Abstract:

Minimizing the busy time of servers is an effective way to save energy in parallel scheduling applications in cloud computing, however, most of the existing busy time energy saving strategies are at the expense of job scheduling performance, and they are unable to cooperate with other job scheduling algorithms. We propose an effective energyefficient optimization algorithm based on busy time in parallel scheduling, called BTEOA. Firstly, based on the available resources of current servers, we divide the job request queue into a special job window and a general job window. Secondly, according to the completion time of job requests in the special job window, we can execute these job requests in the servers so as to ensure that the total busy time of all servers is locally optimal. Finally, when the execution of all job requests in the special job window finishes completely, the BTEOA continues to divide the general job window into a new special job window and a new general job window until the remaining job requests are all executed completely. During the job scheduling, the BTEOA can ensure the job scheduling performance unaffected by keeping the original job queuing model static. Instances and experimental results demonstrate the performance of the BTEOA, which can save energy consumption while does not affect the performance of job scheduling. Besides, it can also be used jointly with other job scheduling algorithms.

Key words: