One of approaches to reducing energy consumption in a data center is to power down a group of servers. In this paper, we consider a power management scheme for distributed parallel processing over clusters of servers, where part of servers in each cluster are turned off in power-saving mode. We model the system as a multi-server queue in which the service time of a job depends on the state of a background process at the beginning of the job service. We analyze the joint distribution of the number of jobs in system and the state of the background process, deriving the mean job-response time and mean amount of energy consumption. In numerical examples, we investigate how the mean job-response time and energy consumption are affected by energy saving level and the number of clusters.
[1]
Yanpei Chen,et al.
Energy efficiency for large-scale MapReduce workloads with significant interactive analysis
,
2012,
EuroSys '12.
[2]
Randolph Nelson,et al.
Probability, Stochastic Processes, and Queueing Theory
,
1995
.
[3]
Vaidyanathan Ramaswami,et al.
Introduction to Matrix Analytic Methods in Stochastic Modeling
,
1999,
ASA-SIAM Series on Statistics and Applied Mathematics.
[4]
Phuoc Tran-Gia,et al.
A queuing analysis of an energy-saving mechanism in data centers
,
2012,
The International Conference on Information Network 2012.