Efficient I/O Merging Scheme for Distributed File Systems

Recently, decentralized file systems are widely used to overcome centralized systems’ load asymmetry between nodes and the scalability problem. Due lack of a metadata server, require more RPC requests control processing clients servers, which adversely impacts I/O performance traffic imbalance by increasing latency. In this paper, we propose an efficient scheme reduce overhead in systems. Instead sending single request at time, enqueued RPCs global queue merged them into larger requests, thus avoiding excessive latency overheads. The experimental results showed that our improves write read up 13% 16%, respectively, compared with those original.