EVERYTHING AIAI engineering, made visual
Visual edition planned
Phase 17 · Lesson 10Learn1 h28 lessons in phase

Cold Start Mitigation for Serverless LLMs

A 20 GB model image takes 5-10 minutes (7B) to 20+ minutes (70B) to go from cold to serving. In a true serverless world, that is not a warm-up — it is an outage. Mitigations operate at five layers: pre-seeded node images (Bottlerocket on AWS, dual-volume arch), model streaming (NVIDIA Run:ai Model Streamer, native in vLLM), GPU memory snapshots (Modal checkpoints, up to 10x faster restart), warm pools (`min_workers=1`), tiered loading (ServerlessLLM's NVMe→DRAM→HBM pipeline, 10-200x latency reduction), and live migration that moves input tokens (KB) rather than KV cache (GB). Modal publishes 2-4s cold starts as a floor; Baseten 5-10s default, sub-second with pre-warming. This lesson teaches you to measure, budget, and stack the five layers.

Visual edition planned

This lesson isn’t interactive yet.

It is part of the curriculum and will get the same treatment as Phase 1 — a visual cover, hands-on labs, derivations with numeric checks and a quiz. Until then, the original lesson is the best place to read it:

Part of Phase 17Infrastructure & Production. Use the previous / next cards below to keep browsing the phase.