Deep Learning Training in Facebook Data Centers: Design of Scale-up and Scale-out Systems
Large-scale training is important to ensure high performance and accuracy of machine-learning models. At Facebook we use many different models, including computer vision, video and language models. However, in this paper we focus on the deep learning recommendation models (DLRMs), which are responsi...
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , , , , , |
|---|---|
| Format: | Artikel |
| Sprache: | eng |
| Schlagworte: | |
| Online-Zugang: | Volltext bestellen |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| container_end_page | |
|---|---|
| container_issue | |
| container_start_page | |
| container_title | |
| container_volume | |
| creator | Naumov, Maxim Kim, John Mudigere, Dheevatsa Sridharan, Srinivas Wang, Xiaodong Zhao, Whitney Yilmaz, Serhat Kim, Changkyu Yuen, Hector Ozdal, Mustafa Nair, Krishnakumar Gao, Isabel Su, Bor-Yiing Yang, Jiyan Smelyanskiy, Mikhail |
| description | Large-scale training is important to ensure high performance and accuracy of
machine-learning models. At Facebook we use many different models, including
computer vision, video and language models. However, in this paper we focus on
the deep learning recommendation models (DLRMs), which are responsible for more
than 50% of the training demand in our data centers. Recommendation models
present unique challenges in training because they exercise not only compute
but also memory capacity as well as memory and network bandwidth. As model size
and complexity increase, efficiently scaling training becomes a challenge. To
address it we design Zion - Facebook's next-generation large-memory training
platform that consists of both CPUs and accelerators. Also, we discuss the
design requirements of future scale-out training systems. |
| doi_str_mv | 10.48550/arxiv.2003.09518 |
| format | Article |
| fullrecord | <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2003_09518</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2003_09518</sourcerecordid><originalsourceid>FETCH-LOGICAL-a678-df4d9881021e9de568647f217c0b30f349ef19ad7be795e67479fa7b50a3b713</originalsourceid><addsrcrecordid>eNotz7FOwzAYBGAvDKjwAEz4BRLs2I5tNpRQQIrUIV2Yot_J78qidSInRfTtEaHT3S0nfYQ8cJZLoxR7gvQTvvOCMZEzq7i5JZ814kQbhBRDPNB9grCWEOkWenTj-EVrWIBWGBdM8zOtcQ6HSEdP2x6OmJ0nCnG4jvG80PYyL3ia78iNh-OM99fckHb7uq_es2b39lG9NBmU2mSDl4M1hrOCox1QlaaU2hdc98wJ5oW06LmFQTvUVmGppbYetFMMhNNcbMjj_-tK66YUTpAu3R-xW4niFwowSws</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Deep Learning Training in Facebook Data Centers: Design of Scale-up and Scale-out Systems</title><source>arXiv.org</source><creator>Naumov, Maxim ; Kim, John ; Mudigere, Dheevatsa ; Sridharan, Srinivas ; Wang, Xiaodong ; Zhao, Whitney ; Yilmaz, Serhat ; Kim, Changkyu ; Yuen, Hector ; Ozdal, Mustafa ; Nair, Krishnakumar ; Gao, Isabel ; Su, Bor-Yiing ; Yang, Jiyan ; Smelyanskiy, Mikhail</creator><creatorcontrib>Naumov, Maxim ; Kim, John ; Mudigere, Dheevatsa ; Sridharan, Srinivas ; Wang, Xiaodong ; Zhao, Whitney ; Yilmaz, Serhat ; Kim, Changkyu ; Yuen, Hector ; Ozdal, Mustafa ; Nair, Krishnakumar ; Gao, Isabel ; Su, Bor-Yiing ; Yang, Jiyan ; Smelyanskiy, Mikhail</creatorcontrib><description>Large-scale training is important to ensure high performance and accuracy of
machine-learning models. At Facebook we use many different models, including
computer vision, video and language models. However, in this paper we focus on
the deep learning recommendation models (DLRMs), which are responsible for more
than 50% of the training demand in our data centers. Recommendation models
