Rhythm: Harnessing data parallel hardware for server workloads

dc.contributor.author

Agrawal, SR

dc.contributor.author

Pistol, V

dc.contributor.author

Pang, J

dc.contributor.author

Tran, J

dc.contributor.author

Tarjan, D

dc.contributor.author

Lebeck, AR

dc.date.accessioned

2015-12-09T18:58:02Z

dc.date.issued

2014-03-14

dc.description.abstract

Trends in increasing web traffic demand an increase in server throughput while preserving energy efficiency and total cost of ownership. Present work in optimizing data center efficiency primarily focuses on the data center as a whole, using off-the-shelf hardware for individual servers. Server capacity is typically increased by adding more machines, which is cheap, though inefficient in the long run in terms of energy and area. Our work builds on the observation that server workload execution patterns are not completely unique across multiple requests. We present a framework - called Rhythm - for high throughput servers that can exploit similarity across requests to improve server performance and power/energy efficiency by launching data parallel executions for request cohorts. An implementation of the SPECWeb Banking workload using Rhythm on NVIDIA GPUs provides a basis for evaluating both software and hardware for future cohort-based servers. Our evaluation of Rhythm on future server platforms shows that it achieves 4× the throughput (reqs/sec) of a core i7 at efficiencies (reqs/Joule) comparable to a dual core ARM Cortex A9. A Rhythm implementation that generates transposed responses achieves 8× the i7 throughput while processing 2.5× more requests/Joule compared to the A9. Copyright © 2014 ACM.

dc.identifier.isbn

9781450323055

dc.identifier.uri

https://hdl.handle.net/10161/11137

dc.publisher

ACM

dc.relation.ispartof

International Conference on Architectural Support for Programming Languages and Operating Systems - ASPLOS

dc.relation.isversionof

10.1145/2541940.2541956

dc.title

Rhythm: Harnessing data parallel hardware for server workloads

dc.type

Conference

duke.contributor.orcid

Lebeck, AR|0000-0003-1893-5464

pubs.begin-page

19

pubs.end-page

34

pubs.organisational-group

Computer Science

pubs.organisational-group

Duke

pubs.organisational-group

Electrical and Computer Engineering

pubs.organisational-group

Pratt School of Engineering

pubs.organisational-group

Trinity College of Arts & Sciences

pubs.publication-status

Published

Files

Original bundle

Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
rhythmasplos.pdf
Size:
3.46 MB
Format:
Unknown data format
Description:
Published version