Split-System: The New Frontier of Cloud Computing

Abstract

Split-protocol computing paradigm uses web services on geographically distributed Web servers. A network of large Split-servers that form the cloud to handles computing and storage task that would otherwise create a heavy CPU utilization for the traditional individual server. The split-protocol concept was devised for load balancing and faster data transmission. Connection and data communication phases are clearly divided between servers without involving the client or the central controlling mechanism, like a dispatcher or a load balancer. In an earlier paper, we demonstrated that the application Split-protocol produces higher performance compared to traditional clusters. Based on the need, different types of split configurations are implemented for throughput, response time, and connection time. The observed throughput improvement was within the range 6.5% - 25% over non-split systems. This paper first time examines the empirical results of Split-system to understand its analytical behavior and compares with traditional non-split systems. Also, this paper prepares the groundwork for the scientific understanding of Split-systems. The split-system technique with given sets of constraints can produce better throughput than conventional equivalent server systems. The performance improvement resulted because of the Split-system operates as a fine-grained pipeline of the parallel system.