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) Sat, 16 May 2026 00:00:00 GMT https://www.youngju.dev/blog/culture/2026-05-16-feature-stores-feast-tecton-hopsworks-databricks-feldera-bytewax-materialize-2026-deep-dive.en https://www.youngju.dev/blog/culture/2026-05-16-feature-stores-feast-tecton-hopsworks-databricks-feldera-bytewax-materialize-2026-deep-dive.en A no-marketing tour of the 2026 feature store landscape: Feast (CNCF sandbox), Tecton (with the Eppo merger), Hopsworks, Vertex AI Feature Store, SageMaker Feature Store, Databricks Feature Engineering in Unity Catalog, Featureform, the streaming feature engines (Bytewax, Feldera, Materialize, ksqlDB, RisingWave), the transactional newcomers (Fennel — now inside Stripe — and Chalk), and the honest 2026 debate: do most teams actually need a separate feature store, or is dbt + warehouse + Redis enough? Online/offline split, point-in-time correctness, streaming freshness, embeddings convergence, and Korea/Japan adoption. Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) feature-storefeasttectonhopsworksdatabricksbytewaxmaterializestreamingmlopslakehouseflinkredis https://www.youngju.dev/blog/culture/2026-05-16-feature-stores-feast-tecton-hopsworks-databricks-feldera-bytewax-materialize-2026-deep-dive.ja https://www.youngju.dev/blog/culture/2026-05-16-feature-stores-feast-tecton-hopsworks-databricks-feldera-bytewax-materialize-2026-deep-dive.ja 2026年5月時点のフィーチャーストア市場を本気で整理する。Feast(CNCF サンドボックス)、Tecton(Eppo 一部資産統合)、Hopsworks、Vertex AI Feature Store、SageMaker Feature Store、Databricks Feature Engineering in Unity Catalog、Featureform、ストリーミング基盤の Bytewax/Feldera/Materialize/ksqlDB/RisingWave、決済領域に強い Fennel(Stripe 買収)と Chalk、そして「そもそも別物のフィーチャーストアは要るのか? dbt + ウェアハウス + Redis キャッシュで十分では?」という 2026 年の本音論争まで。オンライン/オフライン分離、ポイントインタイム整合性、ストリーミング鮮度、埋め込みとの収束、韓国・日本の導入事例まで一気通貫で扱う。 Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) feature-storefeasttectonhopsworksdatabricksbytewaxmaterializestreamingmlopslakehouseflinkredis https://www.youngju.dev/blog/culture/2026-05-16-feature-stores-feast-tecton-hopsworks-databricks-feldera-bytewax-materialize-2026-deep-dive https://www.youngju.dev/blog/culture/2026-05-16-feature-stores-feast-tecton-hopsworks-databricks-feldera-bytewax-materialize-2026-deep-dive 2026년 5월 기준 피처 스토어 시장을 끝까지 본다. Feast(CNCF 샌드박스), Tecton(Eppo 인수 통합), Hopsworks, Vertex AI Feature Store, SageMaker Feature Store, Databricks Feature Engineering in Unity Catalog, Featureform, Bytewax/Feldera/Materialize/ksqlDB/RisingWave 같은 스트리밍 피처 엔진, Fennel(Stripe 인수)·Chalk의 트랜잭셔널 피처 스토어, 그리고 "정말 피처 스토어가 필요한가? dbt + 웨어하우스 + Redis 캐시면 충분한가?"라는 2026년의 진짜 논쟁까지. 온라인/오프라인 스토어 분리, 포인트인타임 정합성, 스트리밍 신선도, 임베딩 통합, 한국·일본 도입 사례까지 한 글에서 정직하게 정리한다. Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) feature-storefeasttectonhopsworksdatabricksbytewaxmaterializestreamingmlopslakehouseflinkredis https://www.youngju.dev/blog/culture/2026-05-16-stream-processing-2026-kafka-streams-flink-sql-risingwave-materialize-arroyo-ksqldb-bytewax-decodable-deep-dive.en https://www.youngju.dev/blog/culture/2026-05-16-stream-processing-2026-kafka-streams-flink-sql-risingwave-materialize-arroyo-ksqldb-bytewax-decodable-deep-dive.en The 2026 stream processing landscape. Flink 2.0 broke the library-vs-DB boundary with Disaggregated State and Materialized Tables; Kafka Streams 4.0 absorbed queue semantics via Share Groups; RisingWave 2.x and Materialize compute incremental views in real time behind the Postgres wire. Arroyo reimplemented Flink SQL in Rust; Bytewax pulled streaming into the PyData ecosystem. We cover ksqlDB, Apache Beam, Pulsar Functions, Quix, Decodable, Aiven Flink, Confluent Cloud Flink, Estuary Flow, Memphis, Redpanda Connect — paradigms, state management, windowing, exactly-once, CEP, Korean/Japanese adoption. Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) stream-processingkafka-streamsflinkrisingwavematerializearroyoksqldbbytewaxdecodablereal-time2026deep-diveenglish https://www.youngju.dev/blog/culture/2026-05-16-stream-processing-2026-kafka-streams-flink-sql-risingwave-materialize-arroyo-ksqldb-bytewax-decodable-deep-dive.ja https://www.youngju.dev/blog/culture/2026-05-16-stream-processing-2026-kafka-streams-flink-sql-risingwave-materialize-arroyo-ksqldb-bytewax-decodable-deep-dive.ja 2026年のストリーム処理地図。Flink 2.0は Disaggregated State と Materialized Tables でライブラリと DB の境界を崩し、Kafka Streams 4.0 は Share Groups でキューも吸収した。RisingWave 2.x と Materialize は Postgres ワイア上で増分ビューをリアルタイムに計算する。Arroyo は Flink SQL を Rust で書き直し、Bytewax は Python で PyData 生態系にストリームを引き込んだ。ksqlDB、Apache Beam、Pulsar Functions、Quix、Decodable、Aiven Flink、Confluent Cloud Flink、Estuary Flow、Memphis、Redpanda Connect まで — パラダイム、状態管理、ウィンドウ、exactly-once、CEP、韓国/日本の採用事例を一気通貫で整理する。 Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) stream-processingkafka-streamsflinkrisingwavematerializearroyoksqldbbytewaxdecodablereal-time2026deep-dive日本語 https://www.youngju.dev/blog/culture/2026-05-16-stream-processing-2026-kafka-streams-flink-sql-risingwave-materialize-arroyo-ksqldb-bytewax-decodable-deep-dive https://www.youngju.dev/blog/culture/2026-05-16-stream-processing-2026-kafka-streams-flink-sql-risingwave-materialize-arroyo-ksqldb-bytewax-decodable-deep-dive 2026년 스트림 처리 지형도. Flink 2.0이 Disaggregated State와 Materialized Tables로 라이브러리·DB 경계를 허물고, Kafka Streams 4.0은 Share Groups로 Queue까지 흡수했으며, RisingWave 2.x와 Materialize는 Postgres 와이어 위에서 incremental view를 실시간으로 계산한다. Arroyo는 Rust로 Flink SQL을 재구현했고 Bytewax는 Python으로 스트림을 PyData 생태계에 끌어들였다. ksqlDB, Apache Beam, Pulsar Functions, Quix, Decodable, Aiven Flink, Confluent Cloud Flink, Estuary Flow, Memphis, Redpanda Connect까지 — 패러다임·상태 관리·윈도잉·exactly-once·CEP·한국/일본 사례를 한 번에 정리한다. Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) stream-processingkafka-streamsflinkrisingwavematerializearroyoksqldbbytewaxdecodablereal-time2026deep-dive