Self-adaptive metadata management model for distributed file system
Cheng Fu-cha · Jisuanji gongcheng yu sheji · 2014
To reduce the cost of metadata management in distributed file systems(DFS)and resolve the problem of single point of failure(SPoF)of metadata node,a distributed memory abstraction called working-backup dataset(WBD)is presented.WBD abstracts the memory areas of nodes into datasets,and accesses the memory with dataset operations,to reduce the complexity of memory operation.Based on WBD,a metadata management model called self-adaptation master-slave(SAMS)model is designed to reduce the performance cost of metadata synchronization.Experimental analysis indicates that adopting SAMS as metadata management model of DFS has achieved high fault-tolerance and high scalability,on the basis of maintaining metadata service performance.