Laser: Load-adaptive group commit in lock-free transaction logging

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Log manager is a key component of DBMS and is considered as the most prominent bottleneck in the modern in-memory OLTP system. In this paper, by addressing two existing performance hurdles in the current procedure, we propose a high-performance transaction logging engine Laser and integrate it into OceanBase, an in-memory OLTP system. First, we present a lock-free transaction logging framework to eliminate the lock contention. Then we make theoretical analysis and propose a judicious grouping strategy to determine an optimized group time for different workloads. Experiment results show that it improves 1.4X–2.4X throughput and reduces more than 60% latency compared with current methods.

Original languageEnglish
Title of host publicationWeb and Big Data - 1st International Joint Conference, APWeb-WAIM 2017, Proceedings
EditorsCyrus Shahabi, Xiang Lian, Christian S. Jensen, Xiaochun Yang, Lei Chen
PublisherSpringer Verlag
Pages320-328
Number of pages9
ISBN (Print)9783319635781
DOIs
StatePublished - 2017
Event1st Asia-Pacific Web and Web-Age Information Management Joint Conference on Web and Big Data, APWeb-WAIM 2017 - Beijing, China
Duration: 7 Jul 20179 Jul 2017

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10366 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference1st Asia-Pacific Web and Web-Age Information Management Joint Conference on Web and Big Data, APWeb-WAIM 2017
Country/TerritoryChina
CityBeijing
Period7/07/179/07/17

Keywords

  • Group commit
  • Load-adaptive
  • Lock-free
  • Logging

Fingerprint

Dive into the research topics of 'Laser: Load-adaptive group commit in lock-free transaction logging'. Together they form a unique fingerprint.

Cite this