Mira storage engine.
Three layers, in dependency order:
- [
schema] — the OTAP-shaped Arrow schemas that are simultaneously the in-memory and the on-disk layout. There is no translation step. - [
identity] — the stable entity key that correlation joins on. - [
attrs] — the attribute tables and Resource-Scope preamble every signal shares. - [
signal] — the one shape every signal's encoder presents to the flusher. - [
logs] / [traces] / [metrics] — OTLP protobuf into those schemas, with block-local id rebasing. - [
block] — atomic publish of immutable block directories and zero-copy mmap reads back out of them.
See docs/architecture.md for why each of those is shaped the way it is.