Multi-dimensional Qos Guided Task Scheduling In Grid Environment Research

Recently, Grid has been widely paying attention, which is considered as a hot topic in the research of Computer Science. The task scheduling is a fundamental issue in achieving high performance in Grid computing systems and plays an important role. There is a great significance in scheduling tasks cross different processors effectively and efficiently within the whole Grid system, both for theoretical and practical application. It has been proved that the Grid task scheduling is a NP-complete problem, and it is hard to find an optimal solution in polynomial by using traditional searching method. So the modern heuristics algorithm is introduced to approach the optimal solution. In this gradual process, quality of service (QoS) should not be ignored.Grid scheduling system is responsible for resource coordinated and task scheduling. In Grid environment, the efficiency of task scheduling is the key factor that affects the efficiency of Grid system. During these years, more and more attentions have been paid to QoS (Quality of Service) in Grid environment, and the ability to deliver nontrivial QoS became one of the key characters of Grid system. Usually, task scheduling means trying to satisfy user's QoS requirement and trying to optimize system performance, such as system throughput and resource utilization. Obviously, supporting QoS in Grid task scheduling is the key issue affects the Grid system deliveries nontrivial QoS.Focusing on the problem of supporting multiple QoS in grid task scheduling, this thesis carries out researches as follows:Firstly, this thesis introduces the background of research, includes the concept of Grid, the concept of Grid resource management, and the role task scheduling acts in Grid resource management.Secondly, this thesis summarizes the mechanism of task scheduling in Grid environment, which includes the summarizes for the characteristics and goals of Grid task scheduling, the analysis for Grid task scheduling mechanism according to the research up to date, the introduction of grid scheduling systems, and the emphasis on the problem of supporting QoS in Grid task scheduling.Moreover, in order to solve multiple QoS requirement, a novel mathematical model for multidimensional QoS and a heuristic algorithm based on this model is proposed. The heuristic for batch scheduling enables the consideration for user's QoS requirements while aiming to optimize system performance. The simulation results from Gridsim demonstrate that, for the multi-QoS supporting batch scheduling heuristic, compared to traditional Min-Min heuristic, its makespan has almost no decrease while its user satisfaction improves 17% in average.At last, the algorithm is evaluated within a simulated Gridsim environment. The experimental results show that the new Multi-QoS guided Min-Min heuristic can lead to significant performance gain for a variety of applications.