The Server Allocation Problem with non-identical machines: A meta-heuristic approach

•The paper addresses the Server Allocation Problem with non-identical machines.•The goal is to minimize the total cost while assuring a minimum target throughput.•The solution approach is a Variable Neighborhood Search meta-heuristic.•The effectiveness has been tested through a comprehensive numerical analysis.•The proposed algorithm is compared with five state-of-the-art meta-heuristics. Firms often need to drastically improve the production rate to meet customer demand, also according to some forecast scenarios. To this end, in the design phase, one of the most significant problem is how to allocate new resources to design efficient production systems. This paper addresses the resource (or server) allocation problem of series-parallel production lines where non-identical machines can be assigned at each stage. Machines can be chosen from a list of versions available on the market, whose purchase cost depends on their related processing speed. The decision problem consists in selecting both number and version of machines to be allocated at each production stage. The goal is to minimize the total cost while assuring a minimum target throughput. To solve the aforementioned server allocation problem (SAP), we accomplished three major research steps. First, we devised an efficient evaluative simulation algorithm that, properly combined with a pattern-based problem representation, allows handling non-identical machines at each stage during the optimization phase. Secondly, we developed a specific constructive heuristic for generating a feasible solution of the SAP. Finally, we used such heuristic solution to speed up the convergence of a new Variable Neighborhood Search (VNS) algorithm, whose effectiveness has been tested through a comprehensive numerical analysis involving five alternative state-of-the-art meta-heuristics.