Contiguitas: The Pursuit of Physical Memory Contiguity in Data Centers

The unabating growth of the memory needs of emerging data center applications has exacerbated the scalability bottleneck of virtual memory. However, reducing the overhead of address translation will remain onerous until the physical memory contiguity predicament gets resolved. To address this proble...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE MICRO 2024-07, Vol.44 (4), p.44-51
Hauptverfasser: Zhao, Kaiyang, Xue, Kaiwen, Wang, Ziqi, Schatzberg, Dan, Yang, Leon, Manousis, Antonis, Weiner, Johannes, Riel, Rik Van, Sharma, Bikash, Tang, Chunqiang, Skarlatos, Dimitrios
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page 51
container_issue 4
container_start_page 44
container_title IEEE MICRO
container_volume 44
creator Zhao, Kaiyang
Xue, Kaiwen
Wang, Ziqi
Schatzberg, Dan
Yang, Leon
Manousis, Antonis
Weiner, Johannes
Riel, Rik Van
Sharma, Bikash
Tang, Chunqiang
Skarlatos, Dimitrios
description The unabating growth of the memory needs of emerging data center applications has exacerbated the scalability bottleneck of virtual memory. However, reducing the overhead of address translation will remain onerous until the physical memory contiguity predicament gets resolved. To address this problem, Contiguitas provides ample physical memory contiguity by design. We identify that the primary cause of memory fragmentation in Meta’s data centers is unmovable allocations scattered across the address space that impedes contiguity. To this end, Contiguitas in the operating system separates movable allocations from unmovable ones by placing them into two different dynamically adjustable regions in physical memory. Furthermore, Contiguitas drastically reduces unmovable allocations through hardware extensions that transparently migrate unmovable pages while they remain in use. Our experiments in production at Meta’s data centers show that Contiguitas achieves end-to-end performance improvements of 2%–18%. Full-system simulations of the Contiguitas hardware show that it can efficiently migrate unmovable allocations without affecting applications.
doi_str_mv 10.1109/MM.2024.3406933
format Article
fullrecord <record><control><sourceid>crossref_RIE</sourceid><recordid>TN_cdi_crossref_primary_10_1109_MM_2024_3406933</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><ieee_id>10543201</ieee_id><sourcerecordid>10_1109_MM_2024_3406933</sourcerecordid><originalsourceid>FETCH-LOGICAL-c146t-d5998835a6edd18ca1535ff0a42adfd6bcb8396e5fc781a285fcd28019f236ab3</originalsourceid><addsrcrecordid>eNpNkDFPwzAUhC0EEqEwszD4D6R9z89ObDYUKCA1okOZIzexaVCbIDsd8u9p1Qox3Z10d8PH2D3CFBHMrCynAoSckoTMEF2wBA3lqURJlywBkYsUcxLX7CbGbwBQAnTC5kXfDe3Xvh1sfOSrjePLfYiHyHvPl5sxtrXd8tLt-jDyv-7I244_28HywnWDC_GWXXm7je7urBP2OX9ZFW_p4uP1vXhapDXKbEgbZYzWpGzmmgZ1bVGR8h6sFLbxTbau15pM5pSvc41W6INphAY0XlBm1zRhs9NvHfoYg_PVT2h3NowVQnXEUJVldcRQnTEcFg-nReuc-9dWkgQg_QK1olkw</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Contiguitas: The Pursuit of Physical Memory Contiguity in Data Centers</title><source>IEEE Electronic Library (IEL)</source><creator>Zhao, Kaiyang ; Xue, Kaiwen ; Wang, Ziqi ; Schatzberg, Dan ; Yang, Leon ; Manousis, Antonis ; Weiner, Johannes ; Riel, Rik Van ; Sharma, Bikash ; Tang, Chunqiang ; Skarlatos, Dimitrios</creator><creatorcontrib>Zhao, Kaiyang ; Xue, Kaiwen ; Wang, Ziqi ; Schatzberg, Dan ; Yang, Leon ; Manousis, Antonis ; Weiner, Johannes ; Riel, Rik Van ; Sharma, Bikash ; Tang, Chunqiang ; Skarlatos, Dimitrios</creatorcontrib><description>The unabating growth of the memory needs of emerging data center applications has exacerbated the scalability bottleneck of virtual