Request Routing in Cache Meshes

Abstract The current document concerns the inclusion of routing data into neighbor cache selection algorithms. The background for this work is derived from experiences in the DFN Caching Project. During this project, 10 cache servers were installed at the central nodes of the German Broadband Research Network (B-WiN) as an integral part of the backbone infrastructure. The fact that B-WiN contains several exchange points to other networks aggravates an intelligent and effective domain name based routing of requests within this cache hierarchy. Especially exchange points in Frankfurt and Hanover to other national networks — all containing Web servers in the same toplevel domain .de — demonstrate the insufficient possibilities in common implementations of neighbor selection algorithms. The proposed solution consists of an extended caching software to gather routing data with the help of Whois requests. These data can be evaluated in the cache selection algorithm. Starting with basic design issues, the focus of this paper is on a detailed description of an extended neighbor selection algorithm. It is concluded by the description of a full implementation in the Squid cache software.

[1]  Patrik Fältström,et al.  How to Interact with a Whois++ Mesh , 1996, RFC.

[2]  Helmut Pralle,et al.  Load and Traffic Balancing in Large Scale Cache Meshes , 1998, Comput. Networks.

[3]  Duane Wessels,et al.  ICP and the Squid web cache , 1998, IEEE J. Sel. Areas Commun..