Database (JouleDB)

A managed columnar/row hybrid database with energy as a first-class billing dimension. Postgres-wire compatible for the row workload, Arrow-native for the columnar workload, both on the same physical store.

Why a new database

Cloud databases bill by provisioned IOPS, instance hours, and storage GB. The actual cost of a query is energy — the work the planner makes the silicon do. JouleDB measures that and bills it. A query that scans a billion rows pays for scanning a billion rows. A query that hits a covering index pays for hitting an index.

Wire protocol

Postgres 16 wire-compatible. Bring psql, libpq, sqlx, pgx, prisma, drizzle — they connect.

psql "postgresql://jc_…@db.greenjoules.cloud:5432/main"
> SELECT * FROM users WHERE id = 42;
SELECT 1
joules: 0.0023   carbon-mg: 0.0005   plan-hash: ab12…

Every result includes a joules + carbon-mg footer. The portal exposes the same numbers as a slow-query log sorted by energy rather than time.

Columnar & analytics

For analytical workloads, JouleDB stores Arrow-IPC chunks in the same physical files as the rows. The planner picks columnar or row based on the query shape. SQL surface for analytics is DuckDB-compatible.

SELECT region, sum(revenue) FROM sales
WHERE date > '2026-01-01'
GROUP BY region;

Replication & placement

Pricing

Two components: (1) storage at-rest in joule-equivalents per GB-month (small — storage is ~0.05 J/GB-second at idle); (2) compute joules per query, billed live. There is no instance-hour charge.

Limits at launch

When to pick something else

If you need a fully-managed graph database, vector-only specialty store, or a multi-master leader-leader replication topology, prefer a specialist. JouleDB is a generalist OLTP+OLAP store with energy-aware billing; it isn't trying to win every category.