present unique challenges in training because they exercise not only compute
but also memory capacity as well as memory and network bandwidth. As model size
and complexity increase, efficiently scaling training becomes a challenge. To
address it we design Zion - Facebook's next-generation large-memory training
platform that consists of both CPUs and accelerators. Also, we discuss the
design requirements of future scale-out training systems.</description><identifier>DOI: 10.48550/arxiv.2003.09518</identifier><language>eng</language><subject>Computer Science - Distributed, Parallel, and Cluster Computing</subject><creationdate>2020-03</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,780,885</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2003.09518$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2003.09518$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Naumov, Maxim</creatorcontrib><creatorcontrib>Kim, John</creatorcontrib><creatorcontrib>Mudigere, Dheevatsa</creatorcontrib><creatorcontrib>Sridharan, Srinivas</creatorcontrib><creatorcontrib>Wang, Xiaodong</creatorcontrib><creatorcontrib>Zhao, Whitney</creatorcontrib><creatorcontrib>Yilmaz, Serhat</creatorcontrib><creatorcontrib>Kim, Changkyu</creatorcontrib><creatorcontrib>Yuen, Hector</creatorcontrib><creatorcontrib>Ozdal, Mustafa</creatorcontrib><creatorcontrib>Nair, Krishnakumar</creatorcontrib><creatorcontrib>Gao, Isabel</creatorcontrib><creatorcontrib>Su, Bor-Yiing</creatorcontrib><creatorcontrib>Yang, Jiyan</creatorcontrib><creatorcontrib>Smelyanskiy, Mikhail</creatorcontrib><title>Deep Learning Training in Facebook Data Centers: Design of Scale-up and Scale-out Systems</title><description>Large-scale training is important to ensure high performance and accuracy of
machine-learning models. At Facebook we use many different models, including
computer vision, video and language models. However, in this paper we focus on
the deep learning recommendation models (DLRMs), which are responsible for more
than 50% of the training demand in our data centers. Recommendation models
present unique challenges in training because they exercise not only compute
but also memory capacity as well as memory and network bandwidth. As model size
and complexity increase, efficiently scaling training becomes a challenge. To
address it we design Zion - Facebook's next-generation large-memory training
platform that consists of both CPUs and accelerators. Also, we discuss the
design requirements of future scale-out training systems.</description><subject>Computer Science - Distributed, Parallel, and Cluster Computing</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2020</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotz7FOwzAYBGAvDKjwAEz4BRLs2I5tNpRQQIrUIV2Yot_J78qidSInRfTtEaHT3S0nfYQ8cJZLoxR7gvQTvvOCMZEzq7i5JZ814kQbhBRDPNB9grCWEOkWenTj-EVrWIBWGBdM8zOtcQ6HSEdP2x6OmJ0nCnG4jvG80PYyL3ia78iNh-OM99fckHb7uq_es2b39lG9NBmU2mSDl4M1hrOCox1QlaaU2hdc98wJ5oW06LmFQTvUVmGppbYetFMMhNNcbMjj_-tK66YUTpAu3R-xW4niFwowSws</recordid><startdate>20200320</startdate><enddate>20200320</enddate><creator>Naumov, Maxim</creator><creator>Kim, John</creator><creator>Mudigere, Dheevatsa</creator><creator>Sridharan, Srinivas</creator><creator>Wang, Xiaodong</creator><creator>Zhao, Whitney</creator><creator>Yilmaz, Serhat</creator><creator>Kim, Changkyu</creator><creator>Yuen, Hector</creator><creator>Ozdal, Mustafa</creator><creator>Nair, Krishnakumar</creator><creator>Gao, Isabel</creator><creator>Su, Bor-Yiing</creator><creator>Yang, Jiyan</creator><creator>Smelyanskiy, Mikhail</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20200320</creationdate><title>Deep Learning Training in