{ "schema_version": 1, "builder_plan": { "scenario": "Your inference service slows down when several customers send requests together.", "change": "Replay a recorded request trace with only the proposed serving policy changed.", "baseline": "Your current scheduler on the same model and hardware.", "steps": [ "Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.", "Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.", "Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.", "Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test." ], "measure": "Requests meeting your latency target, throughput, failures, and cost per successful request.", "decision": "Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation." }, "builder_review": null, "arxiv_id": "2609.29160", "title": "Cross-Model Autoscaling for Shared LLM Serving", "summary": "TRE replaces model-local reactive autoscaling (e.g., KV-cache-utilization signals) with a calibrated, demand-normalized Token Service Share metric that arbitrates replica capacity across co-hosted models under a fixed GPU budget, separating fast rescue from slower rebalancing. It costs a Kubernetes-based hot-switch serving stack and control-plane instrumentation, though it requires no changes to the inference scheduler and no extra GPUs. It can fail when TSS calibration does not transfer to a model's actual latency-SLO relationship, when hot-switching overheads dominate at small replica counts, or when the workload is a single model with no cross-model contention to arbitrate.", "rationale": "The method targets multi-model shared clusters with replica-level reallocation, which exceeds what a single 24 GB GPU can exploit, but the calibrated service-deficit signal is worth tracking if the reader later co-hosts multiple models.", "technique": "Cross-model token-service-share autoscaling for shared GPU serving", "family": "serving_e_batching", "recommendation": "observar", "published": "2026-09-24", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.29160", "independent_implementations": 1, "total_implementations": 2, "stars": 1, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "TRE replaces model-local reactive autoscaling (e.g., KV-cache-utilization signals) with a calibrated, demand-normalized Token Service Share metric that arbitrates replica capacity across co-hosted models under a fixed GPU budget, separating fast rescue from slower rebalancing. It costs a Kubernetes-based hot-switch serving stack and control-plane instrumentation, though it requires no changes to the inference scheduler and no extra GPUs. It can fail when TSS calibration does not transfer to a model's actual latency-SLO relationship, when hot-switching overheads dominate at small replica counts, or when the workload is a single model with no cross-model contention to arbitrate.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.29160", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Reduces P95 end-to-end latency by 11.9-79.0% and P99 by 12.5-72.6% versus a KV-cache-based reactive autoscaler on the same hot-switch runtime, across seven serving traces.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.29160", "source_page": null, "source_excerpt": "" } ], "exposure_map": [ { "dimension": "quality", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compute", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "latency", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "operations", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compatibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "security", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "data_and_training", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "reproducibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" } ], "risks": [], "minimum_test": [], "open_questions": [], "independent_tests": [], "lab_reproductions": [], "report_available": false, "validation_tier": "unknown", "evidence_tier": "unknown", "infrastructure_basis": "unknown", "equations": [], "equations_status": "not_fetched", "core_kind": "", "equations_fetched_at": "", "equations_source_url": "", "equations_source_sha256": "", "exposures_reviewed_at": "", "exposures_source_sha256": "" }