{"paper":{"title":"ADAPT: A Self-Calibrating Proactive Autoscaler for Container Orchestration","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"An online EWMA estimator of varying cold-start durations lets an MPC controller hold SLA violations below 5 percent across all tested workloads.","cross_cats":["cs.LG"],"primary_cat":"cs.DC","authors_text":"Himanshu Singh Baghel","submitted_at":"2026-05-15T09:46:43Z","abstract_excerpt":"Proactive autoscaling for containerized workloads depends on knowing the provisioning delay, i.e., the time between a scaling decision and the moment new capacity is ready to serve traffic. In practice, this cold-start duration can vary substantially across environments and even across consecutive scale-out events. We present ADAPT (Adaptive Duration Approximation for Predictive Timing), an online EWMA estimator that tracks coldstart duration at runtime. ADAPT feeds a dynamic planning horizon, FH-OPT, into a Model Predictive Controller (MPC) that optimizes replica counts over a rolling window."},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"MPC+LSTM achieves below 5% SLA violation on all workloads, compared with 7-19% for reactive HPA and up to 28.7% for MPC+Prophet on bimodal traffic.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that an online EWMA estimator can reliably track and adapt to varying cold-start durations across environments and consecutive scale-out events without additional sensors or external calibration.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ADAPT uses an EWMA estimator for cold-start durations to set a dynamic horizon in an MPC-based proactive autoscaler, achieving under 5% SLA violations with MPC+LSTM across tested workloads versus higher rates for HPA and MPC+Prophet.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"An online EWMA estimator of varying cold-start durations lets an MPC controller hold SLA violations below 5 percent across all tested workloads.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"170e69f396829d3b2bc08fdee2cdad8befe83e29f465ee0082789d718ce94f0c"},"source":{"id":"2605.15788","kind":"arxiv","version":1},"verdict":{"id":"e086140a-db79-4f57-af05-c29f56835d25","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T19:22:29.413991Z","strongest_claim":"MPC+LSTM achieves below 5% SLA violation on all workloads, compared with 7-19% for reactive HPA and up to 28.7% for MPC+Prophet on bimodal traffic.","one_line_summary":"ADAPT uses an EWMA estimator for cold-start durations to set a dynamic horizon in an MPC-based proactive autoscaler, achieving under 5% SLA violations with MPC+LSTM across tested workloads versus higher rates for HPA and MPC+Prophet.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that an online EWMA estimator can reliably track and adapt to varying cold-start durations across environments and consecutive scale-out events without additional sensors or external calibration.","pith_extraction_headline":"An online EWMA estimator of varying cold-start durations lets an MPC controller hold SLA violations below 5 percent across all tested workloads."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.15788/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-19T19:31:29.767515Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T19:31:19.106933Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T17:33:48.746470Z","status":"skipped","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T17:21:55.916094Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"2b3f1a9a41c646f271c0c032d294943d683097bafd71efdaa620ec8641853246"},"references":{"count":28,"sample":[{"doi":"","year":2024,"title":"Kubernetes Authors, “Horizontal Pod Autoscaling,” https: //kubernetes.io/docs/tasks/run-application/horizontal-pod-autoscale/, 2024, [Accessed 2026-05-01]","work_id":"e4e6cf93-c9a0-4caa-8e82-72ab6cd4eb92","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"AWS Lambda cold start latency — performance under load,","work_id":"ef1a304d-a3a9-487b-b526-b2c4a3682f36","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"An experimental evaluation of the Kubernetes cluster autoscaler in the cloud,","work_id":"64e37680-5dee-45d9-93fb-a45194fb7ce1","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"Machine learning-based scaling management for Kubernetes edge clusters,","work_id":"8c8d9cc9-19da-44f0-80eb-97c0d6519658","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Toward optimal load prediction and customizable autoscaling scheme for Kubernetes,","work_id":"636d4f63-18c3-4ca4-b990-b5e1e372813e","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":28,"snapshot_sha256":"5aceb401925d14c0d0dafd3941e0fd81cd46a7e169c9a81e8d51fc6e4e650d8d","internal_anchors":2},"formal_canon":{"evidence_count":2,"snapshot_sha256":"cb232750595321a1c2b5dd3fc772b7887346234287c6a6f47e5c71da6a64e264"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}