🧭 Architecture

This section explains how HestiaStore works under the hood: the core on‑disk structures, read/write data paths, caching and concurrency models, and the knobs that influence performance and reliability. Use these pages to reason about behavior, tune the system, and understand limits and trade‑offs.

• Sparse Index — structure and sampling that bound seeks into the main SST for fast point lookups.
• Segment Design — segment lifecycle, delta cache files, compaction, and when/why segments split.
• Data Block Format — chunked storage layout, block/cell sizes, and chunk headers.
• Write Path — API → buffer → flush to delta files → compaction/split; atomic commits with temp files and rename.
• Read Path — delta cache first, then Bloom + sparse index to a bounded on‑disk scan.
• Consistency & Recovery — crash‑safety model (no WAL), transactional writers, and check/repair steps.
• Filters & Integrity — magic number, CRC32, Snappy, and XOR pipelines on write/read.
• Chain of Filters — shared pipeline helper used by segment search and split flows.
• Caching Strategy — index write buffer, per‑segment caches (Bloom/scarce/delta), LRU, and warm‑up.
• Performance Model & Sizing — key tuning knobs, I/O patterns, and practical recipes.
• Concurrency — thread‑safe mode, per‑segment iteration safety, and process exclusivity.
• On‑Disk Layout & File Names — directory contents, file naming, and commit pattern.
• Limitations & Trade‑offs — hard limits, anti‑patterns, and recommended mitigations.
• Glossary — concise definitions for terms used across the architecture.