Energy & utilities
Grid and meter data in one engine — interval reads at scale, windowed KPIs, and current grid state in constant time.
The problem
Why this is hard today
Smart-meter interval reads and grid sensor data are a firehose with a long memory: billing needs years of durable history, operations needs the state of the grid right now, and both read the same streams.
Splitting those needs across a historian, a cache, and a warehouse multiplies schemas and makes the operational questions — which feeders are overloaded, which meters went dark — depend on pipeline lag.
The fit is line-rate durable ingest with retention policies, keyed current-state per meter or feeder as O(1) reads, and windowed aggregation running continuously in the same SQL.
Where NYXDB fits
Use-case journeys
AMI interval ingest
Absorb meter-read bursts durably with backpressure; history stays queryable for billing.
Time seriesGrid observability
Live per-feeder KPI tables maintained by continuous transforms.
Observability & logsOperations dashboards
Control-room views read keyed state tables and subscribe to live deltas.
Real-time dashboardsArchitecture
How NYXDB fits grid data
Interval reads append durably at line rate; TTL predicates age raw readings out on schedule while windowed rollups persist; keyed tables answer per-feeder and per-meter state questions in constant time.
- 01
Durable ingest
WAL-backed line-rate appends, governed under pressure.
- 02
Retention
TTL predicates age raw interval reads out on schedule (ADR-019).
- 03
Rollups
Continuous transforms maintain hourly/daily windowed tables.
- 04
Grid state
Keyed tables keep the latest per (feeder, metric) — O(1) reads.
Real SQL
Representative queries
SELECT feeder_id, load_mw, rated_mwFROM feeder_stateWHERE load_mw > rated_mw * 0.95;Every statement follows the engine’s own test SQL shapes. See the SQL reference for full syntax.
Capabilities
What you get
Line-rate ingest
Durable, governed, WAL-backed.
TTL retention
Raw reads age out; rollups persist.
O(1) grid state
Latest per feeder/meter, instantly.
Substation to cloud
One 154MB image runs anywhere.
Proof
Measured on the vetted benchmark lane
durable meter-read ingest, single core
1M-row batches, best durable mode
vetted table
View benchmark~1510×skip-index pruning (minmax, 16 parts)
zero disk bytes on pruned reads
vetted table
View benchmark268nscurrent grid state, keyed count()
O(1), flat 1k–16k keys
vetted table
View benchmarkMeasured on AMD EPYC 9554P 64-Core Processor (64t), 1505GB RAM, 6x NVMe — measured on server-class hardware. Release build, median of 5 (re-captured stragglers: median of 3), commit-pinned (7945ae33, 2026-07-09). Ingest figures are engine-side. See the full benchmark suite.
Learn more
Related documentation
One engine from substation to billing
Explore the time-series solution or run a node.