Towards Load-Aware Prefill Deflection for Disaggregated LLM Serving (arXiv 2607.02043)

Fields

id

sources/sched-kairos-arxiv

URL

https://arxiv.org/abs/2607.02043

Publisher

arXiv

Kind

paper

Published

2026-07

Accessed

2026-10-04

Archived URL

https://arxiv.org/html/2607.02043v1

notes

Microsoft Research paper on prefill deflection for disaggregated serving. Key finding: prefill execution is only 2-23% of P95 TTFT; queue wait and KV transfer dominate. Kairos reduces P95 TTFT by up to 81%. The arXiv HTML was fetched and verified.

Notes

Microsoft Research paper on prefill deflection for disaggregated serving. Key finding: prefill execution is only 2-23% of P95 TTFT; queue wait and KV transfer dominate. Kairos reduces P95 TTFT by up to 81%. The arXiv HTML was fetched and verified.

Cross-references

This record references no other records.