The Research And Implementation Of Priority Constrained Job Scheduling Model And Algorithm Based On ECT

With the development of science and technology, the Internet spread rapidly around the world. It has become an effective tool for information communication and cooperation among people. Through which, thousands of computing resources, storage resources, software resources, information resources, as well as other digital equipment together constitute the major carriers for production, transmittal and the use of knowledge. However, Grid just like a new computing infrastructure, it has covered all these resources geographically-connected and has implemented the function of resource sharing, cooperation and combined computing, and it has provided various of users with all kinds of services in science, engineering, finance, military etc.Because the computing resources provided by the Internet are widely distributed geographically and belong to different users, the structure of hardware, the deployment of software and facilities are not quite the same, then when Grid computing is used as solutions for large-scale problems requiring super-computing power in various fields (such as high-energy physics, bioinformatics, chemical molecule simulation and digital weather prediction), we need to allocate these distributed heterogeneous resources with super computing power to users of different applications reasonably to make problems dealt with efficiently, meanwhile we also should ensure the resource utilization remain balance. Therefore, we need a reliable and efficient scheduling algorithm to help to solve the problem for job scheduling and resource allocation. There are different objective functions for scheduling algorithms according to distinct objectives in problem solution, such as resource-centric and application centric. For instance, some grid scheduling management algorithms based on QoS focus on application demanding, and during the scheduling process, the effects of users are taken into consideration in job scheduling and resource allocation. For users, the most important things they concern are the efficiency of job execution and the total cost in resource sharing. Such as jobs must be completed within deadline, or the total charge in resource sharing can't exceed some upper-limit, etc. What's more, there are other constraints for resource demanding, like fault tolerance and security. And now most jobs submitted by users are lodged with different requirements based on actual objectives in Grid resource sharing.This paper presents priority constrained scheduling model based on ECT as well as the corresponding algorithm. The model regards the heterogeneous resources in Grid as the main environment, adopts stochastic Petri networks (SPN) technology, and takes the fault-tolerance grid job scheduling model for reference. And in modeling, we consider the effects of priorities of different users, as well as the coordination of the expected completion time (ECT) and the actual completion time of jobs, to guarantee the jobs of all users to be completed within ECT (Expected Completion Time), especially to ensure the jobs of users with high priorities to be dealt with as soon as possible, making these users get services of good quality, in order to coordinate the conflict of resource sharing between different users.Based on the background of large-scale scientific and engineering computing, we take Grid as basic environment, and focus on the research of Grid scheduling, resource allocation and resource sharing etc. We have given a scheduling model meeting the actual needs and the corresponding scheduling algorithm. The former interprets the relationship between resources, tasks and user as well as some constraints among them in scheduling process; the latter describes the scheduling algorithm in detail. Finally, we compare our algorithm with Mm-min, Max-min and XSuffrage algorithms in performance, in our algorithm, can we ensure the majority of the jobs, in particular, whose users having high priorities, to be completed within ECT, so as to make users generally satisfied, thus achieving the goal of mult-users sharing heterogeneous resources reasonably and without conflict.