Service Consolidation with End-to-End Response Time Constraints

Abstract

In this paper, we address the service consolidation problem: given a data-center, a set of servers and a set of multi-tiered services or applications, the problem is to allocate services to the available servers in order to minimize the number of servers to use while avoiding the overloading of system resources and satisfying end-to-end response time constraints. Exploiting queueing networks theory, we describe a number of linear and non-linear combinatorial optimization problems related to the server consolidation problem. Since their solution is difficult to obtain through standard solution techniques, we propose accurate heuristics which quickly compute a sub-optimal solution and let us deal with hundreds of servers and applications. Experimental results illustrate the impact of the consolidation in data-centers and show that the heuristic solution is almost very close to the optimum.