memory. However, reducing the overhead of address translation will remain onerous until the physical memory contiguity predicament gets resolved. To address this problem, Contiguitas provides ample physical memory contiguity by design. We identify that the primary cause of memory fragmentation in Meta’s data centers is unmovable allocations scattered across the address space that impedes contiguity. To this end, Contiguitas in the operating system separates movable allocations from unmovable ones by placing them into two different dynamically adjustable regions in physical memory. Furthermore, Contiguitas drastically reduces unmovable allocations through hardware extensions that transparently migrate unmovable pages while they remain in use. Our experiments in production at Meta’s data centers show that Contiguitas achieves end-to-end performance improvements of 2%–18%. Full-system simulations of the Contiguitas hardware show that it can efficiently migrate unmovable allocations without affecting applications.</description><identifier>ISSN: 0272-1732</identifier><identifier>EISSN: 1937-4143</identifier><identifier>DOI: 10.1109/MM.2024.3406933</identifier><identifier>CODEN: IEMIDZ</identifier><language>eng</language><publisher>IEEE</publisher><subject>Data centers ; Hardware ; Linux ; Memory management ; Production ; Resource management ; Scalability ; Servers</subject><ispartof>IEEE MICRO, 2024-07, Vol.44 (4), p.44-51</ispartof><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><cites>FETCH-LOGICAL-c146t-d5998835a6edd18ca1535ff0a42adfd6bcb8396e5fc781a285fcd28019f236ab3</cites><orcidid>0009-0000-0945-4429 ; 0009-0005-6697-8276 ; 0000-0003-0067-0701 ; 0009-0006-5702-7687 ; 0009-0002-5311-2826 ; 0009-0004-0133-4800 ; 0009-0003-8130-4435 ; 0000-0002-7856-4274 ; 0009-0001-8347-8207 ; 0009-0009-8010-3635 ; 0000-0002-0289-5499</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://ieeexplore.ieee.org/document/10543201$$EHTML$$P50$$Gieee$$H</linktohtml><link.rule.ids>314,780,784,796,27924,27925,54758</link.rule.ids><linktorsrc>$$Uhttps://ieeexplore.ieee.org/document/10543201$$EView_record_in_IEEE$$FView_record_in_$$GIEEE</linktorsrc></links><search><creatorcontrib>Zhao, Kaiyang</creatorcontrib><creatorcontrib>Xue, Kaiwen</creatorcontrib><creatorcontrib>Wang, Ziqi</creatorcontrib><creatorcontrib>Schatzberg, Dan</creatorcontrib><creatorcontrib>Yang, Leon</creatorcontrib><creatorcontrib>Manousis, Antonis</creatorcontrib><creatorcontrib>Weiner, Johannes</creatorcontrib><creatorcontrib>Riel, Rik Van</creatorcontrib><creatorcontrib>Sharma, Bikash</creatorcontrib><creatorcontrib>Tang, Chunqiang</creatorcontrib><creatorcontrib>Skarlatos, Dimitrios</creatorcontrib><title>Contiguitas: The Pursuit of Physical Memory Contiguity in Data Centers</title><title>IEEE MICRO</title><addtitle>MM</addtitle><description>The unabating growth of the memory needs of emerging data center applications has exacerbated the scalability bottleneck of virtual memory. However, reducing the overhead of address translation will remain onerous until the physical memory contiguity predicament gets resolved. To address this problem, Contiguitas provides ample physical memory contiguity by design. We identify that the primary cause of memory fragmentation in Meta’s data centers is unmovable allocations scattered across the address space that impedes contiguity. To this end, Contiguitas in the operating system separates movable allocations from unmovable ones by placing them into two different dynamically adjustable regions in physical memory. Furthermore, Contiguitas drastically reduces unmovable allocations through hardware extensions that transparently migrate unmovable pages while they remain in use. Our experiments in production at Meta’s data centers show that Contiguitas achieves end-to-end performance improvements of 2%–18%. Full-system simulations of the Contiguitas hardware show that it can efficiently migrate unmovable allocations without affecting applications.