Skip to main content
Version: Next

Glossary

See also:

Core data and storage terms​

Series​

A logical time series identified by:

  • A series string (for example weather.room1)
  • Optional tags (a JSON object) that become part of the series identity in HTTP surfaces

See also:

series_id​

A stable u64 identifier derived from series (and tags) using a hash.

See:

Point​

A single data point: timestamp + value. Most storage structures ultimately store a sequence of points per series.

Implementation: types.Point.

WAL (write-ahead log)​

An append-only log used for crash safety. SydraDB records ingested points to the WAL before making them durable in segment files.

See also:

Memtable​

An in-memory buffer that holds recent points before they are flushed to disk.

See also:

Segment​

An on-disk file containing points for a single series within a time bucket (currently hour-aligned). Range queries scan relevant segments for the requested time window.

See also:

Manifest​

An on-disk index of which segment files exist and what time ranges they cover. It is used for segment selection during reads and for “highwater” timestamps during WAL recovery.

See also:

Compaction​

A background process that rewrites and merges segment files to reduce redundancy and improve read efficiency (for example: de-duplicating points with the same timestamp).

See also:

Retention​

A best-effort deletion pass that removes old segment files beyond a configured TTL (retention_days).

See also:

Query and compatibility terms​

sydraQL​

SydraDB’s query language and query execution pipeline. It is exposed over HTTP at POST /api/v1/sydraql.

See also:

pgwire / PostgreSQL compatibility​

An optional surface that aims to support PostgreSQL clients and ORMs over the PostgreSQL v3 wire protocol by translating SQL and emulating catalog/introspection.

See also: