Showing papers by "Jia Wang published in 2002"
06 Nov 2002
TL;DR: The high volume and good stability properties of P2P traffic suggests that the P1P workload is a good candidate for being managed via application-specific layer-3 traffic engineering in an ISP's network.
Abstract: The use of peer-to-peer (P2P) applications is growing dramaticaliy, particularly for sharing large video/audio files and software. In this paper, we analyze P2P traffic by measuring flow-level information collected at multiple border routers across a large ISP network, and report our investigation of three popular P2P systems -- FastTrack, Gnutella, and DirectConnect. We characterize the P2P traffic observed at a single ISP and its impact on the underlying network. We observe very skewed distribution in the traffic across the network at different levels of spatial aggregation (IP, prefix, AS). All three P2P systems exhibit significant dynamics at short times scale and particularly at the IP address level Still, the fraction of P2P traffic contributed by each prefix is much more stable than the corresponding distribution of either Web traffic or overall traffic. The high volume and good stability properties of P2P traffic indicates that the P2P workload is a good candidate for being managed via application-specific layer-3 traffic engineering in an ISP's network.
499 citations
06 Nov 2002
TL;DR: It is shown that the small number of popular destinations responsible for the bulk of Internet traffic have remarkably stable BGP routes, and operators can engineer their networks under the assumption that the BGP advertisements associated with most of the traffic are reasonably stable.
Abstract: The Border Gateway Protocol (BGP) plays a crucial role in the delivery of traffic in the Internet. Fluctuations in BGP routes cause degradation in user performance, increased processing load on routers, and changes in the distribution of traffic load over the network. Although earlier studies have raised concern that BGP routes change quite often, previous work has not considered whether these routing fluctuations affect a significant portion of the traffic. This paper shows that the small number of popular destinations responsible for the bulk of Internet traffic have remarkably stable BGP routes. The vast majority of BGP instability stems from a small number of unpopular destinations. We draw these conclusions from a joint analysis of BGP update messages and flow-level traffic measurements from AT&T's IP backbone. In addition, we analyze the routing stability of destination prefixes corresponding to the NetRating's list of popular Web sites using the update messages collected by the RouteViews and RIPE-NCC servers. Our results suggest that operators can engineer their networks under the assumption that the BGP advertisements associated with most of the traffic are reasonably stable.
328 citations
Proceedings Article•
10 Jun 2002
TL;DR: A novel, precise, and efficient technique for finding the associations of client to local DNS servers and concludes that DNS is good for very coarse-grained server selection, since 64% of the associations belong to the same Autonomous System.
Abstract: Content Distribution Networks (CDNs) attempt to improve Web performance by delivering Web content to end-users from servers located at the edge of the network. An important factor contributing to the performance improvement is the ability of a CDN to select servers in the proximity of the requesting clients. Most CDNs today use the Domain Name System (DNS) to make such server selection decisions. However, DNS provides only the IP address of the client’s local DNS server to the CDN, rather than the client’s IP address. Therefore, CDNs using DNS-based server selection assume that clients are “close” to their local DNS servers. To quantify the proximity between clients and their local DNS servers, we propose a novel, precise, and efficient technique for finding the associations of client to local DNS servers. We collected more than 4.2 million such unique associations in three months. From this data, we study the impact of proximity on DNS-based server selection using four different proximity metrics. We conclude that DNS is good for very coarse-grained server selection, since 64% of the associations belong to the same Autonomous System. DNS is less useful for finergrained server selection, since only 16% of the client and local DNS associations are in the same network-aware cluster [13] (based on BGP routing information from a wide set of routers). As an application of this methodology, we evaluate DNS-based server selection in three of the largest commercially deployed CDNs to study its accuracy.
174 citations
06 Nov 2002
TL;DR: This work proposes application specific peering that allows searches for a resource to be directed to a copy of resource on an ISP’s network, using a novel technique that automaticly classifies traffic as belonging to an ISP, its c ustomers, peers, or their customers.
Abstract: Peer-to-Peer (P2P) applications [4], [6], [2], [3] have increased traffic significantly. We propose application specific peering that allows searches for a resource (using existing or modified P2P protocols) to be directed to a copy of resource on an ISP’s network, using a novel technique that automatica lly classifies traffic (shown in Fig 1) as belonging to an ISP, its c ustomers, peers, or their customers. We examined a large amoun t of traffic belonging to three popular P2P protocols and used a graph transformation to detect connected components. We pa rtition traffic in each component into signaling and data sect ions based on request/response sizes of P2P protocols and extrac t high traffic volume (by bytes) entities by examining large co nnected components in the graph. We discuss the important com ponents of our traffic classification system here. We analyzed router-level data from multiple Internet Gatew y Routers (IGRs) across AT&T IP backbone using Cisco’s NetFlow services. Note that statistics are from a subset of IGRs at some AT&T’s peering links. They are not representative of ov erall traffic pattern on AT&T IP backbone. We examined DirectConnect[2], Gnutella[3], and FastTrack[4], extracting re cords that matched the default application ports (source or desti nation, 411/412, 6346/6347, and 1214 (FastTrack) respective ly), involving TCP traffic. Data was gathered over three separate months in 2001, each lasting 5-7 days. The number of netflow records varies from a low of 0.5M (DirectConnect) to a high
14 citations