Research on resource allocation based on global optimization in distributed environment

Research on resource allocation based on global optimization in distributed environment

Dan of Dalian University of Technology

With the rapid development of technology and informatization, computer system has begun to change from stand-alone systems to distributed systems. How to integrate different resources to provide higher computing and service capabilities in a distributed environment has become a focus of attention. In this paper, a grid platform based on resource agent is designed and implemented to study the resource allocation problem in two typical distributed environments. When users are applied in a distributed environment, they are often divided into independent or independent tasks, and there is a sequence of execution and data transfer between independent tasks. This paper establishes the problem description model for independent task and non independent Task resource allocation problem, and makes a game analysis to find Nash equilibrium state and prove its existence, and analyze the influence of game between tasks on the performance of the whole system. In the independent task resource allocation model, the collective interest can be represented by the total cost of all tasks, which is an important index of the performance of the reaction system. This paper puts forward a resource allocation algorithm based on the maximization of the collective benefit by changing the payment function of the participant in the game (RACGR) in the resource allocation model of the independent task, the relationship between tasks can be represented by the Dag graph and the collective interest is represented as the scheduling length, this paper presents a resource allocation algorithm based , the task is sorted to reduce the effect of static game between tasks on the performance of the system, which shortens the scheduling length and improves the system performance. In this paper, two resource allocation algorithms are compared with polling algorithm, max-min algorithm and min-min algorithm according to different performance indexes of independent and independent tasks. RACGR algorithm has a good performance in total cost and resource fairness, and analyzes the effect of resource expenditure function on the task execution quantity. The TSBS algorithm can effectively shorten the scheduling length and improve the utilization rate of the DAG graph and the Dag graph of the actual application.

Research on resource allocation based on global optimization in distributed environment