Resource Allocation and Consolidation in a Multi-Core Server Cluster Using a Markov Decision Process Model

Yanzhi Wang,  Shuang Chen,  Hadi Goudarzi,  Massoud Pedram
University of Southern California


Resource allocation is one of the most important challenges in the distributed computing systems especially when the clients have some Service Level Agreements (SLAs) and the total profit depends on how the system can meet these SLAs. In this paper, an SLA-based resource allocation problem in a server cluster is considered. The objective is to maximize the total profit, which is the total price gained from serving the clients subtracted by the energy cost of the server cluster. A joint optimization framework is proposed, comprised of request dispatching, dynamic voltage and frequency scaling (DVFS) for individual cores, as well as server-level and core-level consolidations. Each core in the cluster is modeled using a continuous-time Markov decision process (CTMDP). A near-optimal hierarchical solution is proposed, consisting of a central manager and distributed local agents. Each local agent employs linear programming-based CTMDP solving method to solve the DVFS problem for the corresponding core. The central manager solves the request dispatching problem and finds the optimal number of turned on cores and servers for request processing, thereby achieving a desirable tradeoff between service request response time and power consumption. Experimental results demonstrate that the proposed near-optimal algorithm consistently outperforms baseline algorithms.