Research On Heterogeneous Cluster Scheduler Algorithm Based On Shuffled Frog Leaping Algorithm

Along with social progress and expanded Information, big data and cloud computing have became a fashion in computer realm. In High Performance Computing cluster occupies the important position because of it’s cheap, easy expand, flexibility and versatility. With regard to Homogeneous cluster the scientific research have had an in-depth study, but to heterogeneous cluster just started. Heterogeneous cluster improve the performance of the cluster because of it’s isomerism, while just about it’s isomerism increase the complexity of scheduler, traditional scheduler algorithm can’t resolve the problem about how to match tasks and heterogeneous resource, which have tremendously restrict heterogeneous cluster’s system performance. In recent years, intelligent scheduling algorithms by simulate natural ecology were payed close attention, especially Shuffled Frog Leaping Algorithm have became an important research direction due to it has a good global search.The paper in order to resolve the problem of heterogeneous cluster’s scheduler, has done the following work:Firstly, analyze, test and compare the traditional scheduler algorithm and a part of intelligence algorithm, prove that Shuffled Frog Leaping Algorithm have a good performance.Secondly, In the base of SFLA, we put forward a part of improvement strategy: take advantage of Max-Int algorithm improve initial frogs’ quality; As for the excellent frogs, treat them Individually, strengthen the excellent frogs’ evolution, and enhance the optimal frogs leading effect, in order to accelerate the speed of evolution in frog populations; Meanwhile, replace the inefficient frog with the new frog with randomly generated, to expand the searching scope of frog populations. In addition, all the compared tests have been taken for each step of the modified shuffled frog leaping algorithm, to prove that the improved measures have sped up the frog populations’ evolution speed, improved the quality of optimal solution, and enhanced the efficiency of heterogeneous clusters.Lastly, for the heterogeneous cluster of the real seismic data processing, the paper create a set of heterogeneous cluster management system by means of the improved shuffled frog leaping algorithm. Through picking up the heterogeneous cluster and testing the real seismic data, to prove that the heterogeneous cluster management system has sound function, easy operation, effective performance, and reducing the management burden of heterogeneous cluster in heterogeneous cluster seismic data processing aspect.