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-continuous-batching.en https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-continuous-batching.en Explains why static batching leaves the GPU idle, and how vLLM iteration-level scheduling removes that idle time. Covers the difference in character between prefill and decode, V1 unified scheduler and its token budget, and how chunked prefill affects latency and throughput, verified against the official documentation. Part 3 of the Inside vLLM series. Wed, 12 Aug 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) vllmcontinuous-batchingchunked-prefillllminference https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-continuous-batching.ja https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-continuous-batching.ja 静的バッチングがなぜGPUを遊ばせるのか、vLLMのイテレーション単位のスケジューリングがそれをどうなくすのかを整理した。prefillとdecodeの性質の違い、V1の統合スケジューラとトークン予算、chunked prefillがレイテンシと処理量に与える影響を公式ドキュメント基準で確認した。vLLM内部構造シリーズ第3回。 Wed, 12 Aug 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) vllmcontinuous-batchingchunked-prefillllminference https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-continuous-batching https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-continuous-batching 정적 배칭이 왜 GPU를 놀리는지, vLLM의 이터레이션 단위 스케줄링이 그것을 어떻게 없애는지 정리했습니다. prefill과 decode의 성격 차이, V1의 통합 스케줄러와 토큰 예산, chunked prefill이 지연과 처리량에 미치는 영향을 공식 문서 기준으로 확인했습니다. vLLM 내부 구조 시리즈 3편. Wed, 12 Aug 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) vllmcontinuous-batchingchunked-prefillllminference