The Implementation Of Grid File Copying Strategy Based On Distributed File Sharing

So-called grid is to use the Internet to connect different distribution of various resources(including computation, resources, storage, resources, databases, resource, information resources and bandwidth resources, hardware and software resources) into a virtual logical entry, let users feel like using a supercomputer, which offers a transparent and efficient computational power or information and application services with integration properties. While globus is one of the very influential projects which has close relation with grid computing. Globus is a middleware in grid computing, which provides a application platform for developing grid application. So the establishing of Globus platform is the prerequisites for implementing and researching grid applications.The copies are meaning that there are multiple copies to the same file, and which is located in different places. Copies are also called replica. RLS is Replica Location Service, which is also an important component of RLS in grid environment. RLS is a tool, it provides a ability to track one or more Replica and copies and file. The tool is included in globus. It is very important for users or service to query or position the file in grid.The significance of grid is very important. Initially, the reason of generating grid is to achieve a wide range of high-performance scientific computing and to solve large-scale complex scientific problems. Along with the development of network technology, various complicated application also arises at the historic moment, such as:high nuclear physics, weather forecast, the car crash test, physical simulation experiment, global scientific research objects. All these need to access to large amount of data, but the data are huge and dispersed, so to achieve high performance access frequently and rapid transmit data will become more difficult. The most effective way to solve these difficulties is to implement copy strategy, which is to copy some data files to each node of the grid. While the data grid just provide a shortcuts to relax retrieve data resource. Because the creating of data copies make copies scatter in the grid everywhere widely and reasonability, which improves the network load balance, reduces the bandwidth consumption and network delay brought by remote access data. Meanwhile the reliability and security of data and system fault tolerance has been improved. Therefore how to use the existing copies of cloning technology to be improved to achieve distributed file sharing grid file copying strategy is this thesis's mainly research topic.The thesis mainly realizes the distributed file sharing grid file copying strategy, which solves the problem of creating or deleting or updating or positioning or choosing data copies which uses globus as the platform. By adopting samba share files system as grid data transmission tool replacing the globus GridFTP components selectively. So all the data needed to be visited are stored in its shared storage area. We only need to mount the shared storage area of the node which has the accessed data to their system's specified directory to realize reading and using the data. It utilizes the TC tools to simulate the Internet's. environment, and through calculation network delay to choose copies. At the same time through monitoring the documents to solve the problem of deleting and modifying copies effectively. By implementing client and server program solving the problem of SQLite which cannot be accessed remotely and of creating replication and of computing file access heat. Also through calculating the hash value to the replication to validate whether the documents has changed while transmitting. Finally, through submitting a lot of homework examples to the grid to simulate the process of distributed file access. And realize the waterfall strategy and rapid diffusion strategies and best client strategy and the strategy of willing not create a copy respectively to test them and compare the average time of execution.