https://www.youngju.dev/blog 천천히 올바르게. AI Researcher & DevOps Engineer Youngju's blog. GPU/CUDA, LLM, MLOps, Kubernetes AI workloads, and data engineering — plus mindset essays on confidence, routines, health, and sport psychology. ko fjvbn2003@gmail.com (Youngju Kim) fjvbn2003@gmail.com (Youngju Kim) Wed, 12 Aug 2026 00:00:00 GMT https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-scheduler-and-preemption.en https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-scheduler-and-preemption.en Explains what the vLLM scheduler decides on every step, how the waiting queue and running list move, and how preemption, triggered when the KV cache runs short, wrecks latency and throughput, verified against the official documentation and the V1 scheduler source. Also covers the difference between recompute and swap, and the fcfs and priority policies. Part 4 of the Inside vLLM series. Wed, 12 Aug 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) vllmschedulerpreemptionkv-cachellm https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-scheduler-and-preemption.ja https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-scheduler-and-preemption.ja vLLMスケジューラが毎ステップ何を決定しているのか、待機キューと実行リストがどう動くのか、KVキャッシュが不足したときに起きるプリエンプション(preemption)がレイテンシと処理量をどう崩すのかを、公式ドキュメントとV1スケジューラソースで確認して整理した。recomputeとswapの違い、fcfsとpriorityポリシーも合わせて扱う。vLLM内部構造シリーズ第4回。 Wed, 12 Aug 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) vllmschedulerpreemptionkv-cachellm https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-scheduler-and-preemption https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-scheduler-and-preemption vLLM 스케줄러가 매 스텝 무엇을 결정하는지, 대기 큐와 실행 큐가 어떻게 움직이는지, KV 캐시가 모자랄 때 일어나는 선점(preemption)이 지연과 처리량을 어떻게 무너뜨리는지 공식 문서와 V1 스케줄러 소스로 확인해 정리했습니다. recompute와 swap의 차이, fcfs와 priority 정책도 함께 다룹니다. vLLM 내부 구조 시리즈 4편. Wed, 12 Aug 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) vllmschedulerpreemptionkv-cachellm