Allocation of Computing Resources in Two-Tier Hierarchical Wide Area Networks: An Heuristic Algorithm

This paper addresses the problem of simultaneously design of large WAN satisfying quality demands and allocation of computing resources across the network. Model of the grid structure with hierarchical communication network is presented. Data is generated in the central node, then divided into blocks and distributed among nodes which can be located in different autonomous systems represented by 1st level networks. Nodes participating in the grid structure perform computation of blocks. Next, results are sent to the central repository. Both computation and data transfer between node and the central repository introduces a cost. The goal is to select allocation of computing resources, channel capacities and flow routes in order to minimize linear combination of the total average delay per packet and leasing cost of channel capacities of 2nd level network, cost of computations in network nodes, subject to budget constraint. An heuristic algorithm solving presented problem is proposed.