The Grid Task Scheduling Strategy Research Based On Ant Immune Memory Optimization Algorithm

Grid task scheduling is the key point of grid technology. Because of characteristics the grid computing resources, such as distributing, heterogeneous, dynamic and instability, in the large-scale distributed environment, the task scheduling is a very complex problem. Task scheduling performance and strategy had a direct impact to the overall performance of grid systems, also related to quality of services to grid system. So, how to improve the task scheduling algorithms and strategies, to increase the efficiency of task scheduling and grid users'multi-Qos requirement is a challenging problem.Firstly, this paper is based on ant colony algorithm as the theory of Grid task scheduling. Through the analysis of the process of immune memory and ant colony task scheduling, establishing a excellent resource information library, introducing immune memory mechanism into the ant colony algorithm. Further proposing PCIMACA task scheduling algorithm based on the Ant immune memory. The algorithm is based on task similarity detection, so that the same or similar tasks in the second optimal operation, the algorithm can directly call the excellent resource information library and avoid to search all grids for each resource when scheduling tasks are traversing, effectively reduce the tasks and resources mapping time.Algorithm design when and how the resources pheromone updates not only use the scheduling success and failure factor, but also designed load balancing reward and punishment factor, the algorithm can converge in some special resource. By GridSim grid simulation comparing the PCIMACA algorithm and ACO, Min-Min algorithm, this paper proved that PCIMACA algorithm effectively reduces the task scheduling response time, the successful of scheduling tasks and resource load has improved. The paper proposed mechanism of biological immune memory combined with the ant colony algorithm for grid task scheduling thoughts, the grid task scheduling algorithm has some innovative. Secondly, detailed analyzing of the existing grid users'requirement. This paper subdivided the users'requirements for resources on the grid into two parts, the specific attributes requirement and fuzzy attributes requirement. Through the dynamic clustering method establish a pre-classified database of resource requirement. After that proposed UDDPTS task scheduling model based on user's requirements. Through Agent-based task scheduling model and PCIMACA task scheduling algorithm, pre-classified the grid users' requirement for resources. At the mean time, establishing the fuzzy dynamic cluster, integral resource library by pre-classification, proposing a UDDPTS task scheduling model based on users'requirement. In this paper, the grid pre-qualification based on the users' requirement has the common meaning to other type of grid scheduling.At last, the system reliability theory was introduced into the grid task scheduling model. Analyzing to single resource and series and parallel links for the task scheduling system reliability model based on the random Petri nets. Structuring GSRSPN task scheduling model in considering the reliability of resources based on random Petri nets. The model uses the TRMA algorithms and tasks reachable graph construction algorithm so that the model taking into account the reliability of each resource node and user QoS requirements while finding the optimal scheduling resources path. Simulation proved the scheduling model, not only met the requirement on time and cost from grid users, but also considered the reliability of the grid resource constraints. The model has an important meaning to the reliability and fault tolerance to grid task scheduling system.In this paper, ant colony algorithm, immune memory mechanism, reliability theory, Similarity algorithm, fuzzy dynamic clustering method, integral method are combined together and proposed an approach ants immune memory PCIMACA task scheduling algorithm, the model based on user demand UDDPTS dynamic task scheduling model and task scheduling GSRSPN reliability model based on he random Petri net. By grid scheduling simulation and verification with example, the above algorithm and the model of task scheduling on the grid had an important and significance meaning to grid task scheduling research.