Bounds on the multi-client incremental computing

It has been shown that, in a network computing platform, the arrival order of input data has serious impact on the performance of a computing server. In a multi-client environment, each client competes with the others for the network bandwidth. In this paper, we study the bounds on the time needed to compute the services requested from multiple clients. First, an abstract model of network computing for multiple clients is proposed. Then, based on the model, bounds on the time are developed for a computing server to compute services specified by a class of special sequences. Finally, simulations are conducted to illustrate the effect of message patterns on the performance of a computing server.