{"id":814942,"date":"2022-02-21T02:33:06","date_gmt":"2022-02-21T10:33:06","guid":{"rendered":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/?post_type=msr-research-item&p=814942"},"modified":"2024-10-17T12:51:18","modified_gmt":"2024-10-17T19:51:18","slug":"jarvis-large-scale-server-monitoring-with-adaptive-near-data-processing","status":"publish","type":"msr-research-item","link":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/publication\/jarvis-large-scale-server-monitoring-with-adaptive-near-data-processing\/","title":{"rendered":"Jarvis: Large-scale Server Monitoring with Adaptive Near-data Processing"},"content":{"rendered":"

Rapid detection and mitigation of issues that impact <\/span>performance<\/span> and<\/span> reliability<\/span> is<\/span> paramount<\/span> for<\/span> large-scale<\/span> online <\/span>services.<\/span> For<\/span> real-time<\/span> detection<\/span> of<\/span> such<\/span> issues,<\/span> datacenter<\/span> oper<\/span>ators<\/span> use<\/span> a<\/span> stream<\/span> processor<\/span> and<\/span> analyze<\/span> streams<\/span> of<\/span> monitoring <\/span>data collected from servers (referred to as data source nodes) and <\/span>their<\/span> hosted<\/span> services.<\/span> The<\/span> timely<\/span> processing<\/span> of<\/span> incoming<\/span> streams <\/span>requires<\/span> the<\/span> network<\/span> to<\/span> transfer<\/span> massive<\/span> amounts<\/span> of<\/span> data,<\/span> and <\/span>significant<\/span> compute<\/span> resources<\/span> to<\/span> process<\/span> it.<\/span> These<\/span> factors<\/span> often <\/span>create<\/span> bottlenecks<\/span> for<\/span> stream<\/span> analytics.<\/span><\/p>\n

To<\/span> help<\/span> overcome<\/span> these<\/span> bottlenecks,<\/span> current<\/span> monitoring<\/span> sys<\/span>tems<\/span> employ<\/span> near-data<\/span> processing<\/span> by<\/span> either<\/span> computing<\/span> an<\/span> op<\/span>timal<\/span> query<\/span> partition<\/span> based<\/span> on<\/span> a<\/span> cost<\/span> model<\/span> or<\/span> using<\/span> model-<\/span>agnostic<\/span> heuristics.<\/span> Optimal<\/span> partitioning<\/span> is<\/span> computationally<\/span> ex<\/span>pensive,<\/span> while<\/span> model-agnostic<\/span> heuristics<\/span> are<\/span> iterative<\/span> and<\/span> search <\/span>over<\/span> a<\/span> large<\/span> solution<\/span> space.<\/span> We<\/span> combine<\/span> these<\/span> approaches <\/span>by<\/span> using<\/span> model-agnostic<\/span> heuristics<\/span> to<\/span> improve<\/span> the<\/span> partitioning <\/span>solution from a model-based heuristic. Moreover, current systems <\/span>use<\/span> operator-level<\/span> partitioning:<\/span> if<\/span> a<\/span> data<\/span> source<\/span> does<\/span> not<\/span> have <\/span>sufficient<\/span> resources<\/span> to<\/span> execute<\/span> an<\/span> operator<\/span> on<\/span> all<\/span> records,<\/span> the <\/span>operator<\/span> is<\/span> executed<\/span> only<\/span> on<\/span> the<\/span> stream<\/span> processor.<\/span> Instead,<\/span> we <\/span>perform<\/span> data-level<\/span> partitioning\u2014i.e.,<\/span> we<\/span> allow<\/span> an<\/span> operator<\/span> to<\/span> be <\/span>executed<\/span> both<\/span> on<\/span> a<\/span> stream<\/span> processor<\/span> and<\/span> data<\/span> sources.<\/span><\/p>\n

We<\/span> implement<\/span> our<\/span> algorithm<\/span> in<\/span> a<\/span> system<\/span> called<\/span> Jarvis,<\/span> which <\/span>enables<\/span> quick<\/span> adaptation<\/span> to<\/span> dynamic<\/span> resource<\/span> conditions.<\/span> Our <\/span>evaluation<\/span> on<\/span> a<\/span> diverse<\/span> set<\/span> of<\/span> monitoring<\/span> workloads<\/span> suggests <\/span>that<\/span> Jarvis<\/span> converges<\/span> to<\/span> a<\/span> stable<\/span> query<\/span> partition<\/span> within<\/span> seconds <\/span>of<\/span> a<\/span> change<\/span> in<\/span> node<\/span> resource<\/span> conditions.<\/span> Compared<\/span> to<\/span> current <\/span>partitioning<\/span> strategies,<\/span> Jarvis<\/span> handles<\/span> up<\/span> to<\/span> 75%<\/span> more<\/span> data <\/span>sources<\/span> while<\/span> improving<\/span> throughput<\/span> in<\/span> resource-constrained <\/span>scenarios<\/span> by<\/span> 1.2-4.4<\/span>\u00d7<\/span>.<\/span><\/p>\n","protected":false},"excerpt":{"rendered":"

