Scalable load balancing using virtualization based on approximation

The number of users and services on Internet are increasing day by day resulting in high traffic and load on the web servers. This is in turn increasing the service time of web requests and degrading the quality of service. A well known solution to this problem is replication of content using cluster of web servers. An efficient server load balancing policy is required to achieve scalability and high performance of the service offered by cluster of web servers. Under dynamic, secure and database driven loads, existing load balancing strategies are suffering from performance degradation. In this paper, we proposed Scalable Load Balancing using Virtualization based on Approximation Algorithm (SLBVA). SLBVA is an estimation strategy as it is challenging to correctly measure the load on each web server of a cluster. SLBVA algorithm is capable of offering guarantees for different client priorities, such as premium customers and default customers. We show that using SLBVA strategy web servers are able to maintain Service Level Agreements (SLA) without the need of a priori over-dimensioning of server resources. This is achieved by taking the real perspective of the service requests using the measurement of arrival rates at that time and judiciously discard some requests from the default clients if the default customers traffic is high. If the arrival rate of premium customers goes beyond the capacity of cluster, we will increase the capacity of cluster using virtualization by utilizing the unused servers from the under-utilized server farms. We analyzed and compared the experimental results of SLBVA algorithm with the results of the very popular load balancing algorithm, Weighted Round Robin (WRR). We show that even though the SLBVA strategy takes a little more server processing resources than WRR, it is capable to render assurances unlike WRR.