Evaluation of Data Optimal Storage System in Hybrid Cloud

Abstract

A hybrid cloud is a mix of compute systems, storage systems and services consisting of on-premises systems, private clouds and public clouds for satisfying requirements such as security and performance. In the case of hybrid cloud configuration where data are in on-premises storage systems and compute systems is in public clouds, data should be transferred to the public cloud when a read/write command is issued by compute systems. Thus, long data transfer time is a problem since the on-premises systems and public clouds are connected by the internet. One possible solution is to create an additional cache area inside the compute system; however, it is not clear how effective it is. In this paper, we verify the effect of the data cache by creating the hybrid cloud environment with/without the cache area and measuring input/output latency. We created the measurement environment in the hybrid cloud with dm-cache technology and set up the assumed workload of TPC Benchmark TM E with the write-through mode. The average latency with the cache area reduces by up to 79% compared to without the cache area, and we confirm the effect of installing cache areas in this environment.