Architecture
How DPU is structured, and where each piece runs.
DPU is composed of a small number of layers that sit directly on top of your existing Snowflake environment.
Ingestion
Raw meter reads, customer records, and asset hierarchy data land as files (direct upload or Snowpipe auto-ingest) and are staged into raw landing tables. See Metering for what each domain looks like.
Data Enrichment
A configurable validation/estimation/editing (VEE) pipeline checks, estimates, and scales raw reads before they're promoted to the marts layer — see Data Enrichment.
Marts
The Core mart exposes modeled, query-ready dimension and fact tables for downstream consumers: BI tools, APIs, and other applications. A separate Consumption mart, plus optional per-client extensions like Rate Analysis and Load Research, sit alongside it — see Core.
Your upstream source systems
│
▼
Ingestion → Data Enrichment → Marts (Core, Consumption, extensions)
│
▼
Snowflake (your environment)Because DPU runs natively in your warehouse, there's no separate infrastructure to provision and no data egress step.