The Research And Implementation Of Key Technology In Computing Application Grig Based On Science&Engineering

With the development of science and technology, Internet spreads rapidly around the world. It has become an effective tool for information communication and cooperation. Through which, thousands of computing resources, storage resources, information management resources, together with other digital equipments constitute the major carriers of production, transmittal and the use of knowledge. Meanwhile, Grid, as a new computing infrastructure, has covered all these resources geographically-connected and has implemented the function of resource sharing, cooperation and combined computing. It has provided various of users with all kinds of services in science, engineer, finance and military, etc.The emergence of Computing Grid provides feasibility for sharing of heterogeneous resources distributed in different locations or belonging to various users. It makes means to design a platform based on science and technology computing, integrate and organize kinds of computing and storage resources in heteregeneous environment, through the function of application service access, data access and management service provided by system, to shield the heterogeneity of bottom resource domains, thereby provide users with an intuitionistic file view and canonical usage method. In addition, how to guarantee uaser tasks being finished with efficiency is also a study point. Because of the agility of Grid environment, we must design logical scheduling architecture and adopt specific algorithm to enhance usage efficiency and user satisfaction.Firstly, this paper designs a platform based on science and engineer application. In the whole process, from portal to hardware resources, it realizes seamless connection of each layers, which has strict restriction and nesting. The top layer realies on the bottom layer, the bottom layer serves the top layer. In this way, this platform provides users with a usable service and shields the bottom information, thereby integrally represents the layer concept of the system. In the bottom layer, we adopt the gird middleware Globus to design an extensible heterogeneous system, which can satisfy different application requirements. In the top layer, we adopt grid portal middleware Gridsphere and Gridports to design grid application portal, it realizes the encapsulation of different computing modules, meanwhile, integrates grid accounting and grid monitoring to provide users with a flexible, dynamicly deploying grid application environment.In addition, based on the large-scale science and engineer computing and grid environment, this paper also studys questions like job scheduling, resource allocation and sharing, gives up a job scheduling model and according algorithm. Based on the existing platform deployment and task restriction, we design the DAG scheduling algorithm. Firstly, we rank the tasks in an upwarding order, and then group all these tasks according its rank, in the end, re-schedule tasks to realize optimization. This algorithm realizes two purposes: firstly, minimizing the effection of ranking way to the algorithm performance, secondly, reducing the makespan, finally, this algorithm realizes the aim of multi-users sharing resources with high efficience.