Facebook Data Centers: Design of Scale-up and Scale-out Systems</title><author>Naumov, Maxim ; Kim, John ; Mudigere, Dheevatsa ; Sridharan, Srinivas ; Wang, Xiaodong ; Zhao, Whitney ; Yilmaz, Serhat ; Kim, Changkyu ; Yuen, Hector ; Ozdal, Mustafa ; Nair, Krishnakumar ; Gao, Isabel ; Su, Bor-Yiing ; Yang, Jiyan ; Smelyanskiy, Mikhail</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a678-df4d9881021e9de568647f217c0b30f349ef19ad7be795e67479fa7b50a3b713</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2020</creationdate><topic>Computer Science - Distributed, Parallel, and Cluster Computing</topic><toplevel>online_resources</toplevel><creatorcontrib>Naumov, Maxim</creatorcontrib><creatorcontrib>Kim, John</creatorcontrib><creatorcontrib>Mudigere, Dheevatsa</creatorcontrib><creatorcontrib>Sridharan, Srinivas</creatorcontrib><creatorcontrib>Wang, Xiaodong</creatorcontrib><creatorcontrib>Zhao, Whitney</creatorcontrib><creatorcontrib>Yilmaz, Serhat</creatorcontrib><creatorcontrib>Kim, Changkyu</creatorcontrib><creatorcontrib>Yuen, Hector</creatorcontrib><creatorcontrib>Ozdal, Mustafa</creatorcontrib><creatorcontrib>Nair, Krishnakumar</creatorcontrib><creatorcontrib>Gao, Isabel</creatorcontrib><creatorcontrib>Su, Bor-Yiing</creatorcontrib><creatorcontrib>Yang, Jiyan</creatorcontrib><creatorcontrib>Smelyanskiy, Mikhail</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Naumov, Maxim</au><au>Kim, John</au><au>Mudigere, Dheevatsa</au><au>Sridharan, Srinivas</au><au>Wang, Xiaodong</au><au>Zhao, Whitney</au><au>Yilmaz, Serhat</au><au>Kim, Changkyu</au><au>Yuen, Hector</au><au>Ozdal, Mustafa</au><au>Nair, Krishnakumar</au><au>Gao, Isabel</au><au>Su, Bor-Yiing</au><au>Yang, Jiyan</au><au>Smelyanskiy, Mikhail</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Deep Learning Training in Facebook Data Centers: Design of Scale-up and Scale-out Systems</atitle><date>2020-03-20</date><risdate>2020</risdate><abstract>Large-scale training is important to ensure high performance and accuracy of
machine-learning models. At Facebook we use many different models, including
computer vision, video and language models. However, in this paper we focus on
the deep learning recommendation models (DLRMs), which are responsible for more
than 50% of the training demand in our data centers. Recommendation models
present unique challenges in training because they exercise not only compute
but also memory capacity as well as memory and network bandwidth. As model size
and complexity increase, efficiently scaling training becomes a challenge. To
address it we design Zion - Facebook's next-generation large-memory training
platform that consists of both CPUs and accelerators. Also, we discuss the
design requirements of future scale-out training systems.</abstract><doi>10.48550/arxiv.2003.09518</doi><oa>free_for_read</oa></addata></record> |
| fulltext | fulltext_linktorsrc |
| identifier | DOI: 10.48550/arxiv.2003.09518 |
| ispartof | |
| issn | |
| language | eng |
| recordid | cdi_arxiv_primary_2003_09518 |
| source | arXiv.org |
| subjects | Computer Science - Distributed, Parallel, and Cluster Computing |
| title | Deep Learning Training in Facebook Data Centers: Design of Scale-up and Scale-out Systems |
| url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-28T17%3A55%3A40IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Deep%20Learning%20Training%20in%20Facebook%20Data%20Centers:%20Design%20of%20Scale-up%20and%20Scale-out%20Systems&rft.au=Naumov,%20Maxim&rft.date=2020-03-20&rft_id=info:doi/10.48550/arxiv.2003.09518&rft_dat=%3Carxiv_GOX%3E2003_09518%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true |