Weakly Coupled MDP for Load Balancing in Containerized Cloud: A Scalable Control Design

Weakly Coupled MDP for Load Balancing in Containerized Cloud: A Scalable Control Design | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Weakly Coupled MDP for Load Balancing in Containerized Cloud: A Scalable Control Design Adam Houmairi, El Mehdi Kandoussi, Yassine Maleh, Soufyane Mounir This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8051212/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 8 You are reading this latest preprint version Abstract Load balancing is a core challenge in containerized cloud systems. We formulate the dispatcher’s decision problem as a weakly coupled Markov Decision Process (MDP) and derive a scalable policy via a Lagrangian linear-programming relaxation that decouples per-VM control. A toy-scale study is first used to expose structural behavior and guide design; we then run full-scale simulations to assess performance and robustness. We prove that the optimal value function is non-decreasing in total backlog and establish a stochastic-dominance corollary. Several appealing conjectures (e.g., per-VM monotonicity) hold at low demand but fail near saturation, clarifying when “balance everything” heuristics become suboptimal. Motivated by these insights, we propose a deployable load-aware dispatcher that blends JSQ-like behavior at low load with advantage-based routing at high load. Across scaled experiments, the dispatcher consistently reduces blocking and tail delay versus standard baselines, and competes favorably with a state-of-the-art heuristic, while incurring modest control overhead. The study bridges stochastic control and deployable cloud scheduling, offering both analytical insights and a practical policy design. Containerized cloud Load balancing Dispatcher design Queueing networks Markov Decision Process (MDP) Weak coupling Lagrangian relaxation Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 14 May, 2026 Reviewers agreed at journal 06 May, 2026 Reviewers agreed at journal 04 May, 2026 Reviewers agreed at journal 24 Nov, 2025 Reviewers invited by journal 24 Nov, 2025 Editor assigned by journal 24 Nov, 2025 Submission checks completed at journal 10 Nov, 2025 First submitted to journal 06 Nov, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8051212","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":550061342,"identity":"56e12422-50fb-4616-9243-add469b388c9","order_by":0,"name":"Adam Houmairi","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA90lEQVRIiWNgGAWjYFCCBAYGxgYQA0gmVJCu5QxpWkC62ojQwN+e/HQD4w6bxLXth9s+PJx32F63/wDjhx8M2+RxaZE488zsBuOZtMRtZxKbZyRuO5y47UYCs2QPw23DBlx6biQAtbQBVR5IbGYAagFyGRikGRhuM+LSIn8j/RtEy/mHQC1zDtubnT/A/BuoxR6XFoMbOVBbboBsaTjMuO1AAhvIlkRcWgzPvCm7kdiWZrztBtCWhGPpIL1tlj0Gt5NxaZE7nr7txsc2G9lt59MfM/6osQY67PDhGz8qbtvi9D4IJKByQR43wKd+FIyCUTAKRgEhAAC/B2bEGWfCZAAAAABJRU5ErkJggg==","orcid":"","institution":"Université Sultan Moulay Slimane","correspondingAuthor":true,"prefix":"","firstName":"Adam","middleName":"","lastName":"Houmairi","suffix":""},{"id":550061343,"identity":"5e9cf920-efcb-44f9-8b84-efd6dce82678","order_by":1,"name":"El Mehdi Kandoussi","email":"","orcid":"","institution":"Institut National des Postes et Télécommunications","correspondingAuthor":false,"prefix":"","firstName":"El","middleName":"Mehdi","lastName":"Kandoussi","suffix":""},{"id":550061344,"identity":"2288cef3-1f85-4a9c-a310-4d0874aa5d20","order_by":2,"name":"Yassine Maleh","email":"","orcid":"","institution":"Université Sultan Moulay Slimane","correspondingAuthor":false,"prefix":"","firstName":"Yassine","middleName":"","lastName":"Maleh","suffix":""},{"id":550061345,"identity":"26c28d26-6ec2-4bed-804a-bd84e3812b44","order_by":3,"name":"Soufyane Mounir","email":"","orcid":"","institution":"Université Sultan Moulay Slimane","correspondingAuthor":false,"prefix":"","firstName":"Soufyane","middleName":"","lastName":"Mounir","suffix":""}],"badges":[],"createdAt":"2025-11-06 20:38:02","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8051212/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8051212/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":97137835,"identity":"45ec19c6-24e4-4766-aa44-60d9768e3189","added_by":"auto","created_at":"2025-12-01 09:58:13","extension":"json","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5571,"visible":true,"origin":"","legend":"","description":"","filename":"cd60cffc55604ca28b10cc19c2640f87.json","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/dde534ab3b531743ca7e15f7.json"},{"id":97136348,"identity":"71f4da1b-b20b-4d89-be20-d91ff0f92643","added_by":"auto","created_at":"2025-12-01 09:56:25","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":56858,"visible":true,"origin":"","legend":"","description":"","filename":"CoverletterArticle.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/91c91073e2ba49116550279e.pdf"},{"id":97137669,"identity":"218ad79b-c471-4291-a783-6ff14ab207ae","added_by":"auto","created_at":"2025-12-01 09:58:03","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2199382,"visible":true,"origin":"","legend":"","description":"","filename":"Houmairietal.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/0383ccfe0d0c2936a55bbb21.pdf"},{"id":97136083,"identity":"b0b8e836-a91d-4a3b-a6f0-83ebb7a0c74d","added_by":"auto","created_at":"2025-12-01 09:55:26","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":137538,"visible":true,"origin":"","legend":"","description":"","filename":"conjecturesviolationperlambda.