2020
DOI: 10.1016/j.future.2020.05.005
|View full text |Cite
|
Sign up to set email alerts
|

Adaptive request scheduling for the I/O forwarding layer using reinforcement learning

Abstract: I/O optimization techniques such as request scheduling can improve performance mainly for the access patterns they target, or they depend on the precise tune of parameters. In this paper, we propose an approach to adapt the I/O forwarding layer of HPC systems to the application access patterns by tuning a request scheduler. Our case study is the TWINS scheduling algorithm, where performance improvements depend on the time window parameter, which depends on the current workload. Our approach uses a reinforcemen… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

2
12
0

Year Published

2021
2021
2023
2023

Publication Types

Select...
3
1
1

Relationship

1
4

Authors

Journals

citations
Cited by 7 publications
(14 citation statements)
references
References 25 publications
2
12
0
Order By: Relevance
“…The work presented in [8], is the adaptive method to schedule parallel I/O request for any application on any HPC system by tuning the parameters depending on time window of current workload. The adaptive method is formulated using reinforcement learning of the scheduling algorithm.…”
Section: Related Workmentioning
confidence: 99%
See 1 more Smart Citation
“…The work presented in [8], is the adaptive method to schedule parallel I/O request for any application on any HPC system by tuning the parameters depending on time window of current workload. The adaptive method is formulated using reinforcement learning of the scheduling algorithm.…”
Section: Related Workmentioning
confidence: 99%
“…Previous research has shown that I/O performance prediction based on the different parameters settings can result in significant benefits [7][8][9][10][11]. Despite the key differences between this and existing research with respect to parameters and ML techniques, the prediction of SEG-Y file I/O performance beforehand is itself innovative.…”
Section: Introductionmentioning
confidence: 99%
“…Previous research [13] indicate the feasibility of taking decisions and adapting the system every few seconds. Hence for this paper we chose to work with 1-second long patterns, to represent a real usage while keeping a large number of patterns in our data sets.…”
Section: B Methodology For the Evaluation Of Our Proposalmentioning
confidence: 99%
“…We have recently proposed a reinforcement learning technique to learn the best values for the parameter and adapt it at run-time [13]. We use multiple k-armed bandit instances, one per access pattern.…”
Section: Case Studiesmentioning
confidence: 99%
See 1 more Smart Citation