01
Columnar object storage
Write Parquet to S3, GCS, Azure, or MinIO, with Postgres tracking organizations, streams, schemas, and files.
Product · Unified telemetry
Four signal families. One operational context.
Keep logs, metrics, traces, and continuous profiles under one tenant and time model. Columnar telemetry shares one query plane while lossless pprof archives stay portable on object storage.
- telemetry signal families
- 4
- object-storage layer
- 1
- vectorized query engine
- 1
- tenant context
- 1
Shared where it matters. Specialized where it helps.
Each signal keeps an operator-friendly view while tenant, time, object storage, and investigation context stay common.
02
SQL across signals
Use joins, CTEs, and window functions through DataFusion when an incident does not fit a single telemetry type.
03
Signal-native exploration
Logs, PromQL metrics, trace waterfalls, profiling flame graphs, service maps, and dashboards remain optimized for their own job.
04
Planner-enforced tenancy
Organization filters are rewritten into query plans rather than delegated to naming conventions.
workflow
A single context from ingest to investigation.
The data plane carries the information the operator would otherwise reconstruct by hand.
- 101 · Ingest
Accept the protocol already in production
OTLP, Prometheus, Loki, Elasticsearch bulk, Syslog, pprof, Pyroscope, and cloud drains enter one governed data plane.
- 202 · Store
Persist columnar data on your object store
WAL and Arrow buffers flush into Parquet, with Tantivy and metadata supporting selective reads.
- 303 · Query
Use SQL or PromQL for the question
Choose the surface that fits the signal without moving data to another backend.
- 404 · Correlate
Carry tenant, time, and entity context
Move between signal views with the constraints of the investigation intact.
proof / shipped
The shared layer is an implementation fact.
The public architecture and API expose the same cross-signal model described here.
- Parquet 59 and object_store 0.13
- Lossless pprof + zstd archives
- DataFusion 54 and Arrow Flight 59
- Postgres metadata and planner-level org rewrite
Product · Unified telemetry
Four signal families. One operational context.
Keep logs, metrics, traces, and continuous profiles under one tenant and time model. Columnar telemetry shares one query plane while lossless pprof archives stay portable on object storage.