Reserved Ahead Of Deadline Time Constraints In The Computing Grid Job Scheduling Algorithm Research

Many systems, such as cluster, P2P computation, distributed computing, computer network, have some research about advance reservation, and advance reservation have guaranteed the quality of service (QoS) of jobs. It provides quality of services for terminal users, moreover it can enhance the usage of system resources. It is more advantageous to use resources which contains disk space, and computation node sharingly. The question which appears in some applications must rely on advance reservation to solve.In this paper,it firstly in detail introduces the research condition about advance reservation, including the architecture, the software, the systems, the scheduling strategy and the algorithm present situation, and introduces current advance reservation mechanism impacting on the computation grid system.Next, because of introducting Service Level agreements in the grid, it causes more service demands in the grid computing service. For example, the jobs'deadline limit, this request grid system also has real-time property, completing the job scheduling and execution in the specific time in the future. Therefore according to the grid job's type (independent work and non-independent work) in computation grid it also introduces in detail each kind of scheduling strategies and algorithms related real-time jobs. Through summarizes these strategies and algorithms, finding real-time scheduling strategies and algorithms in the grid is different with the general real-time systems' characteristic, and summarizes the existing grid real-time allocating algorithm deficiency.As for general deadline job in the grid,we bring in the First-Fit algorithm,PE- Best-Fit algorithm, PE-Worst-Fit algorithm,Duration- Best-Fit algorithm,Duration- Worst-Fit algorithm, PE-Duration-Best-Fit algorithm and PE-Duration- Worst-Fit algorithm to solve job scheduling problem; We analyse that different advance reservation rate, different deadlineFactor and different arTimeFactor impact on average usage, average waiting time, average slowdown and successful reservation rate, respectly. At last, we confirm the feasibility about these scheduling strategies through the experiment simulation.Because jobs in the computing grid may need several PE simultaneously, this article proposes these algorithms not only consider the deadline time, but also consider that jobs need several processors, namely simultaneously considers the time and the space to satisfy jobs'scheduling.