</description><subject>Data centers</subject><subject>Hardware</subject><subject>Linux</subject><subject>Memory management</subject><subject>Production</subject><subject>Resource management</subject><subject>Scalability</subject><subject>Servers</subject><issn>0272-1732</issn><issn>1937-4143</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>RIE</sourceid><recordid>eNpNkDFPwzAUhC0EEqEwszD4D6R9z89ObDYUKCA1okOZIzexaVCbIDsd8u9p1Qox3Z10d8PH2D3CFBHMrCynAoSckoTMEF2wBA3lqURJlywBkYsUcxLX7CbGbwBQAnTC5kXfDe3Xvh1sfOSrjePLfYiHyHvPl5sxtrXd8tLt-jDyv-7I244_28HywnWDC_GWXXm7je7urBP2OX9ZFW_p4uP1vXhapDXKbEgbZYzWpGzmmgZ1bVGR8h6sFLbxTbau15pM5pSvc41W6INphAY0XlBm1zRhs9NvHfoYg_PVT2h3NowVQnXEUJVldcRQnTEcFg-nReuc-9dWkgQg_QK1olkw</recordid><startdate>202407</startdate><enddate>202407</enddate><creator>Zhao, Kaiyang</creator><creator>Xue, Kaiwen</creator><creator>Wang, Ziqi</creator><creator>Schatzberg, Dan</creator><creator>Yang, Leon</creator><creator>Manousis, Antonis</creator><creator>Weiner, Johannes</creator><creator>Riel, Rik Van</creator><creator>Sharma, Bikash</creator><creator>Tang, Chunqiang</creator><creator>Skarlatos, Dimitrios</creator><general>IEEE</general><scope>97E</scope><scope>RIA</scope><scope>RIE</scope><scope>AAYXX</scope><scope>CITATION</scope><orcidid>https://orcid.org/0009-0000-0945-4429</orcidid><orcidid>https://orcid.org/0009-0005-6697-8276</orcidid><orcidid>https://orcid.org/0000-0003-0067-0701</orcidid><orcidid>https://orcid.org/0009-0006-5702-7687</orcidid><orcidid>https://orcid.org/0009-0002-5311-2826</orcidid><orcidid>https://orcid.org/0009-0004-0133-4800</orcidid><orcidid>https://orcid.org/0009-0003-8130-4435</orcidid><orcidid>https://orcid.org/0000-0002-7856-4274</orcidid><orcidid>https://orcid.org/0009-0001-8347-8207</orcidid><orcidid>https://orcid.org/0009-0009-8010-3635</orcidid><orcidid>https://orcid.org/0000-0002-0289-5499</orcidid></search><sort><creationdate>202407</creationdate><title>Contiguitas: The Pursuit of Physical Memory Contiguity in Data Centers</title><author>Zhao, Kaiyang ; Xue, Kaiwen ; Wang, Ziqi ; Schatzberg, Dan ; Yang, Leon ; Manousis, Antonis ; Weiner, Johannes ; Riel, Rik Van ; Sharma, Bikash ; Tang, Chunqiang ; Skarlatos, Dimitrios</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c146t-d5998835a6edd18ca1535ff0a42adfd6bcb8396e5fc781a285fcd28019f236ab3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Data centers</topic><topic>Hardware</topic><topic>Linux</topic><topic>Memory management</topic><topic>Production</topic><topic>Resource management</topic><topic>Scalability</topic><topic>Servers</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Zhao, Kaiyang</creatorcontrib><creatorcontrib>Xue, Kaiwen</creatorcontrib><creatorcontrib>Wang, Ziqi</creatorcontrib><creatorcontrib>Schatzberg, Dan</creatorcontrib><creatorcontrib>Yang, Leon</creatorcontrib><creatorcontrib>Manousis, Antonis</creatorcontrib><creatorcontrib>Weiner, Johannes</creatorcontrib><creatorcontrib>Riel, Rik Van</creatorcontrib><creatorcontrib>Sharma, Bikash</creatorcontrib><creatorcontrib>Tang, Chunqiang</creatorcontrib><creatorcontrib>Skarlatos, Dimitrios</creatorcontrib><collection>IEEE All-Society Periodicals Package (ASPP) 2005-present</collection><collection>IEEE All-Society Periodicals Package (ASPP) 1998-Present</collection><collection>IEEE Electronic Library (IEL)</collection><collection>CrossRef</collection><jtitle>IEEE MICRO</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Zhao, Kaiyang</au><au>Xue, Kaiwen</au><au>Wang, Ziqi</au><au>Schatzberg, Dan</au><au>Yang, Leon</au><au>Manousis, Antonis</au><au>Weiner, Johannes</au><au>Riel, Rik Van</au><au>Sharma, Bikash</au><au>Tang, Chunqiang</au><au>Skarlatos, Dimitrios</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Contiguitas: The Pursuit of Physical Memory Contiguity in Data Centers</atitle><jtitle>IEEE MICRO</jtitle><stitle>MM</stitle><date>2024-07</date><risdate>2024</risdate><volume>44</volume><issue>4</issue><spage>44</spage><epage>51</epage><pages>44-51</pages><issn>0272-1732</issn><eissn>1937-4143</eissn><coden>IEMIDZ</coden><abstract>The unabating growth of the memory needs of emerging data center applications has exacerbated the scalability bottleneck of virtual memory. However, reducing the overhead of address translation will remain onerous until the physical memory contiguity predicament gets resolved. To address this problem, Contiguitas provides ample physical memory contiguity by design. We identify that the primary cause of memory fragmentation in Meta’s data centers is unmovable allocations scattered across the address space that impedes contiguity. To this end, Contiguitas in the operating system separates movable allocations from unmovable ones by placing them into two different dynamically adjustable regions in physical memory. Furthermore, Contiguitas drastically reduces unmovable allocations through hardware extensions that transparently migrate unmovable pages while they remain in use. Our experiments in production at Meta’s data centers show that Contiguitas achieves end-to-end performance improvements of 2%–18%. Full-system simulations of the Contiguitas hardware show that it can efficiently migrate unmovable allocations without affecting applications.</abstract><pub>IEEE</pub><doi>10.1109/MM.2024.3406933</doi><tpages>8</tpages><orcidid>https://orcid.org/0009-0000-0945-4429</orcidid><orcidid>https://orcid.org/0009-0005-6697-8276</orcidid><orcidid>https://orcid.org/0000-0003-0067-0701</orcidid><orcidid>https://orcid.org/0009-0006-5702-7687</orcidid><orcidid>https://orcid.org/0009-0002-5311-2826</orcidid><orcidid>https://orcid.org/0009-0004-0133-4800</orcidid><orcidid>https://orcid.org/0009-0003-8130-4435</orcidid><orcidid>https://orcid.org/0000-0002-7856-4274</orcidid><orcidid>https://orcid.org/0009-0001-8347-8207</orcidid><orcidid>https://orcid.org/0009-0009-8010-3635</orcidid><orcidid>https://orcid.org/0000-0002-0289-5499</orcidid></addata></record>
fulltext fulltext_linktorsrc
identifier ISSN: 0272-1732
ispartof IEEE MICRO, 2024-07, Vol.44 (4), p.44-51
issn 0272-1732
1937-4143
language eng
recordid cdi_crossref_primary_10_1109_MM_2024_3406933
source IEEE Electronic Library (IEL)
subjects Data centers
Hardware
Linux
Memory management
Production
Resource management
Scalability
Servers
title Contiguitas: The Pursuit of Physical Memory Contiguity in Data Centers
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-20T15%3A51%3A22IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-crossref_RIE&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Contiguitas:%20The%20Pursuit%20of%20Physical%20Memory%20Contiguity%20in%20Data%20Centers&rft.jtitle=IEEE%20MICRO&rft.au=Zhao,%20Kaiyang&rft.date=2024-07&rft.volume=44&rft.issue=4&rft.spage=44&rft.epage=51&rft.pages=44-51&rft.issn=0272-1732&rft.eissn=1937-4143&rft.coden=IEMIDZ&rft_id=info:doi/10.1109/MM.2024.3406933&rft_dat=%3Ccrossref_RIE%3E10_1109_MM_2024_3406933%3C/crossref_RIE%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rft_ieee_id=10543201&rfr_iscdi=true