lora admission control

1 article

Multi-LoRA Serving Capacity Planning: Working Sets and RAM

A 2026 guide to multi-LoRA serving capacity planning on vLLM: adapter working sets, max_loras vs max_cpu_loras, host RAM, cold-load and eviction effects on p99, and per-tenant admission control.

November 7, 2026• 30 min read
multi-lora serving capacity planningvllm loramax_loras