Abstract not found

Abstract — In this paper, we present the design, implementation, and evaluation of the iPlane, a scalable service providing accurate predictions of Internet path performance for emerging overlay services. Unlike the more common black box latency prediction techniques in use today, the iPlane builds an explanatory model of the Internet. We predict end-to-end performance by composing measured performance of segments of known Internet paths. This method allows us to accurately and efficiently predict latency, bandwidth, capacity and loss rates between arbitrary Internet hosts. We demonstrate the feasibility and utility of the iPlane service by applying it to several representative overlay services in use today: content distribution, swarming peer-to-peer filesharing, and voice-over-IP. In each case, we observe that using iPlane’s predictions leads to a significant improvement in end user performance. 1

Abstract — In recent years, BitTorrent has emerged as a very scalable peer-to-peer file distribution mechanism. While early measurement and analytical studies have verified BitTorrent’s performance, they have also raised questions about various metrics (upload utilization, fairness, etc.), particularly in settings other than those measured. In this paper, we present a simulationbased study of BitTorrent. Our goal is to deconstruct the system and evaluate the impact of its core mechanisms, both individually and in combination, on overall system performance under a variety of workloads. Our evaluation focuses on several important metrics, including peer link utilization, file download time, and fairness amongst peers in terms of volume of content served. Our results confirm that BitTorrent performs near-optimally in terms of uplink bandwidth utilization, and download time except under certain extreme conditions. We also show that low bandwidth peers can download more than they upload to the network when high bandwidth peers are present. We find that the rate-based tit-for-tat policy is not effective in preventing unfairness. We show how simple changes to the tracker and a stricter, block-based tit-for-tat policy, greatly improves fairness. I.

The views and conclusions contained in this document are those of the authors and should not be interpreted as The ability to locate network bottlenecks along end-to-end paths on the Internet is of great interest to both network operators and researchers. For example, knowing where bottleneck links are, network operators can apply traffic engineering either at the interdomain or intradomain level to improve routing. Existing bandwidth measurement tools fail to identify the location of bottleneck links. In addition, they often require access to both end points and generate huge amount of probing packets. These drawbacks make them impractical. In this paper, we present a novel light-weight, single-end active probing tool – Pathneck – based a novel probing technique called Recursive Packet Train (RPT), which allows end users to efficiently and accurately locate bottleneck points to destinations on the Internet. We evaluate Pathneck using trace-driven emulations and wide area Internet experiments. In addition, we conduct extensive measurements on the Internet among carefully selected, geographically diverse probing sources and destinations to study Internet bottleneck properties. We find that Pathneck can successfully detect bottlenecks for over 70% of paths, and most of the bottlenecks are fairly stable. We also report our success on bottleneck

PlanetLab is a continuously-evolving global network research testbed that is simultaneously used by hundreds of researchers for diverse tasks, ranging from shortterm self-contained experiments among PlanetLab nodes to continuously-running Web-accessible services with tens of thousands of non-PlanetLab users. While PlanetLab cannot provide a perfectly-customized environment for every experiment, it has been changing over time, and the base of knowledge of how to best utilize it has also been growing. As a result, many of the early observations researchers made about PlanetLab would change if rechecked today. In this paper, we discuss these issues and explain whether they remain, have been addressed via PlanetLab’s evolution, or can be avoided by the use of best practices. Where possible, we provide quantitative evidence showing the realities of PlanetLab and possible research avenues to further broaden the opportunities for using PlanetLab in network research. 1

In this paper, we take the sample-path approach in analyzing the asymptotic behavior of single-hop bandwidth estimation under bursty cross-traffic and show that these results are provably different from those observed under fluid models of prior work. This difference, which we call the probing bias, is one of the previously unknown factors that can cause measurement inaccuracies in available bandwidth estimation. We present an analytical formulation of "packet probing," based on which we derive several major properties of the probing bias. We then experimentally observe the probing bias and investigate its quantitative relationship to several deciding factors such as probing packet size, probing train length, and cross-traffic burstiness. Both our analytical and experimental results show that the probing bias vanishes as the packet-train length or packet size increases. The vanishing rate is decided by the burstiness of cross-traffic.

