Enabling a B+-tree-based Data Management Scheme for Key-value Store over SMR-based SSHD

Owing to the explosive growth of data volume, high areal density storage technologies have been proposed in the past few years. Among them, shingled magnetic recording (SMR) has been regarded as the most promising candidate to replace current conventional hard disk drive based on the perpendicular magnetic recording technology. However, SMR technology not only brings large capacity storage devices but also results in terrible random access performance. For increasing the random access performance of SMR, solid-state hybrid drive (SSHD) seems a possible solution in storage system development. Nevertheless, when an SMR-based SSHD is adopted to a large-scale data management system, a severe performance degeneration will happen because an indexing scheme for access efficiency always maintains data in the large-scale data management system. More specifically, jointly managing indexing keys and data values on an SSHD drive will result in the massive amount of write amplification because of read-merge-write operations and garbage collection processes. Based on such motivations, this work proposed a total solution, namely XsB+-tree, to establish a high-performance B+-tree-based data management scheme for key-value store systems. To the best of our knowledge, this work is the first work to discuss the total solution for the key-value store over an SMR-based SSHD. According to our experimental results, XsB+-tree can improve the access time by 80% on average and prolong the lifetime of SSD up to 19%.