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/afe22d09d7df398235ec0079.png"},{"id":96958680,"identity":"d7c41742-4438-455a-bdc0-644170031903","added_by":"auto","created_at":"2025-11-28 04:22:21","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":100752,"visible":true,"origin":"","legend":"","description":"","filename":"costcomparison.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/5067d2e86628f43a75da1e27.png"},{"id":96958681,"identity":"ffd80bcb-21c1-4bc3-8027-bc9ee79b822c","added_by":"auto","created_at":"2025-11-28 04:22:21","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":155245,"visible":true,"origin":"","legend":"","description":"","filename":"dualitygappercentage.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/e50e52a11bc27156525f86e3.png"},{"id":96958676,"identity":"45f632e9-976b-429c-a453-d997218da157","added_by":"auto","created_at":"2025-11-28 04:22:21","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":83919,"visible":true,"origin":"","legend":"","description":"","filename":"dualitygappercentagecorrected.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/20c1e987301362e5e4ad7c66.png"},{"id":97136964,"identity":"cd4b2567-a957-43ea-914f-08abc0afe490","added_by":"auto","created_at":"2025-12-01 09:57:13","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":125254,"visible":true,"origin":"","legend":"","description":"","filename":"figblockvslambda.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/23e5632190089f1bd78e8f9d.png"},{"id":97136160,"identity":"1058c04d-1ca4-4aba-a91a-028d76fd0af0","added_by":"auto","created_at":"2025-12-01 09:55:51","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":84233,"visible":true,"origin":"","legend":"","description":"","filename":"figblockvslambdaupto120.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/307227c961d5086e50d981f3.png"},{"id":96958686,"identity":"712f4867-e93a-43d1-a48a-73df4ae84964","added_by":"auto","created_at":"2025-11-28 04:22:21","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":143840,"visible":true,"origin":"","legend":"","description":"","filename":"figcostimprovementvslambda.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/9ebee641a749f38f3415edd4.png"},{"id":96958687,"identity":"6fa01c04-23e2-4584-a8d2-0f78ecef1b33","added_by":"auto","created_at":"2025-11-28 04:22:21","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":137337,"visible":true,"origin":"","legend":"","description":"","filename":"figcostvslambda.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/8a674c855c84ea9d2a16e975.png"},{"id":96958685,"identity":"16ad7954-8019-489e-a39f-586e376127c0","added_by":"auto","created_at":"2025-11-28 04:22:21","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":110358,"visible":true,"origin":"","legend":"","description":"","filename":"figcostvslambdaupto120.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/dc85acfb0f03fe15f5516015.png"},{"id":96958691,"identity":"d84551e8-f42e-4c6c-85a3-a2f303d11c18","added_by":"auto","created_at":"2025-11-28 04:22:21","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":135987,"visible":true,"origin":"","legend":"","description":"","filename":"figutilvslambda.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/cc918151798d13f46f54c878.png"},{"id":96958683,"identity":"f0943b3c-30a9-4ea0-8fb8-309c5f023126","added_by":"auto","created_at":"2025-11-28 04:22:21","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":158210,"visible":true,"origin":"","legend":"","description":"","filename":"figwaitvslambda.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/5ff688e61131aff8765be7fd.png"},{"id":97137120,"identity":"b7678504-bf44-4f14-8bd2-c39aa516f9ce","added_by":"auto","created_at":"2025-12-01 09:57:23","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":119873,"visible":true,"origin":"","legend":"","description":"","filename":"figwaitingtimevslambda.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/770b48838de7e22bbba7d737.png"},{"id":96958690,"identity":"d4979a76-76d9-42d0-94cf-381713028468","added_by":"auto","created_at":"2025-11-28 04:22:21","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":180253,"visible":true,"origin":"","legend":"","description":"","filename":"primalvsdualcost.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/ebd07911f84613a4c0d3284a.png"},{"id":96958694,"identity":"0e2e6a85-800a-452b-88ad-5be81827fcc9","added_by":"auto","created_at":"2025-11-28 04:22:22","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":97354,"visible":true,"origin":"","legend":"","description":"","filename":"primalvsdualcostcorrected.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/bc118906e315a5096a847d92.png"},{"id":96958696,"identity":"0ebaff99-7995-4412-b1ef-956fd33003ec","added_by":"auto","created_at":"2025-11-28 04:22:22","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":158635,"visible":true,"origin":"","legend":"","description":"","filename":"schurviolationperS.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/b413208dd80156a890c82f5e.png"},{"id":96958688,"identity":"718ccd29-8c53-4129-ac46-c6db5f6059a8","added_by":"auto","created_at":"2025-11-28 04:22:21","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":76349,"visible":true,"origin":"","legend":"","description":"","filename":"schurviolationratevsS.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/746d78b1284778b3c8c5d6f2.png"},{"id":96958689,"identity":"6d25f720-228f-4f3c-98b1-1e1e3116890b","added_by":"auto","created_at":"2025-11-28 04:22:21","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":58423,"visible":true,"origin":"","legend":"","description":"","filename":"systemarchitecture.