Abstract:
Web caching is one of many applications argued to benefit from a switch from a client-server to a peer-to-peer architecture. Several projects suggested the use of a network of peers to provide cached web content in order to help sites survive a burst of user requests. In this thesis, we present a new system that targets web sites with dynamic content and allows them to use a group of a variable number of (volunteered) peers to provide cached web content to the clients.
The main objective of our system is to increase the capacity of a web server, and to reduce the average end user latency. To clarify how our approach would work in a real world web application, we implemented various components of our system in addition to an Internet forums application as a case study of dynamic web sites.
The Internet forums implementation will be used to evaluate our claims. Success will be measured by measuring the average end user latency and upload bandwidth of the web server using our approach, and comparing these values to their counterparts in a normal client-server system.
In our experiments, we used both static and dynamic content. In the first case, all applied requests are view requests, and in the latter, requests include both views and forums replies requests. We varied various parameters, including the rate of requests per second and the number of DHT peers. Results showed that our approach can help overloaded web sites to withstand a high rate of incoming requests while yielding satisfactory end user latency.
