Research On The Decision Tree-like Hierarchical Grid Resource Organization Model And Scheduling Algorithms

As a distributed heterogeneous computing platform, the gird aims at achievingInternet-wide resource sharing and collaborative computing. Grid resource organizationmodel and scheduling algorithm is the core of the grid technology and the essential part ofthe grid resource management. In this paper, the research status of the grid resourceorganization model and scheduling algorithm is analyzed. According to the grid resourceorganization model and the scheduling algorithm in poor performance, they are analyzedand researched in this paper.Firstly, In view of the poor extensibility of the traditional centralized resourceorganization model and the low efficiency of the traditional distributed resourceorganization model, combined with the grid resource characteristics in distribution,dynamics and heterogeneity, the decision tree-like hierarchical grid resource organizationmodel is put forward. The model is logically divided into two layers. The lower layerconsists of the physical resources nodes, and the upper is composed of routing nodeswhich are formed in decision tree-like grid resource organization model.Secondly, Because of the grid resources characteristics of dynamic joining and exit, amaintenance algorithm on the decision tree-like hierarchical grid resource organizationmodel is constructed. When resources arrive at the grid system, the ID5R algorithm isadopted to construct the model, when resources depart from the grid system, DID3algorithm is adopted to construct the model. The grid resource discovery process isdescribed based on the decision tree-like hierarchical grid resource organization model.Thirdly, Combined with the results of the grid resource discovery, the Min-Min gridresources scheduling algorithm is optimized. The Min-Min scheduling algorithm is simpleand easy to implement. It is worth to mention that Min-Min scheduling algorithm has beenefficiently adopted in some practice projects. However, the Min-Min scheduling algorithmalways preferentially scheduling short jobs on the high computing power resources willlead to load imbalance problem. According to the Min-Min scheduling algorithm inuneven load and low resources utilization rate, the OPT-Min-Min scheduling algorithm isput forward. Finally, Experimental results of the OPT-Min-Min scheduling algorithm obtained forthe benchmark of instances by Braun et al. The high efficiency of OPT-Min-Minscheduling algorithm is verified through comparison testing of the Min-Min schedulingalgorithm, the Max-Min scheduling algorithm and Min-mean scheduling algorithm.