We show how to significantly improve the mean response time seen by both uploaders and downloaders in peer-to-peer data-sharing systems. Our work is motivated by the observation that response times are largely determined by the performance of the peers serving the requested objects, that is, by the peers in their capacity as servers. With this in mind, we take a close look at this server side of peers, characterizing its workload by collecting and examining an extensive set of traces. Using trace-driven simulation, we demonstrate the promise and potential problems with scheduling policies based on shortest-remaining-processing-time (SRPT), the algorithm known to be optimal for minimizing mean response time. The key challenge to using SRPT in this context is determining request service times. In addressing this challenge, we introduce two new estimators that enable predictive SRPT scheduling policies that closely approach the performance of ideal SRPT. We evaluate our approach through extensive single-server and system-level simulation coupled with real Internet deployment and experimentation.

Packet forwarding prioritization (PFP) in routers is one of the mechanisms commonly available to network administrators. PFP can have a significant impact on the performance of applications, the accuracy of measurement tools' results and the effectiveness of network troubleshooting procedures. Despite their potential impact, no information on PFP settings is readily available to end users. In this paper, we present an end-to-end approach for packet forwarding priority inference and its associated tool, POPI. This is the first attempt to infer router packet-forwarding priority through end-to-end measurement. Our POPI tool enables users to discover such network policies through the monitoring and rank classification of loss rates for different packet types. We validated our approach via statistical analysis, simulation, and wide-area experimentation in PlanetLab. As part of our wide-area experiments, we employed POPI to analyze 156 random paths across 162 PlanetLab nodes. We discovered 15 paths flagged with multiple priorities, 13 of which were further validated through hop-by-hop loss rates measurements. In addition, we surveyed all related network operators and received responses for about half of them confirming our inferences.

Overlay-based multicast has been proposed as a key alternative for large-scale group communication. There is ample motivation for such an approach, as it delivers the scalability advantages of multicast while avoiding the deployment issues of a network-level solution. As multicast functionality is pushed to autonomous, unpredictable end systems, however, significant performance loss can result from their higher degree of transiency when compared to routers. Consequently, a number of techniques have recently been proposed to improve overlays' resilience by exploiting path diversity and minimizing node dependencies. Delivering high application performance at relatively low costs and under high degree of transiency has proven to be a difficult task. Each of the proposed resilient techniques comes with a different trade-off in terms of delivery ratio, end-to-end latency and additional network traffic. In this paper, we review some of these approaches and evaluate their effectiveness by contrasting the performance and associated cost of representative protocols through simulation and wide area experimentation.

One of the most important challenges of selforganized, overlay systems for large-scale group communication lies in these systems ability to handle the high degree of transiency inherent to their environment. While a number of resilient protocols and techniques have been recently proposed, achieving high delivery ratios without sacrificing end-to-end latencies or incurring significant additional costs has proven to be a difficult task. In this paper we review some of these approaches and experimentally evaluate their effectiveness by contrasting their performance and associated cost through simulation and widearea experimentation.

We explore the feasibility of streaming applications over DHT-based substrates. In particular, we focus our study on the implications of bandwidth heterogeneity and transiency, both characteristic of these systems target environment. Our discussion is grounded on an initial evaluation of SplitStream, a representative DHT-based cooperative multicast system.

This work introduces Nemo, a novel peer-to-peer multicast protocol that aims at achieving this elusive goal. Based on two techniques: (1) co-leaders; and, (2) triggered negative acknowledgments (NACKs), Nemo's design emphasizes conceptual simplicity and minimum dependencies (Anderson et al., 2002), thus achieving, in a cost-effective manner, performance characteristics resilient to the natural instability of its target environment. Simulation-based and wide-area experimentations show that Nemo can achieve high delivery ratios (up to 99.98%) and low end-to-end latency similar to those of comparable protocols, while significantly reducing the cost in terms of duplicate packets (reductions > 85%) and control related traffic, making the proposed algorithm a more scalable solution to the problem.