DeepServe: SLO-Aware and Cost-Aware Elastic Scheduling for Serverless Multi-Tenant LLM Inference

DeepServe: SLO-Aware and Cost-Aware Elastic Scheduling for Serverless Multi-Tenant LLM Inference | 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 DeepServe: SLO-Aware and Cost-Aware Elastic Scheduling for Serverless Multi-Tenant LLM Inference Xuexian Li, Xiayuan Liu, Zilong Wang, Chun-Yao Hsieh, Yixue Liu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9317057/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Deploying large language model (LLM) inference services in serverless, multi-tenant environments present compounding challenges: cold start latency, GPU memory fragmentation, inter-tenant resource contention, and unpredictable tail latency. Existing systems optimize individual aspects but fail to jointly address service-level objective (SLO) compliance and cost efficiency under dynamic, heterogeneous workloads. We present DeepServe++, an elastic scheduling framework that formulates the joint SLO--cost optimization as a contextual bandit problem. The system introduces a Request Profiler that extracts online features---prompt length, historical KV-cache hit ratio, and predicted generation length---and feeds them into a contextual bandit agent that adaptively selects batch sizes, concurrency levels, KV-cache eviction policies, and warm-standby strategies. We evaluate DeepServe++ on ShareGPT and BurstGPT traces using LLaMA-2-13B and Mixtral-8x7B models on NVIDIA A100 GPUs. Results show that DeepServe++ reduces P99 latency by 38--62% compared to state-of-the-art baselines while improving GPU utilization by 14--23% and reducing per-request cost by up to 27%, with only a modest increase in scheduling overhead. LLM Serving Serverless Inference Scheduling Multi-Tenant KV-Cache Contextual Bandit Full Text Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted 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-9317057","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":617377761,"identity":"6d67ea8c-7a62-4213-812d-1197dacb5a1b","order_by":0,"name":"Xuexian Li","email":"","orcid":"","institution":"Northeastern University","correspondingAuthor":false,"prefix":"","firstName":"Xuexian","middleName":"","lastName":"Li","suffix":""},{"id":617377762,"identity":"df511316-6a52-42f6-b5f4-c0cfc4a2426c","order_by":1,"name":"Xiayuan Liu","email":"","orcid":"","institution":"Northeastern University","correspondingAuthor":false,"prefix":"","firstName":"Xiayuan","middleName":"","lastName":"Liu","suffix":""},{"id":617377763,"identity":"926ad254-a1da-4329-8302-5dfbe4dd8f62","order_by":2,"name":"Zilong Wang","email":"","orcid":"","institution":"New York University","correspondingAuthor":false,"prefix":"","firstName":"Zilong","middleName":"","lastName":"Wang","suffix":""},{"id":617377764,"identity":"3ae582b5-e936-446a-b795-6179f292232a","order_by":3,"name":"Chun-Yao Hsieh","email":"","orcid":"","institution":"Columbia University","correspondingAuthor":false,"prefix":"","firstName":"Chun-Yao","middleName":"","lastName":"Hsieh","suffix":""},{"id":617377765,"identity":"7b04a5bf-d779-4f7b-bc78-3f63a3ec9ce8","order_by":4,"name":"Yixue Liu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA30lEQVRIiWNgGAWjYDACZhBRIMHDxt98AMiSkCFSi4GNDJ/EsQSQFh4irTJIs5FjyDEAMQlrMTjOYybxweAwDxvDmc+vbtRY8DCwHz66Aa+Ww2xpkjNAWph7t1nnHAM6jCct7QZ+LczHpHnAtpzdZpzDBtQiwWNGQAtjm/QfsJacZ8Y5/4jSArQF6H2QFubHuW1EaJE8zJZs2WNgw8MmccyMObcPGEGE/MJ3/ozhjR8VEvby/c2PP+d8q5PjZz98DK8WhQMINpsEmMSnHATkGxBs5g+EVI+CUTAKRsHIBACM0EBzHb7aRAAAAABJRU5ErkJggg==","orcid":"","institution":"Carnegie Mellon University","correspondingAuthor":true,"prefix":"","firstName":"Yixue","middleName":"","lastName":"Liu","suffix":""}],"badges":[],"createdAt":"2026-04-04 02:59:27","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9317057/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9317057/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":106357638,"identity":"7d826a69-fa34-4820-acee-0cf5b94c4dd3","added_by":"auto","created_at":"2026-04-07 19:10:00","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":5161413,"visible":true,"origin":"","legend":"","description":"","filename":"DeepServeSLOAwareandCostAwareElasticSchedulingforServerlessMultiTenantLLMInference.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9317057/v1_covered_413d1f06-ea2c-4c6a-99e8-a4cfe801ea89.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eDeepServe: SLO-Aware and Cost-Aware Elastic Scheduling for Serverless Multi-Tenant LLM Inference\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"LLM Serving, Serverless Inference, Scheduling, Multi-Tenant, KV-Cache, Contextual Bandit","lastPublishedDoi":"10.21203/rs.3.rs-9317057/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9317057/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eDeploying large language model (LLM) inference services in serverless, multi-tenant environments present compounding challenges: cold start latency, GPU memory fragmentation, inter-tenant resource contention, and unpredictable tail latency. Existing systems optimize individual aspects but fail to jointly address service-level objective (SLO) compliance and cost efficiency under dynamic, heterogeneous workloads. We present DeepServe++, an elastic scheduling framework that formulates the joint SLO--cost optimization as a contextual bandit problem. The system introduces a \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003eRequest Profiler\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e\u0026nbsp;that extracts online features---prompt length, historical KV-cache hit ratio, and predicted generation length---and feeds them into a contextual bandit agent that adaptively selects batch sizes, concurrency levels, KV-cache eviction policies, and warm-standby strategies. We evaluate DeepServe++\u0026nbsp;on ShareGPT and BurstGPT traces using LLaMA-2-13B and Mixtral-8x7B models on NVIDIA A100 GPUs. Results show that DeepServe++\u0026nbsp;reduces P99 latency by 38--62% compared to state-of-the-art baselines while improving GPU utilization by 14--23% and reducing per-request cost by up to 27%, with only a modest increase in scheduling overhead.\u003c/strong\u003e\u003c/p\u003e","manuscriptTitle":"DeepServe: SLO-Aware and Cost-Aware Elastic Scheduling for Serverless Multi-Tenant LLM Inference","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-07 19:09:07","doi":"10.21203/rs.3.rs-9317057/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"f6c64390-3474-4982-b247-85e050ba465b","owner":[],"postedDate":"April 7th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-07T19:09:07+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-07 19:09:07","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9317057","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9317057","identity":"rs-9317057","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}