Rapid detection and mitigation of issues that impact performance and reliability is paramount for large-scale online services. For real-time detection of such issues, datacenter operators use a stream processor and analyze streams of monitoring data collected from servers (referred to as data source nodes) and their hosted services. The timely processing of incoming streams requires […]<\/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":"","msr-author-ordering":null,"msr_publishername":"","msr_publisher_other":"","msr_booktitle":"","msr_chapter":"","msr_edition":"","msr_editors":"","msr_how_published":"","msr_isbn":"","msr_issue":"","msr_journal":"","msr_number":"","msr_organization":"","msr_pages_string":"","msr_page_range_start":"","msr_page_range_end":"","msr_series":"","msr_volume":"","msr_copyright":"","msr_conference_name":"38th IEEE International Conference on Data Engineering","msr_doi":"","msr_arxiv_id":"","msr_s2_paper_id":"","msr_mag_id":"","msr_pubmed_id":"","msr_other_authors":"","msr_other_contributors":"","msr_speaker":"","msr_award":"","msr_affiliation":"","msr_institution":"","msr_host":"","msr_version":"","msr_duration":"","msr_original_fields_of_study":"","msr_release_tracker_id":"","msr_s2_match_type":"","msr_citation_count_updated":"","msr_published_date":"2022-5-9","msr_highlight_text":"ICDE Best Paper Award","msr_notes":"","msr_longbiography":"","msr_publicationurl":"","msr_external_url":"","msr_secondary_video_url":"","msr_conference_url":"","msr_journal_url":"","msr_s2_pdf_url":"","msr_year":0,"msr_citation_count":0,"msr_influential_citations":0,"msr_reference_count":0,"msr_s2_match_confidence":0,"msr_microsoftintellectualproperty":false,"msr_s2_open_access":false,"msr_s2_author_ids":[],"msr_pub_ids":[],"msr_hide_image_in_river":0,"footnotes":""},"msr-research-highlight":[246574],"research-area":[13547],"msr-publication-type":[193716],"msr-publisher":[],"msr-focus-area":[],"msr-locale":[268875],"msr-post-option":[],"msr-field-of-study":[],"msr-conference":[],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-814942","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-highlight-award","msr-research-area-systems-and-networking","msr-locale-en_us"],"msr_publishername":"","msr_edition":"","msr_affiliation":"","msr_published_date":"2022-5-9","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":"ICDE Best Paper Award","msr_release_tracker_id":"","msr_original_fields_of_study":"","msr_download_urls":"","msr_external_url":"","msr_secondary_video_url":"","msr_longbiography":"","msr_microsoftintellectualproperty":0,"msr_main_download":"","msr_publicationurl":"","msr_doi":"","msr_publication_uploader":[{"type":"url","viewUrl":"false","id":"false","title":"https:\/\/ieeexplore.ieee.org\/document\/9835523","label_id":"243109","label":0},{"type":"url","viewUrl":"false","id":"false","title":"https:\/\/arxiv.org\/pdf\/2202.06021v1","label_id":"252679","label":0}],"msr_related_uploader":"","msr_citation_count":0,"msr_citation_count_updated":"","msr_s2_paper_id":"","msr_influential_citations":0,"msr_reference_count":0,"msr_arxiv_id":"","msr_s2_author_ids":[],"msr_s2_open_access":false,"msr_s2_pdf_url":null,"msr_attachments":[{"id":815158,"url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-content\/uploads\/2022\/01\/icde2022_jarvis_dist.pdf"}],"msr-author-ordering":[{"type":"text","value":"Atul Sandur","user_id":0,"rest_url":false},{"type":"text","value":"ChanHo Park","user_id":0,"rest_url":false},{"type":"user_nicename","value":"Stavros Volos","user_id":35437,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Stavros Volos"},{"type":"text","value":"Gul Agha","user_id":0,"rest_url":false},{"type":"user_nicename","value":"Myeongjae Jeon","user_id":33040,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Myeongjae Jeon"}],"msr_impact_theme":[],"msr_research_lab":[199561],"msr_event":[],"msr_group":[],"msr_project":[],"publication":[],"video":[],"msr-tool":[],"msr_publication_type":"inproceedings","related_content":[],"_links":{"self":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/814942","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item"}],"about":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-research-item"}],"version-history":[{"count":5,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/814942\/revisions"}],"predecessor-version":[{"id":1088712,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/814942\/revisions\/1088712"}],"wp:attachment":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/media?parent=814942"}],"wp:term":[{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=814942"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=814942"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=814942"},{"taxonomy":"msr-publisher","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-publisher?post=814942"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=814942"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=814942"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=814942"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=814942"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=814942"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=814942"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=814942"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=814942"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}