We address the problem of serving Deep Neural Networks (DNNs) efficiently from a cluster of GPUs. In order to realize the promise of very low-cost processing made by accelerators such as GPUs, it is essential to run them at sustained high utilization. Doing so requires cluster-scale resource management that performs detailed scheduling of GPUs, reasoning about groups of DNN invocations that need to be coscheduled, and moving from the conventional whole-DNN execution model to executing fragments of DNNs. Nexus is a fully implemented system that includes these innovations. In large-scale case studies on 16 GPUs, when required to stay within latency constraints at least 99% of the time, Nexus can process requests at rates 1.8-12.7\u00d7 higher than state of the art systems can. A long-running multi-application deployment stays within 84% of optimal utilization and, on a 100-GPU cluster, violates latency SLOs on 0.27% of requests.<\/p>\n","protected":false},"excerpt":{"rendered":"
We address the problem of serving Deep Neural Networks (DNNs) efficiently from a cluster of GPUs. In order to realize the promise of very low-cost processing made by accelerators such as GPUs, it is essential to run them at sustained high utilization. Doing so requires cluster-scale resource management that performs detailed scheduling of GPUs, reasoning […]<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","footnotes":""},"msr-content-type":[3],"msr-research-highlight":[],"research-area":[13556,13547],"msr-publication-type":[193716],"msr-product-type":[],"msr-focus-area":[],"msr-platform":[],"msr-download-source":[],"msr-locale":[268875],"msr-post-option":[],"msr-field-of-study":[],"msr-conference":[],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-767323","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-area-artificial-intelligence","msr-research-area-systems-and-networking","msr-locale-en_us"],"msr_publishername":"","msr_edition":"","msr_affiliation":"","msr_published_date":"2019-10","msr_host":"","msr_duration":"","msr_version":"","msr_speaker":"","msr_other_contributors":"","msr_booktitle":"","msr_pages_string":"","msr_chapter":"","msr_isbn":"","msr_journal":"","msr_volume":"","msr_number":"","msr_editors":"","msr_series":"","msr_issue":"","msr_organization":"","msr_how_published":"","msr_notes":"","msr_highlight_text":"","msr_release_tracker_id":"","msr_original_fields_of_study":"","msr_download_urls":"","msr_external_url":"","msr_secondary_video_url":"","msr_longbiography":"","msr_microsoftintellectualproperty":1,"msr_main_download":"","msr_publicationurl":"","msr_doi":"","msr_publication_uploader":[{"type":"doi","viewUrl":"false","id":"false","title":"https:\/\/dl.acm.org\/doi\/10.1145\/3341301.3359658","label_id":"243109","label":0},{"type":"file","viewUrl":"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2021\/08\/nexus.pdf","id":"767326","title":"nexus","label_id":"243132","label":0}],"msr_related_uploader":"","msr_attachments":[{"id":767326,"url":"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2021\/08\/nexus.pdf"}],"msr-author-ordering":[{"type":"text","value":"Haichen Shen","user_id":0,"rest_url":false},{"type":"text","value":"Lequn Chen","user_id":0,"rest_url":false},{"type":"text","value":"Yuchen Jin","user_id":0,"rest_url":false},{"type":"text","value":"Liangyu Zhao","user_id":0,"rest_url":false},{"type":"text","value":"Bingyu Kong","user_id":0,"rest_url":false},{"type":"user_nicename","value":"Matthai Philipose","user_id":32834,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Matthai Philipose"},{"type":"text","value":"Arvind Krishnamurthy","user_id":0,"rest_url":false},{"type":"text","value":"Ravi Sundaram","user_id":0,"rest_url":false}],"msr_impact_theme":[],"msr_research_lab":[199565],"msr_event":[],"msr_group":[],"msr_project":[635574],"publication":[],"video":[],"download":[],"msr_publication_type":"inproceedings","related_content":{"projects":[{"ID":635574,"post_title":"Watch For","post_name":"watch-for","post_type":"msr-project","post_date":"2021-08-20 08:33:39","post_modified":"2022-12-22 12:43:09","post_status":"publish","permalink":"https:\/\/www.microsoft.com\/en-us\/research\/project\/watch-for\/","post_excerpt":"Transitioned | Watch For is a large-scale, low-cost, highly programmable media analysis platform built on Azure.","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/635574"}]}}]},"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/767323"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-research-item"}],"version-history":[{"count":1,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/767323\/revisions"}],"predecessor-version":[{"id":767329,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/767323\/revisions\/767329"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=767323"}],"wp:term":[{"taxonomy":"msr-content-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-content-type?post=767323"},{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=767323"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=767323"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=767323"},{"taxonomy":"msr-product-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-product-type?post=767323"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=767323"},{"taxonomy":"msr-platform","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-platform?post=767323"},{"taxonomy":"msr-download-source","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-download-source?post=767323"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=767323"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=767323"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=767323"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=767323"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=767323"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=767323"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=767323"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}