10 Executive summary

[DESIRE] [DESIRE Web cache] [Web cache Architecture]

Drawing on the experiences in the EC 4th Framework Project DESIRE, funded by the European Union, this document identifies some of the pitfalls and possibilities of building Web caching meshes. This document gives guidelines and outlines strategies for building a functional Web caching system with cooperating Web caches.

Institutions and Internet Service Providers should provide Web cache servers for their users; and teach users to use caching. Web caching halves your Web related network traffic and reduces download time for popular Web pages.

In order to have a high quality of service, a mesh of Web cache servers using cooperative Web caching that supports the Internet Cache Protocol is the best solution.

The architecture of a Web caching mesh is highly depending on the network topology and the cost factors of the network.

The emphasis is on building Internet Service Provider wide caching systems, but architectural considerations for inter-network agreements on Web caching is discussed and examples of policies for international Web caching agreements are given. The impact of Web caching as a routing issue is discussed.

A summary of the advice given in the document:

• a single caching server is not a good solution (not scalable), and therefore the use of a mesh of cooperating caching servers (a caching mesh) is necessary
• the operation of a caching mesh must be resistant to the failure of one of its caching servers, so therefore a robust protocol such as the ICP and the use of apc scripts is advisable
• more than 3 levels of caches in a caching mesh should be avoided
• Web clients situated in a LAN, using a local caching server on this LAN should preferably turn off local disk caching; Web clients situated at an dial-up line should preferably use their local disk cache
• a proper architecture for a caching mesh is highly dependent on network topology and care should be taken not to violate routing policies
• place Web cache servers on the users side of a bottleneck
• place Web cache servers close to the flow of traffic
• place a local Web cache server close to your internet connection
• place Web cache servers where more networks join together
• upper level Web caches should be placed where more networks join together
• the caching servers in a caching mesh should tune the lifetimes of the objects they cache, as improper settings may result in less performance of the caching mesh

Ingrid Melve