Abstract

www.coralcdn.org Abstract To improve performance, large-scale Internet systems re-quire clients to access nearby servers. While centralized systems can leverage static topology maps for rough net-work distances, fully-decentralized systems have turned to active probing and network coordinate algorithms toscalably predict inter-host latencies. Internet applications seeking immediate adoption, however, must inter-operatewith unmodified clients running existing protocols such as HTTP and DNS.This paper explores a variety of active probing algorithms for locality prediction. Upon receiving an exter-nal client request, peers within a decentralized system are able to quickly estimate nearby servers, using a minimumof probes from multiple vantages. We find that, while network coordinates may play an important role in scalablychoosing effective vantage points, they are not directly useful for predicting a client's nearest servers. 1 Introduction Many replicated Internet systems can improve perfor-mance by servicing clients at nearby hosts. The performance of a distributed web mirror, for example, is highlydependent on the network distance between client and server. Commercial content distribution networks likeAkamai [1] build large maps of the Internet topology and use this information to redirect clients to nearby hosts.These hosts are carefully deployed at specific access sites or behind bottleneck links. This technique for locality pre-diction, however, requires centralized mapping, aggregation, extensive network knowledge, and often ISP-specificheuristics. But by using existing protocols like HTTP and DNS, these systems can achieve immediate and wide-spread use. More recent distributed systems use self-organizingtechniques to reduce the infrastructure's administrative and operational overhead, while still providing service tounmodified clients. Such peer-to-peer systems include static and dynamic content distribution networks [5, 6],distributed hash storage services [7], and new Internet naming systems [13, 14]. Such decentralized systems,

Citations