An optimization algorithm for I/O performance of cloud storage system based on dynamic mixed partitioning mechanism

An optimization algorithm for I/O performance of cloud storage system based on dynamic mixed partitioning mechanism

Ma Yi Chun, Ran, Feng Keng, Shen Ye

To solve the problem of storage I/O performance optimization in the cloud storage system, based on the data distribution, a storage fragment performance model is established, and a piecewise decision algorithm based on the optimal I/O performance and a calculation method of the optimal slice number are proposed, and a data distribution algorithm ( Dadh). Dhah can fully consider the actual performance of each storage node and client, dynamically make the file slicing decision and the optimal slice number calculation, and can optimize the data distribution according to the actual performance of the storage node. The experimental results show that the dadh has the best I/O performance compared with the traditional data distribution algorithm, and the algorithm performance improvement effect can reach 18% ~44% on average in all kinds of bandwidth.

Keywords: cloud storage; Data distribution; Dynamic mixed fragmentation; I/O performance

An optimization algorithm for I/O performance of cloud storage system based on dynamic mixed partitioning mechanism