Font Size: a A A

Researches On Some Key Issues In Grid Computing Environments

Posted on:2004-04-25Degree:DoctorType:Dissertation
Country:ChinaCandidate:D ZhaoFull Text:PDF
GTID:1118360095460104Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
The third tidal wave of IT industry is gestating from the beginning of the new century, which tries to interconnect all resources on Internet so that resource sharing and collaborating can be implemented in the scope of virtual organizations. Therefore, it is significant and valuable for both researchers and engineers to investigate grid computing technologies for large-scale business applications to support virtual organizations.Based on Meta Computing, orthodox Grid Computing and Peer-to-Peer Computing, the author explores fargoing grid computing methodologies, which incorporates middleware support at three levels, viz., system model, relevant protocols together with algorithms as well as the infrastructure. Some of the research goals include resource management, resource scheduling, fault tolerance and security in grid computing environment. Innovative contributions of the dissertation are partly enumerated as follows:* Proposing a resource management model by syncretizing Web Computing, Orthodox Grid Computing and P2P Computing. Currently a single resource management model can seldom meet various grid computing requirements, such as dynamicity, scalability and adaptability. A new resource management model, which syncretizes web computing, orthodox grid computing and P2P computing, is proposed to solve the problem. In the model, the whole grid computing environment is organized into a hierarchy with two levels: Different management policies and architectures can be adopted in different administrative domains internally to meet adaptability requirement for legacy systemsï¼›P2P model is used to organize multiple administrative domains to meet dynamicity and scalability. Web services are introduced in the model to provide standard interoperability to overcome resource heterogeneity. Dynamic resource deployment model employs the XML-based data model to describe resource properties and solves the distributed garbage collection problem by Soft State mechanism. Whereas highly flexible and extensible dynamic resource discovery algorithm, which overcomes the limitation of LDAP and relational data model, matches requirements for both resource providers and consumers via XQuery.Proposing a new resource scheduling model and related task graph* scheduling algorithms. Currently most of resource scheduling models lack support for fairness and extensibility. Aiming at the problem, a new resource repository model is proposed in the dissertation. The scheduling is divided into user-level and resource-level to adapt to the resource management model. It also embodies fairness since it enables service level agreement negotiation between resource providers and resource consumers. According to the homogeneous and heterogeneous computing environments, different static task graph scheduling algorithms are designed respectively. Dynamic task graph scheduling algorithms are also explored for these environments. In the homogeneous environment, DCP algorithm is reformed to DCPTM by introducing dynamic task migration idea. Both theory analysis and simulation experiments prove that the new algorithm can generate scheduling results with better performance. In the heterogeneous environment, task graphs are classified according to their communication/computation ratio so that different scheduling algorithms could be applied adaptably for task graphs with different classes. Compared with related works, the adaptable algorithms presented in the dissertation are more full-scale in different scheduled application type, different resource scheduling time and different resource scheduling space.* Proposing an adaptable efficient replication protocol. The 7X24 non-stop running of critical services is necessary for grid computing technologies to be applied in business field on a large scale. Therefore these services must be highly available; nevertheless, current researches do not care about it enough. An efficient replication protocol is proposed in the dissertation to support different distributed applications. Its design adopts...
Keywords/Search Tags:grid computing, middleware, dynamic resource deployment, dynamic resource discovery, task graph scheduling, replication, access control, security model
PDF Full Text Request
Related items