| With the rapid development of Internet technology, people pay more and more attention to grid computing. Resource sharing is a core issue that is necessary for grid computing to resolve, and is also the main purpose of grid computing. In a manner of speaking, if there is no resource sharing, there would be no grid computing. Because of the distributed, dynamic and heterogeneous characteristics of the grid resource, the management of various grid resources turns into the focal point and difficulty of the grid research. The grid resource discovery that serves as the basic grid service, is also the basis of grid resource management, which searchs for resources that meet the application requirements for those requesters.Traditional gird resource discovery mostly uses the directory-based centralized resource discovery mechanism. This mechanism can be simply implemented and managed, but, as the grid scale continues to expand, the centralized directory will become a bottleneck, and the single point of failure is prone. The poor scalability should also be considered. At present, grid resource discovery mechanism can mainly make accurate query based on the static information, which can not well adapt to the dynamic properties of grid resource and can not implement the range query, multi-attribute query.P2P technology which comes up to implement the resource sharing with the same purpose as grid, is based on the distributed peer-to-peer network technology, and it solves the drawbacks of the traditional centralized control in C/S structure. With the rise of P2P networks, the new directions and new concepts raised in P2P provide new ideas in grid research. And one attempt is to combine P2P related concepts and research with grid, especially in the network topology and resource search. In recent years, there has been a lot of research about combining P2P resource management with Grid, which has laid a foundation for the new direction of grid research.In this thesis, the feasibility of combining grid with P2P has been determined, by analysing the similarities and differences in grid and P2P resource discovery, and researching the related work in recent years. Because of the remaining differences in the resource sharing mechanism between grid and P2P, the combination is not a simple superposition, and it is necessary to modify the P2P mechanism, to design a new mechanism for resource discovery and a new resource representation method. In this thesis, a hierarchical grid resource discovery mechanism is put forward to solve the lack of the traditional grid resource discovery, and a new XML-based resource representation method is also introduced to support dynamic query and multi-attribute query.The hierarchical grid resource discovery mechanism logically divide the grid resources into two layers. The bottom is called the grid layer, which is composed of a number of virtual organizations (VO) formed by grid nodes. A virtual organization is a complete grid system, including the management node and resource node, and the WS-MDS Index Service of GT4 is deployed in each virtual organization, which is responsible for the resource discovery inside the organization. The introduction of virtual organizations makes it more convenient to manage the grid resources, makes the resource query inside and outside virtual organizations logically independent of each other, and enhances the system scalability. The upper layer is called the P2P layer, which is formed by a number of DHT nodes. Each DHT node takes charge of a virtual organization, and all DHT nodes form a Chord ring in which nodes can join, leave and search is implemented by the use of P2P Chord protocol. The P2P layer forms a DHT based P2P system.The nodes inside the virtual organization are also organized by hierarchical mode, and a virtual organization deploys one or more index services in which the grid services are registered. The nodes in a virtual organization are divided into a number of small sites, and each site can deploy their own index service to retrieve resources inside the site. A number of small sites can also be organized into a large site. The virtual organization has a general index service, which is used to register the index services of each site.DHT node is a node of the P2P layer, which is also the gathering node of a virtual organization, and takes charge of the resource discovery inside the virtual organization. Each virtual organization has a DHT node. All DHT nodes in P2P layer that form a DHT ring, are organized by Chord protocol, and each DHT node has saved information about a number of other DHT nodes. When receiving query messages, DHT node forwards the query message with the stored information about other DHT nodes.Resources are represented by the XML-based format, including two types of XML documents, such as DHTNode and ResourceNode. ResourceNode represents the XML document on the resource node, which keeps the information of the services provided by the node, for example, ServiceName that is the name of the Service, ServiceDescription that is the description, DHTNodeIP that is the DHTnode IP address of the virtual organization which the nodes belong to, EndPointReference that is the reference address of the service provided by the node, MachineMessageEPR that is the reference address of a service MachineMessage provided by the node, which is used to provide the information of CPU quantity, CPU frequency , memory size, amount of free memory, disk space and other information available about the host where the node locates.According to the resource routing in P2P layer, the algorithms for maintaining the hierarchical structure about node join and node leave are introduced, in particular, the maintaining method by DHT node of the XML document used to preserve information about resource nodes has been put forward. Then the resource discovery method for the hierarchical system has been raised, which is divided into the single-attribute resource discovery, and the multi-attribute resource discovery. The single-attribute resource discovery is similar to the traditional ways, and the multi-attribute resource discovery is a improvement to the none supporting for complex queries in the past, which takes use of MachineMessageEPR service provided by resource nodes to do further match on the machine information.At last, an improvement to the Chord routing table is raised, due to a large number of redundant information existing in the routing table and the inefficiency for searching, which is discovered in the use of the search algorithm in Chord. It distributes the items of the routing table evenly to the Chord identifier space, which reduces the redundancy of Chord routing table, and improves the efficiency of the query. This has also been tested in a simulation environment.In this thesis, the introduction of the hierarchical resource discovery mechanism and related algorithms, reforms the organization of the grid resources, improves the quality of the resource discovery to a certain extent and supports the multi-attribute queries. The improvement of Chord routing table enhances query efficiency and reduces the query hops. The related work that has been done in this article makes a foundation for further study about resource monitoring and subsequent resource allocation. |
PDF Full Text Request