Optimization Issues in Distributed Computing Systems Design

In recent years, we observe a growing interest focused on distributed computing systems. Both industry and academia require increasing computational power to process and analyze large amount of data, including significant areas like analysis of medical data, earthquake, or weather forecast. Since distributed computing systems – similar to computer networks – are vulnerable to failures, survivability mechanisms are indispensable to provide the uninterrupted service. Therefore, in this paper we propose a novel 1 + 1 protection mechanism. We formulate an ILP model related to optimization of survivable distributed computing systems. The objective is to allocate computational tasks to computing nodes and dimension network capacity in order to minimize the operational cost of the computing system and satisfy survivability constraints. To facilitate high computational complexity caused by NP-completeness in solving the ILP problem, we propose additional cut inequalities that can be applied for the branch-and-cut algorithm. We consider the cut-and-branch variant of the B&C algorithm. To construct additional cut inequalities we use the idea of cover inequalities and mixed integer rounding (MIR) inequalities. Results of experiments conducted using CPLEX solver are provided and discussed.