KVIMR: Key-Value Store Aware Data Management Middleware for Interlaced Magnetic Recording Based Hard Disk Drive

Authors: 

Yuhong Liang, Tsun-Yu Yang, and Ming-Chang Yang, The Chinese University of Hong Kong

Abstract: 

Log-Structured Merge-Tree (LSM-tree) based key-value (KV) store provides write-intensive applications with high throughput on Hard Disk Drive (HDD). Recently, the emerging Interlaced Magnetic Recording (IMR) technology makes the IMR based HDD become another desirable option to construct a cost-effective KV store because of its high areal density. Nevertheless, we observe that deploying LSM-tree based KV store on IMR based HDD may suffer noticeable degradation on throughput of incoming reads/writes. Thus, this paper presents KVIMR, a data management for constructing a cost-effective yet high-throughput LSM-tree based KV store on IMR based HDD. KVIMR is architected as a middleware, interposed between LSM-tree based KV store and IMR based HDD, to embrace the compatibility for mainstream LSM-tree based KV store implementations with limited modifications. Technically, KVIMR adopts a novel Compaction-aware Track Allocation scheme, which leverages the special properties behind the compaction process to remedy the throughput degradation. KVIMR further utilizes a novel Merged RMW approach to improve the efficiency of persisting a multi-track-sized file of KV store into IMR tracks with the ensured crash consistency. Our evaluations on several well-known LSM-tree based KV store implementations reveal that KVIMR not only improves the overall throughput by up to 1.55× under write-intensive workloads but even achieves 2.17× higher throughput under high space usage of HDD, as compared with the state-of-the-art track allocation scheme for IMR.

Open Access Media

USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.

BibTeX
@inproceedings {273933,
author = {Yuhong Liang and Tsun-Yu Yang and Ming-Chang Yang},
title = {{KVIMR}: {Key-Value} Store Aware Data Management Middleware for Interlaced Magnetic Recording Based Hard Disk Drive},
booktitle = {2021 USENIX Annual Technical Conference (USENIX ATC 21)},
year = {2021},
isbn = {978-1-939133-23-6},
pages = {657--671},
url = {https://www.usenix.org/conference/atc21/presentation/liang},
publisher = {USENIX Association},
month = jul,
}

Presentation Video