Internet topology discovery consists of inferring the inter-router connectivity (“links”) and the mapping from IP addresses to routers (“alias resolution”). Current topology discovery techniques use TTL-limited “traceroute ” probes to discover links and use direct router probing to resolve aliases. The often-ignored record route (RR) IP option provides a source of disparate topology data that could augment existing techniques, but it is difficult to properly align with traceroute-based topologies because router RR implementations are under-standardized. Correctly aligned RR and traceroute topologies have fewer false links, include anonymous and hidden routers, and discover aliases for routers that do not respond to direct probing. More accurate and feature-rich topologies benefit overlay construction and network diagnostics, modeling, and measurement. We present DisCarte, a system for aligning and cross-validating RR and traceroute topology data using observed engineering practices. DisCarte uses disjunctive logic programming (DLP), a logical inference and constraint solving technique, to intelligently merge RR and traceroute data. We demonstrate that the resultant topology is more accurate and complete than previous techniques by validating its internal consistency and by comparing to publicly available topologies. We classify irregularities in router implementations and introduce a divide-and-conquer technique used to scale DLP to Internet-sized systems.

Abstract—We consider the design of bandwidth-demanding broadcasting applications using overlays in environments characterized by hosts with limited and asymmetric bandwidth, and significant heterogeneity in upload bandwidth. Such environments are critical to consider to extend the applicability of overlay multicast to mainstream Internet environments where insufficient bandwidth exists to support all hosts, but have not received adequate attention from the research community. We leverage the multitree framework and design heuristics to enable it to consider host contribution and operate in bandwidth-scarce environments. Our extensions seek to simultaneously achieve good utilization of system resources, performance to hosts commensurate to their contributions, and consistent performance. We have implemented the system and conducted an Internet evaluation on PlanetLab using real traces from previous operational deployments of an overlay broadcasting system. Our results indicate for these traces, our heuristics can improve the performance of high contributors by 10–240 % and facilitate equitable bandwidth distribution among hosts with similar contributions. Index Terms—Bandwidth detection, incentive, multitree, overlay multicast, NAT, saturation detection.

Abstract. With the lack of end-to-end QoS guarantees on existing networks, applications that require certain performance levels resort to periodic measurements of network paths. Typical metrics of interest are latency, bandwidth and loss rates. While the latency metric has been the focus of many research studies, the bandwidth metric has received comparatively little attention. In this paper, we report our bandwidth measurements between PlanetLab nodes and analyze various trends and insights from the data. For this work, we assessed the capabilities of several existing bandwidth measurement tools and describe the difficulties in choosing suitable tools as well as using them on PlanetLab. 1

Abstract – It is increasingly common that computers in residential and hotspot scenarios see multiple access points (APs). These APs often provide high speed wireless connectivity but access the Internet via independent, relatively low-speed DSL or cable modem links. Ideally, a client would simultaneously use all accessible APs and obtain the sum of their backhaul bandwidth. Past work can connect to multiple APs, but can neither aggregate AP backhaul bandwidth nor can it maintain concurrent TCPs across them. This paper introduces FatVAP, an 802.11 driver that aggregates the bandwidth available at accessible APs and also balances their loads. FatVAP has three key features. First, it chooses the APs that are worth connecting to and connects with each AP just long enough to collect its available bandwidth. Second, it ensures fast switching between APs without losing queued packets, and hence is the only driver that can sustain concurrent high throughput TCP connections across multiple APs. Third, it works with unmodified APs and is transparent to applications and the rest of the network stack. We experiment with FatVAP both in our lab and hotspots and residential deployments. Our results show that, in today’s deployments, FatVAP immediately delivers to the end user a median throughput gain of 2.6x, and reduces the median response time by 2.8x. 1