Lcio: large scale filesystem aging

Matthew Bachstein, Feiyi Wang, Sarp Oral

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

1 Scopus citations

Abstract

Performance of file systems shift during their life cycles. Evaluating this performance change over time is not trivial. Complexity arises in the interplay between external (i.e. application I/O workloads) and internal (i.e. the filesystem state) factors. Many benchmarks can test how a filesystem performs at the current snapshot state, but to observe the change over time necessitates that the filesystem state mutate (age) between benchmark runs. For a large-scale HPC parallel filesystem, the sheer scale and amount of interacting components during I/O operations magnify these challenges. There have been several approaches that address different aspects of filesystem aging, from creating statistically realistic filesystem images to file age distributions. The common drawbacks are the scale to be evaluated and the time needed to converge; none of the methods in literature targeted network or parallel file systems. Also, none were evaluated with a filesystem image over 300 GiB, most under 50 GiB, yet almost all took between a half hour to 7 h to converge. For a large-scale parallel file system, these methods are impractical as far as time and resources needed (a typical large PFS is in the PB range). Additionally, HPC filesystem I/O workloads are drastically different from local system workloads used in earlier studies. This paper presents the design, implementation and evaluation of LCIO synthetic filesystem aging benchmark, which aims to address the question of “how will the filesystem perform at different stages of its life cycle?”. As such, being able to answer that question as realistically as feasible in a reasonable time is where LCIO contributes.

Original languageEnglish
Title of host publicationBenchmarking, Measuring, and Optimizing - 2nd BenchCouncil International Symposium, Bench 2019, Revised Selected Papers
EditorsWanling Gao, Jianfeng Zhan, Geoffrey Fox, Xiaoyi Lu, Dan Stanzione
PublisherSpringer
Pages251-261
Number of pages11
ISBN (Print)9783030495558
DOIs
StatePublished - 2020
Event2nd International Symposium on Benchmarking, Measuring, and Optimization, Bench 2019 - Denver, United States
Duration: Nov 14 2019Nov 16 2019

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12093 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference2nd International Symposium on Benchmarking, Measuring, and Optimization, Bench 2019
Country/TerritoryUnited States
CityDenver
Period11/14/1911/16/19

Fingerprint

Dive into the research topics of 'Lcio: large scale filesystem aging'. Together they form a unique fingerprint.

Cite this