achieves transparent dependency maintenance-IPC, file access, memory access, process creation/termination, process relationships-under dynamic load balancing. The two principal performance features are push/pull-based active and passive end-point caching and communication-sensitive load balancing. Collectively, they mitigate the overhead introduced by the transparent dependency maintenance mechanisms. Communication-sensitive load balancing, in addition, affects the scheduling of distributed resources to application processes where both communication and computation costs are explicitly taken into account. The DUNES architecture endows commodity operating systems with distributed operating system functionality while achieving transparency with respect to their existing application base. DUNES also preserves semantic correctness with respect to single processor semantics.

With the proliferation of high-speed networks and networked services, provisioning differentiated services to a diverse user base with heterogeneous QoS requirements has become an important problem. The traditional approach of resource reservation and admission control provides both guarantees and graded services, however, at the cost of potentially underutilized resources and limited scalability. We describe a WAN QoS provision architecture that adaptively organizes best-effort bandwidth into stratified services with graded QoS properties such that the QoS needs of a diverse user base can be effectively met. This architecture-SBS (stratified best-effort service)-promotes a simple user/simple network realization where neither the user nor the network is burdened with complex computational responsibilities. SBS is scalable, efficient, and adaptive, and it complements the guaranteed service architecture, sharing a common network substrate comprised of GPS routers. It is also a functional complement, provisioning QoS efficiently commensurate with user needs, albeit at the cost of weaker protection. SBS is suited to noncooperative network environments where users behave selfishly and resource contention resolution is mediated by the principle of competitive interaction. A principal feature of SBS is the transformation of user-centric QoS provision mechanisms-a defining characteristic of competitive interaction entailing intimate user control of internal network resources-into network-centric mechanisms while preserving the former's resource allocation paradigm. End-to-end QoS control is facilitated by decentralized control based on Lagrangian optimization-achieve a target end-to-end QoS at minimum cost or resource usage-which, in turn, is amenable to distributed implementation. SBS achieves per-flow QoS control with zero per-flow state at routers and a packet header whose size is independent of hop count. SBS, in spite of foregoing both resource reservation and admission control, is able to provision stable, graded QoS.

This paper presents an adaptive protocol for packet-level forward error correction in dynamic networks. The objective is to facilitate end-to-end transport of real-time traffic whose timing constraints rule out the use of retransmission-based congestion control and quality of service (QoS) provision schemes. The degree of redundancy injected into the network is adjusted as a function of the network state, decreasing when the network is well-behaved and increasing when it is not. The control problem is nontrivial due to the fact that increased redundancy, beyond a certain point, can backfire resulting in self-induced congestion which impedes the timely recovery of information at the receiver. First, we give a comprehensive analysis of the control problem associated with adaptive forward error correction, concentrating on the dynamics of a particular protocol called adaptive forward error correction (AFEC). We show that instabilities can arise from two distinct sources-the desired operating point and network delay-and we give solutions to handle them. The first factor is determined by the stringency of the application's QoS requirements, making its achievement potentially perilous. Second, we present simulation results showing the dynamics of AFEC and comparing its performance against static FEC. We study the efficacy of AFEC under structural perturbations and bursty traffic conditions, and we quantitatively estimate the redundancy recovery rate function which relates redundancy to the QoS rendered at the receiver. We show that the curve is unimodal, its curvature being determined by whether a connection is packet loss or delay dominated.

This paper presents an architecture for multi-class quality of service (QoS) provision in wide area networks. Users or applications are assumed to be selfish and end-to-end QoS is determined by the service levels received by an application traffic flow at each router or switch along a path. In previous work, we have given a comprehensive analysis of the noncooperative multi-class QoS provision game for single-switch systems showing when Nash equilibria exist and under what conditions they are Pareto and/or system optimal. In this paper, we propose a specific network architecture for facilitating noncooperative and of provision in many-switch systems such as the Internet with emphasis on realizability. We shield the user from having to choose the service classes on the switches along a route-a hard combinatorial optimization problem even assuming perfect knowledge about network state-while preserving the basic premise of selfishness. This is achieved by employing a set of QoS agents installed at routers which act on behalf of an user's traffic flow. The QoS agent intercepts packets entering a switch implementing generalized processor sharing (GPS) packet scheduling-and using only constant space packet header overhead and zero per-connection state at the routers-determines which service class to assign the packet to to satisfy the user's end-to-end QoS requirement at minimum cost. We present simulation results which show that our architecture is able to provide stable, stratified services to application traffic with diverse QoS requirements.

