Linux Cluster Environment, Job Scheduling Algorithm Realization

In cluster system, the job scheduling receives user requests and the scheduling policies used to choose the appropriate nodes. Job scheduling policies determines the efficiency of the cluster system; in particular, when users submit the large jobs, good job scheduling policies can greatly speed up the implementation of speed, therefore, job scheduling policies is a key factor in the utilization of the cluster system and the efficiency of the parallel implementation.This thesis deep analysis the job scheduling process and job scheduling policies of OpenPBS, focus on the reservations backfill scheduling strategies A cluster job scheduling strategy is put forward, which called combination of nodes load and reservations backfill scheduling strategy. The main work of this paper is as follows:ⅰWe study the job scheduling strategy of OpenPBS and these classic job scheduling strategies, analyze their advantages and disadvantages, on this basis, discuss the reservations backfill scheduling.ⅱThe load-balancing is the key factor to the utilization ratio and performance of cluster system, this paper presented a combination of nodes load and reservations backfill scheduling strategy, make full use of cluster system resources and increase system throughput and utilization ratio.ⅲWe added this job scheduling strategy in the scheduling module of OpenPBS, the scheduling module made full use of the job management framework of OpenPBS, and afforded reservation and scheduling functions.ⅳIn the experimental test, we set up a small cluster system and compare this job scheduling strategy with FirstFit scheduling strategy and FCFS scheduling strategy. These results showed that this job scheduling strategy had better performance than other on load-balancing and system resources throughput. l;...