Research

Paper

AI LLM March 17, 2026

FleetOpt: Analytical Fleet Provisioning for LLM Inference with Compress-and-Route as Implementation Mechanism

Authors

Huamin Chen, Xunzhuo Liu, Yuhan Liu, Junchen Jiang, Bowei He, Xue Liu

Abstract

Modern LLM GPU fleets are provisioned for worst-case context lengths that the vast majority of requests never approach, wasting GPU capacity on idle KV-cache slots. We present FleetOpt, a framework that starts from first principles: given a workload's prompt-length CDF and a P99 TTFT target, derive the minimum-cost fleet analytically, then deploy it in practice. The analytical core models each pool as an M/G/c queue and derives that the minimum-cost fleet is a two-pool architecture -- a short-context pool and a long-context pool -- with an optimal boundary B* satisfying an equal marginal GPU cost condition across both pools. The fundamental barrier to achieving B* is the cost cliff: a hard routing step where requests just above B* consume 8x--42x more GPU capacity than requests just below it (depending on the context window ratio), creating a structural disincentive to lower the boundary. Compress-and-Route (C&R) is the implementation mechanism that resolves this barrier. Gateway-layer extractive compression trims borderline requests below B* before the engine ever sees them, converting the hard hardware boundary into a software parameter read from the workload CDF. The two components are unified in the FleetOpt offline planner: given a CDF and SLO, it returns the optimal (n_s*, n_l*, B*, gamma*) in under 1 ms. On three production traces, the combined framework reduces total GPU cost by 6--82% versus a homogeneous fleet, with C&R contributing 1--44 percentage points beyond plain pool routing depending on workload archetype. The analytical model is validated against a discrete-event simulator (inference-fleet-sim) with <= 3% error on predicted GPU utilization across all pools and workloads.

Metadata

arXiv ID: 2603.16514
Provider: ARXIV
Primary Category: cs.DC
Published: 2026-03-17
Fetched: 2026-03-18 06:02

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.16514v1</id>\n    <title>FleetOpt: Analytical Fleet Provisioning for LLM Inference with Compress-and-Route as Implementation Mechanism</title>\n    <updated>2026-03-17T13:41:21Z</updated>\n    <link href='https://arxiv.org/abs/2603.16514v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.16514v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Modern LLM GPU fleets are provisioned for worst-case context lengths that the vast majority of requests never approach, wasting GPU capacity on idle KV-cache slots. We present FleetOpt, a framework that starts from first principles: given a workload's prompt-length CDF and a P99 TTFT target, derive the minimum-cost fleet analytically, then deploy it in practice.\n  The analytical core models each pool as an M/G/c queue and derives that the minimum-cost fleet is a two-pool architecture -- a short-context pool and a long-context pool -- with an optimal boundary B* satisfying an equal marginal GPU cost condition across both pools. The fundamental barrier to achieving B* is the cost cliff: a hard routing step where requests just above B* consume 8x--42x more GPU capacity than requests just below it (depending on the context window ratio), creating a structural disincentive to lower the boundary.\n  Compress-and-Route (C&amp;R) is the implementation mechanism that resolves this barrier. Gateway-layer extractive compression trims borderline requests below B* before the engine ever sees them, converting the hard hardware boundary into a software parameter read from the workload CDF. The two components are unified in the FleetOpt offline planner: given a CDF and SLO, it returns the optimal (n_s*, n_l*, B*, gamma*) in under 1 ms.\n  On three production traces, the combined framework reduces total GPU cost by 6--82% versus a homogeneous fleet, with C&amp;R contributing 1--44 percentage points beyond plain pool routing depending on workload archetype. The analytical model is validated against a discrete-event simulator (inference-fleet-sim) with &lt;= 3% error on predicted GPU utilization across all pools and workloads.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.DC'/>\n    <published>2026-03-17T13:41:21Z</published>\n    <arxiv:comment>Work in progress</arxiv:comment>\n    <arxiv:primary_category term='cs.DC'/>\n    <author>\n      <name>Huamin Chen</name>\n    </author>\n    <author>\n      <name>Xunzhuo Liu</name>\n    </author>\n    <author>\n      <name>Yuhan Liu</name>\n    </author>\n    <author>\n      <name>Junchen Jiang</name>\n    </author>\n    <author>\n      <name>Bowei He</name>\n    </author>\n    <author>\n      <name>Xue Liu</name>\n    </author>\n  </entry>"
}