| With the development of information technology and widely apply, the requirementsof data storage ability in computer systems are also increasing. The problems of storagesystem include: the regional of data storage space become broader; data quantity emergeas explosive growth; data transmission speed cannot meet demand. Distributed storagesystem through connect a large number of nodes together and then collaborate by someway, it can break through the single computer’s CPU, memory, I/O bottleneck resources,and provide higher storage ability. In the distributed file system, when the load of dataservers is not proportionate, the system’s service ability will be reduced. Therefore, theload balancing of system is important for maintaining the system’s performance. Datamigration is an effective base method for system load balancing.Parallel file system pNFS is expanded version of NFS; it has well parallelism,scalability, but its data servers lack of load balancing ability. This restricts the performanceof the system. In this paper, based on parallel file system pNFS (Parallel Network FileSystem), research on data access of load balance problems in Parallel Network FileSystem, focus on how to use data migration to balance the load of the system. This paperpresents a method called ALDM (Adaptive load data migration) to balance the system dataservers’ load. First, different types of resource load on data servers will be classifiedevaluation, and then based on the different types of loads’ conditions, we choose datamigration source data server, target data server and the migration data files dynamically.Through the selected data migration, we can prompt balance the load of the system andimprove the performance of the system.Experiment show that, ALDM data migration technology degrade systemperformance less when data migrating, when migration is complete, about10%systemperformance will be improved, and it has better adaptability with different system loads. |
PDF Full Text Request