Which can Accelerate Distributed Machine Learning Faster: Hybrid Optical/Electrical or Optical Reconfigurable DCN?
We run various distributed machine learning (DML) architectures in a hybrid optical/electrical DCN and an optical DCN based on Hyper-FleX-LION. Experimental results show that Hyper-FleX-LION gains faster DML acceleration and improves acceleration ratio by up to 22.3%.
Published in: | Optical Fiber Communication Conference (OFC) 2022 |
---|---|
Main Authors: | , , , |
Other Authors: | , , , |
Format: | Conference Object |
Language: | English |
Published: |
Institute of Electrical and Electronics Engineers Inc.
2022
|
Subjects: | |
Online Access: | https://hdl.handle.net/11583/2973049 https://doi.org/10.1364/OFC.2022.Th1G.5 |
id |
ftpoltorinoiris:oai:iris.polito.it:11583/2973049 |
---|---|
record_format |
openpolar |
spelling |
ftpoltorinoiris:oai:iris.polito.it:11583/2973049 2024-04-14T08:10:54+00:00 Which can Accelerate Distributed Machine Learning Faster: Hybrid Optical/Electrical or Optical Reconfigurable DCN? Yang H. Zhu Z. Proietti R. Ben Yoo S. J. Yang, H. Zhu, Z. Proietti, R. Ben Yoo, S. J. 2022 ELETTRONICO https://hdl.handle.net/11583/2973049 https://doi.org/10.1364/OFC.2022.Th1G.5 eng eng Institute of Electrical and Electronics Engineers Inc. info:eu-repo/semantics/altIdentifier/isbn/978-1-55752-466-9 info:eu-repo/semantics/altIdentifier/wos/WOS:000828152500432 ispartofbook:2022 Optical Fiber Communications Conference and Exhibition, OFC 2022 - Proceedings 2022 Optical Fiber Communications Conference and Exhibition, OFC 2022 numberofpages:3 https://hdl.handle.net/11583/2973049 doi:10.1364/OFC.2022.Th1G.5 info:eu-repo/semantics/altIdentifier/scopus/2-s2.0-85128913656 https://doi.org/10.1364/OFC.2022.Th1G.5 info:eu-repo/semantics/openAccess Distributed machine learning optical switching datacenter network info:eu-repo/semantics/conferenceObject 2022 ftpoltorinoiris https://doi.org/10.1364/OFC.2022.Th1G.5 2024-03-21T16:08:55Z We run various distributed machine learning (DML) architectures in a hybrid optical/electrical DCN and an optical DCN based on Hyper-FleX-LION. Experimental results show that Hyper-FleX-LION gains faster DML acceleration and improves acceleration ratio by up to 22.3%. Conference Object DML PORTO@iris (Publications Open Repository TOrino - Politecnico di Torino) Optical Fiber Communication Conference (OFC) 2022 Th1G.5 |
institution |
Open Polar |
collection |
PORTO@iris (Publications Open Repository TOrino - Politecnico di Torino) |
op_collection_id |
ftpoltorinoiris |
language |
English |
topic |
Distributed machine learning optical switching datacenter network |
spellingShingle |
Distributed machine learning optical switching datacenter network Yang H. Zhu Z. Proietti R. Ben Yoo S. J. Which can Accelerate Distributed Machine Learning Faster: Hybrid Optical/Electrical or Optical Reconfigurable DCN? |
topic_facet |
Distributed machine learning optical switching datacenter network |
description |
We run various distributed machine learning (DML) architectures in a hybrid optical/electrical DCN and an optical DCN based on Hyper-FleX-LION. Experimental results show that Hyper-FleX-LION gains faster DML acceleration and improves acceleration ratio by up to 22.3%. |
author2 |
Yang, H. Zhu, Z. Proietti, R. Ben Yoo, S. J. |
format |
Conference Object |
author |
Yang H. Zhu Z. Proietti R. Ben Yoo S. J. |
author_facet |
Yang H. Zhu Z. Proietti R. Ben Yoo S. J. |
author_sort |
Yang H. |
title |
Which can Accelerate Distributed Machine Learning Faster: Hybrid Optical/Electrical or Optical Reconfigurable DCN? |
title_short |
Which can Accelerate Distributed Machine Learning Faster: Hybrid Optical/Electrical or Optical Reconfigurable DCN? |
title_full |
Which can Accelerate Distributed Machine Learning Faster: Hybrid Optical/Electrical or Optical Reconfigurable DCN? |
title_fullStr |
Which can Accelerate Distributed Machine Learning Faster: Hybrid Optical/Electrical or Optical Reconfigurable DCN? |
title_full_unstemmed |
Which can Accelerate Distributed Machine Learning Faster: Hybrid Optical/Electrical or Optical Reconfigurable DCN? |
title_sort |
which can accelerate distributed machine learning faster: hybrid optical/electrical or optical reconfigurable dcn? |
publisher |
Institute of Electrical and Electronics Engineers Inc. |
publishDate |
2022 |
url |
https://hdl.handle.net/11583/2973049 https://doi.org/10.1364/OFC.2022.Th1G.5 |
genre |
DML |
genre_facet |
DML |
op_relation |
info:eu-repo/semantics/altIdentifier/isbn/978-1-55752-466-9 info:eu-repo/semantics/altIdentifier/wos/WOS:000828152500432 ispartofbook:2022 Optical Fiber Communications Conference and Exhibition, OFC 2022 - Proceedings 2022 Optical Fiber Communications Conference and Exhibition, OFC 2022 numberofpages:3 https://hdl.handle.net/11583/2973049 doi:10.1364/OFC.2022.Th1G.5 info:eu-repo/semantics/altIdentifier/scopus/2-s2.0-85128913656 https://doi.org/10.1364/OFC.2022.Th1G.5 |
op_rights |
info:eu-repo/semantics/openAccess |
op_doi |
https://doi.org/10.1364/OFC.2022.Th1G.5 |
container_title |
Optical Fiber Communication Conference (OFC) 2022 |
container_start_page |
Th1G.5 |
_version_ |
1796308564820623360 |