Fast Serverless LLM Cold Starts by Sharing and Prewarming Startup Work

When a serverless LLM platform launches a model worker on demand, the cold start lasts roughly a minute, yet only 5-11% of it is spent loading weights. Most of the remainder goes to compilation, memory profiling and runtime setup, which each server performs again even though the outcome is identical across servers or independent of the model. We present Cairn. The server that is first to start a model publishes the weights and engine artifacts to a CXL memory pool shared across the cluster, and the remaining servers read them in place. Model-independent setup is completed ahead of time inside warm containers, and a template process replaces each container as soon as it is used. With ServerlessLLM on vLLM v0.20, six models from 1B to 14B, and two servers that share an emulated CXL pool, Cairn speeds up cold starts by 4.1-5.2x on servers without a checkpoint of their own and reduces median TTFT on a request stream by 64%.

Authors

Publication Details

Journal
arXiv (Cornell University)
Published
2026-11-12
DOI
https://doi.org/10.48550/arxiv.2611.16180
Primary Topic
Cloud Computing and Resource Management
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Fast Serverless LLM Cold Starts by Sharing and Prewarming Startup Work

Jaehyuk Huh, Kwanghoon Choi, Igjae Kim, Minho Ha et al.
arXiv (Cornell University)
Cloud Computing and Resource Management
preprint

Fast Serverless LLM Cold Starts by Sharing and Prewarming Startup Work

Jaehyuk Huh, Kwanghoon Choi, Igjae Kim, Minho Ha, Hongjin Kim
preprint en

Abstract

When a serverless LLM platform launches a model worker on demand, the cold start lasts roughly a minute, yet only 5-11% of it is spent loading weights. Most of the remainder goes to compilation, memory profiling and runtime setup, which each server performs again even though the outcome is identical across servers or independent of the model. We present Cairn. The server that is first to start a model publishes the weights and engine artifacts to a CXL memory pool shared across the cluster, and the remaining servers read them in place. Model-independent setup is completed ahead of time inside warm containers, and a template process replaces each container as soon as it is used. With ServerlessLLM on vLLM v0.20, six models from 1B to 14B, and two servers that share an emulated CXL pool, Cairn speeds up cold starts by 4.1-5.2x on servers without a checkpoint of their own and reduces median TTFT on a request stream by 64%.

arXiv (Cornell University)
Cloud Computing and Resource Management
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon