Fast Heterogeneous Serving: Scalable Mixed-Scale LLM Allocation for SLO-Constrained Inference
This work addresses the challenge of scalable and cost-effective LLM deployment for cloud providers and enterprises, representing an incremental improvement over existing optimization methods.
The paper tackles the problem of efficiently allocating heterogeneous GPUs for large language model inference under strict latency, accuracy, and budget constraints, achieving over 260x speedup on large-scale instances while closely approaching optimal cost.
Deploying large language model (LLM) inference at scale requires jointly selecting base models, provisioning heterogeneous GPUs, configuring parallelism, and distributing workloads under tight latency, accuracy, and budget constraints. Exact mixed-integer linear programming (MILP) approaches guarantee optimality but scale poorly. We propose two constraint-aware heuristics: a Greedy Heuristic (GH) for single-pass allocation, and an Adaptive Greedy Heuristic (AGH) that enhances GH via multi-start construction, relocate-based local search, and GPU consolidation. Three constraint-aware mechanisms -- TP-aware feasibility selection, cost-per-effective-coverage ranking, and TP upgrade -- ensure feasibility under tightly coupled memory, delay, error, and budget constraints. On workloads calibrated with the Azure LLM Inference Trace (2025), both heuristics produce feasible solutions in under one second, with AGH closely approaching optimal cost while achieving over 260x speedup on large-scale instances. Under out-of-sample stress tests with up to 1.5x parameter inflation, AGH maintains controlled SLO violations and stable cost, whereas the exact solver's placement degrades sharply.