Freshness-based Consistency and its Implementation for High-Available Key-Value Store
-
- Hiroshi Horii
- IBM Japan
-
- Tamiya Onodera
- IBM Japan
Bibliographic Information
- Other Title
-
- 高可用キー・バリュー ストアにおけるデータ鮮度に基づいた一貫性の提案とその実装
Search this article
Description
Key-value stores that emphasize their availability and partition-tolerance sacrifice consistency, as the CAP theorem states. In existing key-value stores, servers detect inconsistencies with versions of values and clients reconcile the inconsistencies while reading values with versions in multiple servers. However, such solutions reduce the throughput caused by duplicated processing for each read in servers. We have devised a new key-value store that provides freshness of value for how up-to-date the value is. Servers recognize freshness of values by exchanging the latest version numbers of values, and clients get the freshness for every read. Whenever the freshness is enough for clients, the clients read a value of only one server. In our evaluation, our key-value store showed up to 2.0 times higher throughputs for reads and competitive throughputs for writes with existing key-value stores.
Journal
-
- IPSJ SIG Notes
-
IPSJ SIG Notes 2013 (20), 1-6, 2013-07-15
Information Processing Society of Japan (IPSJ)
- Tweet
Keywords
Details 詳細情報について
-
- CRID
- 1573387452695819776
-
- NII Article ID
- 110009585828
-
- NII Book ID
- AN10112482
-
- Text Lang
- ja
-
- Data Source
-
- CiNii Articles