Research Of Process Migration Mechanism Based On Checkpoint In Computational Grid

With the rapid development and wide application of Internet technologies, a new computing pattern—Grid computing, which substantially changes the way we perceive and use computational resources and services is booming. Grid is an Internet-based infrastructure for cooperative resource sharing and problem-solving, which brings the Internet into a new stage. For cooperative resource sharing and problem-solving, many challenging problems anticipate solving in such a wide-area distributed and substantially heterogeneous computing environment as Grid, many challenging problems anticipate to be solved, one of which is how to guarantee the high availability and reliability of Grid in circumstances with frequent resource failure.The research in this thesis focused on computational Grid. We analyzed the particularity of the availability of computational Grid, then studied the key technology of process migration and checkpoint in Grid for the purpose of providing high availability and fault-tolerance with Grid middleware-Globus Toolkit for applications. Main innovations of this thesis include:(1) Through analyzing the engineering of computational Grid and scientific computing, a High Availability Service Layer of Architecture based on Globus Toolkit (HASLA) was proposed which includes process migration, checkpoint data management and resource backup. It takes the high-availability and fault-tolerance into account as well as performance and service quality.(2) Using the technologies of process migration and checkpoint for reference, the thesis proposed a Process Migration Model based on Checkpoint (PMMC), and expounded the course of the model from four aspects: checkpoint setting, process state dump, process state migration and process recovery. This model has provided the high availability and fault-tolerance for running, effectively balanced the node load and increased nodes'using rate and throughput.(3) Based on analyzing classical algorithms of process migration and combined with Checkpoint intercalating strategy, this paper improved a kind of algorithm of process migration based on Checkpoint which was similar to File Server. At last, we tested the performance on the experimental platform. The results showed that this algorithm is effective in reducing the migrated freezing time and eliminating remains dependence, and it provided real time and fault-tolerance for the Grid system.