In order to reduce average delay and bandwidth usage in the Web, geographically dispersed servers often store copies of popular objects. For example, with network caching, the origin server stores a master copy of the object and geographically dispersed cache servers pull and store copies of the object. With site replication, objects stored at master are replicated into secondary sites. In this paper we propose a newnetwork application, Location Data System (LDS), that allows an arbitrary host to obtain the IP addresses of the servers that store a specified URL. Our networking application is an extension to the Domain Name System (DNS), requires only small changes to the domain name servers, and can be deployed incrementally. For the case of network Web caching, we elaborate on our proposal to allow a cache to (i) up-date a distributed database when it stores or evicts objects, and (ii) push objects to parent caches in order to improve delay and bandwidth usage. For the case of mirrored servers, we show how a client can obtain a list of all servers mirroring all or part of the desired site. LDS applied to partially mirrored sites generates substantially less DNS traffic than LDS applied to caching. Finally, we discuss how a host can use the location data in order to make intelligent decisions about where to retrieve desired objects.
