# Push-based KV transfer to improve TTFT and TPOT for disaggregated inference in vLLM

vLLM's prefill-decode disaggregation splits LLM inference across engine instances but must transfer the KV cache between them. This post introduces a push-based NIXL connector that moves transfer coordination off the critical path, delivering up to 3x better TTFT and 30% better T



## 질문과 답변



## 출처
- AWS Machine Learning Blog: https://aws.amazon.com/blogs/machine-learning/push-based-kv-transfer-to-improve-ttft-and-tpot-for-disaggregated-inference-in-vllm/ (발표 2026-10-19T16:05:17.000Z)

작성: TUMN 큐레이션 · AI 보조: false
발행: 2026-10-19T16:05:17.000Z
확인: 2026-10-19T21:25:34.296Z
검토: 원문 제목과 공개된 발췌를 인용하고 원문으로 연결합니다. 내용을 재작성하거나 요약하지 않았습니다.
정식 URL: https://api.tumn.com/news/6f26adbc-8f70-4421-971f-8e63d9a97401