This article presents a comprehensive overview of Iceberg MOR principles, Arctic‑based performance optimizations, benchmark evaluations using CH‑benchmark, and future roadmap items, highlighting how various file‑type strategies, self‑optimizing mechanisms, and task balancing improve real‑time data lake query efficiency.
This article describes how ByteDance tackled petabyte‑scale feature storage by adopting Apache Iceberg, detailing the problem background, design choices, implementation of COW and MOR back‑fill strategies, performance optimizations, and future plans such as lake‑cold‑layering and materialized views.
This article explains the rectangular layout of data‑center server rooms, the POD concept, and compares three common access‑switch architectures—EOR, MOR, and TOR—detailing their physical placement, cabling advantages, and suitability for modern high‑density, cloud‑driven environments.
This article examines the challenges of using Apache Hudi for real‑time data lake writes, analyzes the COW and MOR write models, and presents a custom master‑worker architecture with index optimization and repartitioning that reduces write resource consumption by over 85% while boosting throughput up to 300‑fold.
This article explains Apache Hudi's two table formats—Copy On Write and Merge On Read—by introducing key terminology, describing their file structures and versioning, comparing write and read latency, I/O cost, and write amplification, and concluding with guidance on choosing the appropriate format.
