CDNs with Global Adaptive Request Distribution

Abstract

This paper presents the application of fuzzy logic and neural networks to HTTP request dispatching performed within Content Delivery Network. We propose a global request distribution algorithm called GARD to support request routing to the surrogate servers that deliver the requested content in an efficient manner. The algorithm uses the fuzzy-neural decision-making mechanism to assign each incoming request to the server with the least expected response time. The response time include the transmission time over the network, both for the request and for the response, as well as the time elapsed on the server responding to the request. We demonstrate through the simulations that our algorithm is more effective than popular dispatching policies as Round-Robin and Weighted Round-Robin. We also show that in case of non-evenly loaded environment the GARD algorithm outperforms RTT algorithm which is often used in CDNs.