Performance Analysis of an Admission Controller for CPU- and I/O-Intensive Applications in Self-Managing Computer Systems
Mais Nijim, Tao Xie, and Xiao Qin
Department of Computer Science New Mexico Institute of Mining and Technology 801 Leroy Place, Socorro, New Mexico 87801
With rapid advances in processing power, network bandwidth, and storage capacity, computer systems are increasingly becoming extremely complex. Consequently, it becomes expensive and difficult for human beings to manually manage complex computer systems. This problem can be effectively tackled by self-managing computer systems, which are intended to meet high performance requirements in a dynamic computing environment. In this paper, we develop a performance model for self-manage computer systems under dynamic workload conditions, where both CPU- and I/O-intensive applications are running in the systems. In particular, we design in this paper a 2-dimenssional Markov chain model with two different arrival and service rate of CPU- and I/O-intensive jobs. Importantly, two serving probabilities with respect to CPU- and I/O intensive jobs are derived. To validate the analytical model, we developed an adaptive admission controller in which the model is incorporated. Experimental results demonstratively show that the controller is capable of achieving high performance for computer systems under workloads exhibiting high variability.
This paper appeared in ACM Operating Systems Review, Vol. 39, No. 4, pp.37-45, October, 2005.