2015 | OriginalPaper | Buchkapitel
SASS: A High-Performance Key-Value Store Design for Massive Hybrid Storage
verfasst von : Jiangtao Wang, Zhiliang Guo, Xiaofeng Meng
Erschienen in: Database Systems for Advanced Applications
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Key-value(KV) store is widely used in data-intensive applications due to its excellent scalability. It supports tremendous working data set and frequent data modifications. In this paper, we present SSD-assisted storage system (SASS), a novel high-throughput KV store design using massive hybrid storage. SASS meets three exclusive requirements of enterprise-class data management: supporting billions of key-value pairs, processing thousands of key-value pairs per second, and taking advantage of the distinct characteristics of flash memory as much as possible. To make full use of the high IOPS of sequential write on the SSD, all modification operations are packaged as operation logs and appended into SSD in the time order. To handle the tremendous number of key-value pairs on hard disk, a novel sparse index, which can be always kept in the SSD, is proposed. Moreover, we also propose an in-memory dense index for the operation logs on SSD. Our evaluation mainly characterizes the throughput of read and write, namely the ops/sec(
get-set
operations per second). Experiments show that our SASS design enjoys up to 96806 write ops/sec and 3072 read ops/sec over 2 billion key-value pairs.