Font Size: a A A

Platform EGO WSG Client Resource Broker Model

Posted on:2008-01-31Degree:MasterType:Thesis
Country:ChinaCandidate:W K SunFull Text:PDF
GTID:2178360242460218Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
The market for traditional grid computing faces an impending disruption as enterprises look to move to virtual environments and fully leverage their hardware and software assets. When making this transition, enterprises need solutions with broad compatibility and the intelligence to manage and allocate resources where and when they're needed. In striking the balance between service oriented infrastructure(SOI) and service oriented architectures(SOA) you need a layer in between to ensure diverse IT resources come together seamlessly-Platform EGO has the opportunity to become this layer。Platform EGO is a SOA based grid platform. Unlike traditional cluster computing environments, Platform EGO is built on open standards and architectures(Web services, SOA and virtualization) to allow for extensibility and flexibility to manage the shared resources across geographically dispersed sites for diverse enterprise applications, services and workloads.Platform EGO Web Service Gateway, is a grid middleware, enable applications, called web service clients(WSC), to access Platform EGO services as web services, has high reliability, performance and interoperability.As the client of Platform EGO WSG, WSC is a web application correspond to the WSG, uses the WS-Security UsernameToken enable clients access WSG and get service. Achieve a notification mechanism that WSC can receive resource from notification port. User can use the interface that provide by WSC access WSG, ask EGO for resources. When users receive enough resources for its tasks, it submits its job.The model that EGO allocates resources is to satisfy users try its best. EGO has how many resources, it will allocate all for users. So, when user needs more resources and EGO can not allocate all for user because of the workload. EGO will be able to allocate resources to users and return these resources to the user first. If user can not execute its job before it can get all resources that he needs, user must be wait.Because single user know the workload not so clearly, it can't judge the current resources ability that provide by EGO accurately. Therefore, EGO's workload is higher and the ability to provide resources is very limited when single user request resources. User must wait for a long time to gets enough resources and wastes much money in the resource waiting.This paper recording to the need of user's control budget and job's deadline, design a resource broker provides these services for user.The resource broker is a WSC that working under EGO WSG for a long time. And because this WSC requests resources continually, it has rich experiences about the EGO's ability to provide resources. Through the log storage, resource broker realize the accumulation of experience. Using time series model predict the EGO's ability in the future time becomes possible.The resource broker can confirm how many resource should be applied by analyze all task waiting for dispose, and the ability of resource broker to forecast how many resource he can really got in a exactitude time. When the broker confirm the number of resource, he can use this as the capability of resource to scheme when those task will be run. but this problem is NP-complete, to find a way to solve it by a solve time is the most important problem the broker must solve.Actual method can be compartmentalization to two general kind: the optimization method and the approximately method. the optimization method can find the best result, but usually it can not made it in a reason time when the problem is more big. the approximately method can equipoise the time to solve the problem and the result quality, made a satisfactory result in a reason time.By seeing about those method and think about the problems of the broker faced, we design a mix method to support scheme those task in a way of parallel and serial. Use the capability of resource to made those task which can schemed at the same time a combination to be schemed. And we design a effective heuristic function to make our method more efficiency, so the broker can scheme a great deal of tasks in a reason time.The broker profit by offer service to client. Client use broker to make sure his task can accomplish befor the deadline, and control the budget of his task. Also, because the broker made a great deal of smaller resource application to a big one, can reduce the WSG's burden, and made EGO more efficiency. So, client, the broker, WSG and EGO can all benefit from the broker's work.
Keywords/Search Tags:Platform
PDF Full Text Request
Related items