The broadening adoption of machine learning in the enterprise is increasing the pressure for strict governance and cost-effective performance, in particular for the common and consequential steps of model storage and inference.<\/p>\n
The RDBMS provides a natural starting point, given its mature infrastructure for fast data access and processing, along with support for enterprise features, such as encryption, auditing, and high-availability. To take advantage of all of the above, we need to address a key concern: Can in-RDBMS scoring of ML models match (outperform?) the performance of dedicated frameworks?<\/p>\n
We answer the above positively by building Raven, a system that leverages native integration of ML runtimes (such as ONNX Runtime) deep within SQL Server and a unified intermediate representation (IR) to enable advanced cross-optimizations between ML and database operators. In this optimization space, we discover the most exciting research opportunities that combine DB\/compiler\/ML thinking. Our initial evaluation on real data demonstrates performance gains of up to 5.5\u00d7 from the native integration of ML in SQL Server and up to 24\u00d7 from cross-optimizations. An early preview of the ONNX Runtime integration is currently available with Azure\u2019s SQL Database Edge.<\/p>\n","protected":false},"excerpt":{"rendered":"
The broadening adoption of machine learning in the enterprise is increasing the pressure for strict governance and cost-effective performance, in particular for the common and consequential steps of model storage and inference. The RDBMS provides a natural starting point, given its mature infrastructure for fast data access and processing, along with support for enterprise features, […]<\/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":[13556,13563],"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":[246784,252424,246685,252427,249145],"msr-conference":[],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-725959","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-area-artificial-intelligence","msr-research-area-data-platform-analytics","msr-locale-en_us","msr-field-of-study-analytics","msr-field-of-study-in-database-machine-learning","msr-field-of-study-machine-learning","msr-field-of-study-model-inference","msr-field-of-study-query-optimization"],"msr_publishername":"","msr_edition":"","msr_affiliation":"","msr_published_date":"2020-1-1","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":"http:\/\/cidrdb.org\/cidr2020\/papers\/p24-karanasos-cidr20.pdf","label_id":"243109","label":0}],"msr_related_uploader":"","msr_attachments":[],"msr-author-ordering":[{"type":"user_nicename","value":"Konstantinos Karanasos","user_id":32565,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Konstantinos Karanasos"},{"type":"user_nicename","value":"Matteo Interlandi","user_id":39784,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Matteo Interlandi"},{"type":"text","value":"Doris Xin","user_id":0,"rest_url":false},{"type":"user_nicename","value":"Fotis Psallidas","user_id":40057,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Fotis Psallidas"},{"type":"user_nicename","value":"Rathijit Sen","user_id":39450,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Rathijit Sen"},{"type":"user_nicename","value":"Kwanghyun Park","user_id":40078,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Kwanghyun Park"},{"type":"text","value":"Ivan Popivanov","user_id":0,"rest_url":false},{"type":"text","value":"Supun Nakandal","user_id":0,"rest_url":false},{"type":"user_nicename","value":"Subramaniam Venkatraman Krishnan","user_id":33746,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Subramaniam Venkatraman Krishnan"},{"type":"user_nicename","value":"Markus Weimer","user_id":33038,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Markus Weimer"},{"type":"user_nicename","value":"Yuan Yu","user_id":35054,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Yuan Yu"},{"type":"user_nicename","value":"Raghu Ramakrishnan","user_id":40051,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Raghu Ramakrishnan"},{"type":"user_nicename","value":"Carlo Curino","user_id":31352,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Carlo Curino"}],"msr_impact_theme":[],"msr_research_lab":[],"msr_event":[],"msr_group":[684024],"msr_project":[],"publication":[],"video":[],"download":[],"msr_publication_type":"inproceedings","related_content":[],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/725959"}],"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\/725959\/revisions"}],"predecessor-version":[{"id":725962,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/725959\/revisions\/725962"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=725959"}],"wp:term":[{"taxonomy":"msr-content-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-content-type?post=725959"},{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=725959"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=725959"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=725959"},{"taxonomy":"msr-product-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-product-type?post=725959"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=725959"},{"taxonomy":"msr-platform","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-platform?post=725959"},{"taxonomy":"msr-download-source","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-download-source?post=725959"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=725959"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=725959"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=725959"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=725959"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=725959"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=725959"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=725959"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}