← System overview

STORAGE / PROFESSIONAL

LogStore

Append-oriented storage with disk-backed indexes and bounded memory use.

01 / DESIGN

How LogStore stores data.

LogStore separates durable record bytes from lookup state. Checksummed mutations are appended to a data log, a disk-backed primary hash index maps item IDs to current offsets, secondary indexes remain on disk, and bounded caches accelerate hot values and index pages.

AVAILABILITYAvailable as the logstore storage provider, with log and ls aliases.
  1. 01Mutation frame
  2. 02Checksummed data log
  3. 03Disk primary index
  4. 04Bounded caches
  5. 05Checkpoint + tail replay

02 / TRADE-OFFS

Choose the engine for the workload.

Storage engines balance memory, I/O, recovery, concurrency and operational complexity. These profiles show where each engine fits.

STRENGTHS

  • Memory use does not grow linearly with record count
  • Sequential append path suits sustained writes
  • Fast point lookup without replaying the full log
  • Checkpoints, integrity checks and compaction support recovery

WEAKNESSES

  • More moving parts than replay storage
  • Disk-index tuning affects lookup efficiency
  • Compaction temporarily needs additional storage
  • Durability modes trade latency against the size of the loss window

03 / ABILITIES

What it provides.

LogStore combines an append-only, checksummed data log with disk-backed primary and secondary indexes. Bounded value and index-page caches keep memory use configurable as datasets grow, while checkpoints and tail replay provide recoverable startup.

01

Disk-backed primary and secondary indexes

02

Configurable sync, group and async durability

03

Bounded value and page caches

04

Checkpoint recovery, verification and compaction

BEST FIT

  • Large production datasets
  • Write-heavy workloads
  • Bounded-memory operation
  • Portable single-node deployments

CORE CAPABILITIES

  • Append-only checksummed data log
  • Disk-backed primary and secondary indexes
  • Bounded value and page caches
  • Checkpoint recovery and compaction

CONFIGURATION

  • MVDB_LOGSTORE_DATA_ROOT
  • MVDB_LOGSTORE_DURABILITY_MODE
  • MVDB_LOGSTORE_GROUP_COMMIT_RECORDS
  • MVDB_LOGSTORE_VALUE_CACHE_BYTES
  • MVDB_LOGSTORE_PRIMARY_PAGE_CACHE_BYTES
  • MVDB_LOGSTORE_PRIMARY_BUCKETS