docs(blog): Async Compaction — Background Maintenance as an Architectural Primitive (MOR series part 4)

[nsivabalan]

Summary

Fourth post in the MOR-as-architectural-shift blog series.

• Argues that async compaction (introduced in Hudi in 2019) is not a performance tweak layered onto ingestion, but a direct consequence of separating correctness from optimization — the same principle mature database engines (PostgreSQL VACUUM, RocksDB/LevelDB/Cassandra SSTable compaction, online index rebuilds, checkpointing) have followed for decades.
• Traces the shift in lakehouse workloads (CDC, streaming, feature stores, AI pipelines) that makes background maintenance an architectural requirement rather than a periodic batch job.
• Sets up the next post in the series: a head-to-head benchmark of Apache Hudi, Apache Iceberg, and Delta Lake under continuous-ingestion + concurrent maintenance workloads, measuring sustained forward progress rather than raw throughput/latency.

Series so far:

1. MOR Isn't a Storage Optimization. It's an Architectural Shift
2. Why Metadata Has to Be Mutation-Friendly
3. Building Indexes on a Moving Target
4. Async Compaction: Background Maintenance as an Architectural Primitive (this PR)

Tags

All tags used (mor, merge on read, compaction, architecture, data lakehouse, metadata, clustering) are already in the existing allowlist — no .github/scripts/validate-blog.py change needed.

Test plan

• `python3 .github/scripts/validate-blog.py website/blog/2026-07-17-async-compaction-background-maintenance.mdx` passes
• Local site preview renders frontmatter, cover image, and all three inline diagrams at high resolution
• Tag filter bar surfaces the post under existing tags (`mor`, `merge on read`, `compaction`, `architecture`, `data lakehouse`, `metadata`, `clustering`)
• Series navigation links to prior three posts and back-references from those posts (if updated) resolve correctly

🤖 Generated with Claude Code

[hudi-agent]

⚠️ 🤖 This review was generated by an AI agent and may contain mistakes. Please verify any suggestions before applying.

No reviewable code files in this PR.

cc @yihua

[nsivabalan]

[wombatu-kun]

The LevelDB reference links to a facebook/rocksdb wiki page instead of LevelDB documentation. Point it at LevelDB's own docs (github.com/google/leveldb) or relabel the link as RocksDB.