Modern state-of-the-art deep learning (DL) applications tend to scale out to a large number of parallel GPUs. Unfortunately, we observe that the collective communication overhead across GPUs is often the key limiting factor of performance for distributed DL. It under-utilizes the networking bandwidth by frequent transfers of small data chunks, which also incurs a substantial I\/O overhead on GPU that interferes with computation on GPU. The root cause lies in the inefficiency of CPU-based communication event handling as well as the inability to control the GPU’s internal DMA engine with GPU threads.<\/p>\n
To address the problem, we propose a GPU-driven code execution system that leverages a GPU-controlled hardware DMA engine for I\/O offloading. Our custom DMA engine pipelines multiple DMA requests to support efficient small data transfer while it eliminates the I\/O overhead on GPU cores. Unlike existing GPU DMA engines initiated only by CPU, we let GPU threads to directly control DMA operations, which leads to a highly efficient system where GPUs drive their own execution flow and handle communication events autonomously without CPU intervention. Our prototype DMA engine achieves a line-rate from a message size as small as 8KB (3.9x better throughput) with only 4.3us of communication latency (9.1x faster) while it incurs little interference with computation on GPU, achieving 1.8x higher all-reduce throughput in a real training workload.<\/p>\n","protected":false},"excerpt":{"rendered":"
Modern state-of-the-art deep learning (DL) applications tend to scale out to a large number of parallel GPUs. Unfortunately, we observe that the collective communication overhead across GPUs is often the key limiting factor of performance for distributed DL. It under-utilizes the networking bandwidth by frequent transfers of small data chunks, which also incurs a substantial […]<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"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-field-of-study":[246658],"msr-conference":[263941],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-934836","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-field-of-study-deep-learning"],"msr_publishername":"","msr_edition":"","msr_affiliation":"","msr_published_date":"2023-4-13","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":"url","viewUrl":"false","id":"false","title":"https:\/\/cp5555.github.io\/publications\/ark-nsdi23.pdf","label_id":"243109","label":0}],"msr_related_uploader":[{"type":"file","viewUrl":"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/04\/ark-nsdi23.pdf","id":"934845","title":"ark-nsdi23","label_id":"243112","label":0}],"msr_attachments":[{"id":934845,"url":"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2023\/04\/ark-nsdi23.pdf"}],"msr-author-ordering":[{"type":"user_nicename","value":"Changho Hwang","user_id":41844,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Changho Hwang"},{"type":"user_nicename","value":"Ran Shu","user_id":37926,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Ran Shu"},{"type":"user_nicename","value":"Peng Cheng","user_id":33225,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Peng Cheng"},{"type":"user_nicename","value":"Yongqiang Xiong","user_id":35049,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Yongqiang Xiong"},{"type":"guest","value":"kyoungsoo-park","user_id":934854,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=kyoungsoo-park"}],"msr_impact_theme":[],"msr_research_lab":[],"msr_event":[],"msr_group":[],"msr_project":[],"publication":[],"video":[],"download":[],"msr_publication_type":"inproceedings","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/934836"}],"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\/934836\/revisions"}],"predecessor-version":[{"id":934857,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/934836\/revisions\/934857"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=934836"}],"wp:term":[{"taxonomy":"msr-content-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-content-type?post=934836"},{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=934836"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=934836"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=934836"},{"taxonomy":"msr-product-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-product-type?post=934836"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=934836"},{"taxonomy":"msr-platform","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-platform?post=934836"},{"taxonomy":"msr-download-source","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-download-source?post=934836"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=934836"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=934836"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=934836"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=934836"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=934836"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=934836"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}