IMPROVING USER RESPONSE TIMES IN APPLICATION DELIVERY NETWORKS THROUGH REDUCTION OF NETWORK AND SERVER LATENCIES

Abstract

An application delivery network (ADN) consists of a set of servers distributed over a large wide area network. In general, ADNs employ two approaches to improve performance in terms of response time: (1) use topological proximity on the network to redirect clients to a closest server, and (2) balance the load of the servers in the original source or the content delivery network using a load balancing algorithm. These two approaches aim to minimize the two major parameters, network latency and server latency, respectively. To accelerate delivery of static content, network latency is the major parameter to minimize. However, to accelerate delivery of dynamically generated content, both parameters are important and a solution requires to achieve a balance between these two approaches. In this paper, we illustrate the need for an integrated approach to this problem. We describe methods for improving the observed performance of an application delivery network by assigning the end user requests to servers. We describe an algorithm which computes such assignments efficiently, so that the assignment task can be performed and adjusted often as the environment changes. The experimental results show that the assignment computation is accurate and close to the optimal.