Performance Analysis of IO Intensive Task Allocation Strategies for Heterogeneous Web Servers

Abstract

The current rate of growth of the World Wide Web has led to an explosion in internet traffic for many popular websites. To overcome the problem of falling quality of service for its customers an efficient approach would be to use a
heterogeneous cluster of nodes which replicate the entire site data. In a centralized system, a master node would load balance the user requests and allocate them to the appropriate node. A web application which mainly provides file sharing services to its users offers a system where the tasks are basically of retrieval based nature and hence more IO intensive. In order to address the allocation problem of these tasks, several IO aware policies have been designed and compared with respect to certain standard performance metrics. The study shows that considering the IO nature of tasks yields significantly better results than other existing algorithms.