A Highly Reliable Storage Systems Based on SSD Array for IoE Environment

摘要

Devices in IoE (Internet of Everything) environment generate massive data from various sensors. To store and process the rapidly incoming large-scale data, SSDs are used for improving performance and reliability of storage systems. However, they have typical problem called write amplification which is caused by out-of-place updates characteristics. As the write amplification increases, it degrades I/O performance and shortens SSDs' lifetime. This paper presents a new approach to reduce write amplification of SSD arrays. To solve the problem, this paper proposes a new parity update scheme, called LPUS. LPUS transforms random parity updates to sequential writes with additional log blocks in SSD arrays by using parity logs and lazy parity updates. The experimental results show that, LPUS reduces write amplification up to 37% and the number of erases up to 50% with the reasonable size of log space.