Influence of Network Characteristics on Application Performance in a Grid Environment

With the availability of high-performance off-the-shelf computers and high-speed widearea networks, large-scale distributed computing environments are growing at an amazing speed. These environments enable massive computations to be performed using a large number of computers connected over WAN (Wide Area Networks). This form of distributed computing (grid computing) dynamically involves a number of heterogeneous computing resources (e.g., CPU, memory, storage, application program, data, and so on.) connected by heterogeneous network resources across geographically dispersed organizations[1][2]. The fundamental challenge for grid computing is to achieve a good performance from multiple distributed applications that share limited and heterogeneous computing and network resources (e.g., in terms of the completion time for each independent application and/or set of related applications). Large amounts of data are now being handled by distributed applications and the time required to transmit data is increasing, which has a significant effect on the performance of an application. Effective share of network resources is therefore critical. There is considerable work on traffic engineering to improve the utilization of network resources. To name a few, Elwalid et al. proposed a decentralized method to balance flows over multiple paths based on the traffic load of 3-5 On Influence of Network Characteristics on Application Performance in a Grid Environment