How to Avoid Herd Behavior: A Stochastic Multi-Choice Scheduling Algorithm and Parameters Analysis in Grid Scheduling

Abstract

Large distributed systems, such as grid computing and cloud computing, promise to supply users with high performance. Consequently, scheduling is currently becoming a crucial problem. Herd behavior is a common phenomenon which causes severe performance decrease in the systems caused by bad scheduling behaviors. In this paper, based on the theoretical results of the homogeneous balls and bins model, it is proposed that a new and unique stochastic algorithm is used to avoid herd behavior. Experiments address that the multi-choice strategy can decrease herd behavior in large-scale sharing environment, at the same time providing increased scheduling performance and causing less scheduling burden than greedy algorithms. Distributed Hash Table (DHT) is used to organize grid computing resources. In the case of 1000 resources, the simulations show that for the heavy load (i.e., system utilization rate 0.5), the multi-choice algorithm reduces the number of incurred herds by a factor of 36, the average job waiting time by a factor of 8, and the average job turn-around time by 12% compared to the greedy algorithm. Moreover, in the cases of 2000 and 4000 nodes, two parameters (replica and d-group) are analyzed based on how they affect the performance of the algorithm. It is observed that there is an inflexion in the performance curve. Finally, a theoretic analysis of the algorithm performance is presented.