https://www.youngju.dev/blog 천천히 올바르게. AI Researcher & DevOps Engineer Youngju's tech blog. GPU/CUDA, LLM, MLOps, Kubernetes AI workloads, distributed training, and data engineering. ko fjvbn2003@gmail.com (Youngju Kim) fjvbn2003@gmail.com (Youngju Kim) Mon, 25 May 2026 00:00:00 GMT https://www.youngju.dev/blog/culture/2026-05-25-durable-workflow-temporal-inngest-trigger-hatchet-restate-dbos-2026-deep-dive.en https://www.youngju.dev/blog/culture/2026-05-25-durable-workflow-temporal-inngest-trigger-hatchet-restate-dbos-2026-deep-dive.en In 2026, workflow engines sit on top of two massive waves — first, long-running AI agents (LLM jobs that run for minutes to days) have effectively mandated durable execution; second, Postgres is becoming the backbone of message queues and workflow engines, a huge convergence. Temporal (Uber Cadence fork), Inngest, Trigger.dev v3, Hatchet, Restate, DBOS (Mike Stonebraker), AWS Step Functions, Cadence (Uber), Apache Airflow, Prefect, Dagster, Argo Workflows, Camunda, Conductor (Netflix). Durable execution semantics (event sourcing, deterministic replay, idempotency), saga patterns, exactly-once semantics, real cases from Toss, KakaoPay, Mercari, Rakuten. Which engine fits where. Mon, 25 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) durable-executionworkflow-enginetemporalinngesttrigger-devhatchetrestatedbosstep-functionsairflowprefectdagsterargo-workflowsevent-sourcingsaga-pattern https://www.youngju.dev/blog/culture/2026-05-25-durable-workflow-temporal-inngest-trigger-hatchet-restate-dbos-2026-deep-dive.ja https://www.youngju.dev/blog/culture/2026-05-25-durable-workflow-temporal-inngest-trigger-hatchet-restate-dbos-2026-deep-dive.ja 2026年のワークフローエンジンは二つの巨大な潮流の上に立っている。第一に、長時間実行のAIエージェント(LLMが数分から数日かかる作業)がデュラブル実行を事実上必須化した。第二に、Postgresがメッセージキューとワークフローエンジンのバックボーンになるという大きな合流である。Temporal(Uber Cadenceのフォーク)、Inngest、Trigger.dev v3、Hatchet、Restate、DBOS(Mike Stonebraker)、AWS Step Functions、Cadence(Uber)、Apache Airflow、Prefect、Dagster、Argo Workflows、Camunda、Conductor(Netflix)。デュラブル実行の意味論(イベントソーシング、決定論的リプレイ、冪等性)、サガパターン、厳密に一回の意味論、トス・カカオペイ・メルカリ・楽天の決済事例。どのエンジンをどこに使うかまで。 Mon, 25 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) durable-executionworkflow-enginetemporalinngesttrigger-devhatchetrestatedbosstep-functionsairflowprefectdagsterargo-workflowsevent-sourcingsaga-pattern https://www.youngju.dev/blog/culture/2026-05-25-durable-workflow-temporal-inngest-trigger-hatchet-restate-dbos-2026-deep-dive https://www.youngju.dev/blog/culture/2026-05-25-durable-workflow-temporal-inngest-trigger-hatchet-restate-dbos-2026-deep-dive 2026년 워크플로우 엔진은 두 가지 거대한 흐름 위에 서 있다 — 첫째는 장기 실행 AI 에이전트(LLM이 몇 분에서 며칠을 도는 작업)가 듀러블 실행을 사실상 강제했다는 점, 둘째는 Postgres가 메시지 큐와 워크플로우 엔진의 백본이 되어가는 거대한 합류다. Temporal(Uber Cadence 포크), Inngest, Trigger.dev v3, Hatchet, Restate, DBOS(Mike Stonebraker), AWS Step Functions, Cadence(Uber), Apache Airflow, Prefect, Dagster, Argo Workflows, Camunda, Conductor(Netflix). 듀러블 실행 의미론(이벤트 소싱, 결정론적 리플레이, 멱등성), 사가 패턴, 정확히 한 번 의미론, 토스·카카오페이·Mercari·Rakuten 결제 사례. 어느 엔진을 어디에 쓸지까지. Mon, 25 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) durable-executionworkflow-enginetemporalinngesttrigger-devhatchetrestatedbosstep-functionsairflowprefectdagsterargo-workflowsevent-sourcingsaga-pattern