The Design Of Distributed Compute Environment Base On Grid

Now,in the Automobile industry , only depending on itself ,no factory can design a whole car's mould in time ,so we need put all the resource in the internet into a logic entity,just like a super computer which can provide the integrative information and application services. At last we could construct a High Throughput Computing (HTC)envirnment to distributively accomplish the task. How to construct the kind of envirnment is a challenge,there are some HTCs .First ,Condor is an expert load manage system for compute -Intensive Computing ,which provides a lot of functions ,for exampe,queue mechanism,schedule policy,priorities,resource monitor and manage.When a client submits the job to condor,the condor puts it to the queue,runs the job on the fittest resource depending on some policies,montor the jobs'lifecycle ,at last returns the result. Unfortunately the condor can only work in the intranet,so it is not the grid which we talk about.Second ,the Globus is the well-known grid project ,it provides a tools called Globus Toolkit,which implements many functions ,for example resource manage,network security,information services,data services.Depending on the tools,we can put the separate compute resource and information resource together.But this tools don't support some functions ,for example ,resource schedule ,resource reservation and accouting,and it cannot guarantee the QoS,so it cannot be used in the real envirnment straightly. So,we design a distributed compute envirnment based on the Globus Toolkit infrastructure and advanced a high-Qos-job-first schedule alogrithm based on the economic model, implement some services for example authentication service ,information service,schedule service,job management service,queue service ,reservation service,acconting service.Depending on them,we can schedule the job with accuracy,and use the resource efficently,at last we prove the feasibility of our system depending on some test. First ,how to match the resource is the key of efficient utilization of the resource,so it is very important that we get the information of resource in the system fast and accurately.The information service in our system uses the index service framework based on the webservice and the subscribtion/notification mechanism which can obtain and aggregate all kinds of grid's information efficently in the distributed and isomerous envirnment.So it can provide the cogent reference for matching the resource . We have the eyes and ears ,further more we must have a decision-making part, the resource scheduler is in charge of the job. The resource scheduler uses the High-QoS-job-first schedule algorithm based on the economic model ,first the scheduler analyses the job specification which the client submited ,then sorts the job queue depending on the jobs'QoS requests,and selects the resource whose capacity matches the job's request.If there are many matched resources ,we schedule them by the schedule algorithm based on the economic model,which simulates the economic rule in the real world,and consider that we reserve and hold the resource,we must...