Design and Development of Efficient Multipath TCP using GMM clustering for Big Data in Public Cloud Data Center

Abstract

Currently new topologies have been introduced in many data centers that provide location independence and larger aggregate bandwidth by making different multiple paths in the network core. This work proposes transport of data from TCP (Transmission control Protocol) to multi-path TCP (MPTCP) for maximum utilization of paths over network flow. In spite of its added advantages, some sort of work on MPTCP to be carried out on cloud environment and further, efficient way of using MPTCP on real-world cloud application still looks like unclear problem. Further, the work also concerned on MPTCP usage in most effective and feasible way for cloud and data center environments over various conditions on network. The experiment is conducted by clustering the public cloud data using Gaussian Mixture Model (GMM) based Expectation and Maximization (EM) algorithm and communicated over a network using MPTCP. The results shows that the proposed method yields high-speed data transfer and low communication delay when compare to traditional TCP technique.