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/3f8bec76e68e21a71f59438c.png"},{"id":96958693,"identity":"d25a67af-c8b2-4bfa-8ccd-9dd06028ef4e","added_by":"auto","created_at":"2025-11-28 04:22:22","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":124475,"visible":true,"origin":"","legend":"","description":"","filename":"waitingtimecomparison.png","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/ca7e67aac797dc4fc0611802.png"},{"id":96958695,"identity":"4f14e6ef-249c-4986-baf2-bce16cc9a238","added_by":"auto","created_at":"2025-11-28 04:22:22","extension":"xml","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":212437,"visible":true,"origin":"","legend":"","description":"","filename":"cd60cffc55604ca28b10cc19c2640f871structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1/ced7a88ad8ca176eb7c30b45.xml"},{"id":97248328,"identity":"86c83522-92b5-4641-884b-56b445046a98","added_by":"auto","created_at":"2025-12-02 12:53:39","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1946376,"visible":true,"origin":"","legend":"","description":"","filename":"Houmairietal.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8051212/v1_covered_e6fdac50-c357-4e91-9914-f75eb4214168.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Weakly Coupled MDP for Load Balancing in Containerized Cloud: A Scalable Control Design","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"cluster-computing","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Cluster Computing](https://www.springer.com/journal/10586)","snPcode":"10586","submissionUrl":"https://submission.nature.com/new-submission/10586/3","title":"Cluster Computing","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Containerized cloud, Load balancing, Dispatcher design, Queueing networks, Markov Decision Process (MDP), Weak coupling, Lagrangian relaxation","lastPublishedDoi":"10.21203/rs.3.rs-8051212/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8051212/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Load balancing is a core challenge in containerized cloud systems. We formulate the dispatcher’s decision problem as a weakly coupled Markov Decision Process (MDP) and derive a scalable policy via a Lagrangian linear-programming relaxation that decouples per-VM control. A toy-scale study is first used to expose structural behavior and guide design; we then run full-scale simulations to assess performance and robustness. We prove that the optimal value function is non-decreasing in total backlog and establish a stochastic-dominance corollary. Several appealing conjectures (e.g., per-VM monotonicity) hold at low demand but fail near saturation, clarifying when “balance everything” heuristics become suboptimal. Motivated by these insights, we propose a deployable load-aware dispatcher that blends JSQ-like behavior at low load with advantage-based routing at high load. Across scaled experiments, the dispatcher consistently reduces blocking and tail delay versus standard baselines, and competes favorably with a state-of-the-art heuristic, while incurring modest control overhead. The study bridges stochastic control and deployable cloud scheduling, offering both analytical insights and a practical policy design.","manuscriptTitle":"Weakly Coupled MDP for Load Balancing in Containerized Cloud: A Scalable Control Design","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-28 04:22:16","doi":"10.21203/rs.3.rs-8051212/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-05-14T10:33:50+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"215747670305128735443336818174670515162","date":"2026-05-06T15:01:44+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"57709042890288390880536476897560748614","date":"2026-05-04T07:23:09+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"165283507544769675958954215901334057962","date":"2025-11-24T15:34:18+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-11-24T11:33:20+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-11-24T08:29:33+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-11-10T14:29:39+00:00","index":"","fulltext":""},{"type":"submitted","content":"Cluster Computing","date":"2025-11-06T20:22:03+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"cluster-computing","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Cluster Computing](https://www.springer.com/journal/10586)","snPcode":"10586","submissionUrl":"https://submission.nature.com/new-submission/10586/3","title":"Cluster Computing","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"1fde1ae9-42ea-4a1a-bfd1-4543930d1d56","owner":[],"postedDate":"November 28th, 2025","published":true,"recentEditorialEvents":[{"type":"editorInvitedReview","content":"","date":"2026-05-14T10:33:50+00:00","index":89,"fulltext":""},{"type":"reviewerAgreed","content":"215747670305128735443336818174670515162","date":"2026-05-06T15:01:44+00:00","index":86,"fulltext":""},{"type":"reviewerAgreed","content":"57709042890288390880536476897560748614","date":"2026-05-04T07:23:09+00:00","index":81,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-11-28T04:22:16+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-28 04:22:16","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8051212","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8051212","identity":"rs-8051212","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}