Skip to content

Startup Memory Estimate

HestiaStore logs a rough startup memory estimate after the effective SegmentIndex configuration is resolved. The report estimates memory pressure from configuration values; it is not a memory cap and it is not a measurement of allocated objects.

The estimate is meant to explain why the configured index may need memory. Use it as a sizing hint, then verify real usage with JVM and application monitoring.

How to read the report

The report uses fixed-width text tables with three columns:

  • Component: short name for the calculated area, input, or constant.
  • Estimate: estimated memory or configured value.
  • Description: formula, input context, or reason why an estimate is unknown.

Long descriptions wrap onto continuation rows so each report line stays within the fixed report width. The final estimate appears first, followed by the calculated memory areas:

  • Total index memory: all segments, chunk-store page cache, and maintenance overhead.
  • All segments: cached segment memory plus the segment routing map.
  • One cached segment: per-segment memory blocks that are multiplied by the configured segment cache slots.
  • Delta cache: configured maximum number of keys in the cache multiplied by the estimated key/value entry.
  • Chunk-store page cache: cached chunk-store pages and their key/value entries.
  • Maintenance overhead: temporary headroom for short-lived objects, maintenance work, and snapshots.
  • Key/value entry: key size, value size, and fixed per-entry overhead used by key/value cache estimates.
  • Key/position entry: key size, integer position size, and fixed per-entry overhead used by scarce-index estimates.
  • Key/segment-id entry: key size and segment id size used by the segment routing map.
  • Inputs and constants: configuration values and fixed assumptions used by the estimator. Values such as Write-buffer keys are reported for context and are not added to the total.

Area rows such as Delta cache, Chunk-store page cache, and Maintenance overhead summarize the formula in the description column. Supporting rows in Entry sizes and Inputs and constants show the entry-size assumptions and configuration values used by those formulas.

What the report uses

The estimator combines resolved configuration, descriptor size estimates, and a small set of fixed assumptions:

  • fixed per-entry overhead: applied when estimating key/value entries and key/position entries.
  • page overhead: applied per chunk-store page-cache page.
  • segment id: applied with key size when estimating key/segment-id route-map entries.
  • scarce index position: integer position stored with each scarce-index key.
  • fixed per loaded/cached segment: applied once for each loaded or cached segment.
  • maintenance overhead: extra space above memory before maintenance, calculated as the larger of a percentage of memory before maintenance or one delta cache.

These assumptions are deliberately visible because they are not exact JVM object-layout measurements. They are stable estimator constants used to make the report understandable and repeatable.

Incomplete estimates

If a key or value TypeDescriptor cannot provide getEstimatedAverageSizeInBytes(), HestiaStore still starts normally. The report marks dependent line items as unknown and still shows independent items such as Bloom filters and fixed segment infrastructure.