Asynchronous parallel computing can result in high message generation rates, thus triggering network congestion. We characterize the communication requirements of a large class of supercomputing applications falling under the category of fixed-point problems amenable to solution by parallel iterative methods. In particular, we concentrate on asynchronous iterative algorithms whose communication/computation ratio is especially high resulting in degraded effective throughput if communication is not managed properly. Second, we show the effects of network contention and asynchrony on application performance in a local-area network environment and investigate methods of solution. Our approach is based on a congestion control algorithm called ‘warp control’ whose adaptive properties are exploited to yield significant performance enhancements when network contention is high. Although tested in a LAN environment for experimental control purposes, our solution follows the end-to-end paradigm and refrains from exploiting special MAC-layer properties to achieve applicability to general WAN environments. Third, we provide a framework wherein efficient congestion control can be facilitated, encompassing methods acting at the application layer as well as the transport/network layer, with emphasis on application-driven control. We conclude with a discussion of our experimental results and special issues arising in high-bandwidth ATM networks.

Measurements of LAN and WAN traffic show that network traffic exhibits variability on different scales. We examine a mechanism that gives rise to self-similar network traffic and discuss performance. The mechanism we study is the transfer of files or messages whose size is drawn from a heavy-tailed distribution. In a realistic client/server network the degree to which file sizes are heavy-tailed can directly determine the degree of traffic self-similarity at the link level. This causal relationship is robust relative to changes in network resources, network topology, the influence of cross-traffic, and the distribution of interarrival times. Properties of the transport layer play an important role in preserving and modulating this relationship. The reliable transmission and flow control mechanisms of TCP serve to maintain the long-range dependency structure induced by heavy-tailed file size distributions. In contrast, if a non-flow-controlled and unreliable (UDP-based) transport protocol is used, the resulting traffic shows little self-similarity: although still bursty at short time scales, it has little long-range dependence. Performance implications of self-similarity are discussed as represented by various performance measures. Increased self-similarity as expected, results in degradation of performance. Queueing delay, in particular is discussed. Throughput-related measures such as packet loss and retransmission rate, however increase only gradually with increasing traffic self-similarity as long as reliable, flow-controlled transport protocol is used.

For communication-intensive parallel applications, the maximum degree of concurrency achievable is limited by the communication throughput made available by the network. In previous work (Heddaya et al., 1994) we showed experimentally that the performance of certain parallel applications running on a workstation network can be improved significantly if a congestion control protocol is used to enhance network performance. We characterize and analyze the communication requirements of a large class of supercomputing applications that fall under the category of fixed-point problems, amenable to solution by parallel iterative methods. This results in a set of interface and architectural features sufficient for the efficient implementation of the applications over a large-scale distributed system. In particular, we propose a direct link between the application and network layer, supporting congestion control actions at both ends. This in turn enhances the system's responsiveness to network congestion, improving performance. Measurements are given showing the efficacy of our scheme to support large-scale parallel computations.>

This paper presents a distributed, end-to-end congestion control protocol for use in high-traffic packet switched networks. The network is represented as a stochastic single-server queue, with arrival rates being the control variables. A time-stamp based measure of network state called warp is defined, and it is shown to be an estimator of network utilization. Congestion is modeled explicitly using unimodal load-service rate functions, and its monotonicity property is exploited to yield characterizations of stability and optimality. A protocol based on "perfect" information is analyzed, whose prowess is then shown to be emulated by one which only uses locally computable, delayed information. The main effect of a unimodal load-service function is to induce a division of the phase space into stable and unstable regions, the optimal operating point being its "boundary." Protocols are devised for dealing with each regime separately, rate adjustment protocol being the control that guides the system to the optimal operating point. Proactive rate protocol and reactive rate protocol deal with the issue of the optimal operating point being near to the unstable zone. Protocols for handling fairness and structural perturbation augment the basic suite. The analysis is supported by simulations showing the global dynamical properties of the system.