Skip to content
CSA Loom — the Microsoft Fabric experience for Azure tenants where Fabric isn't yet available: lakehouses, warehouses, notebooks, semantic models, Activator rules, Data Agents, across Commercial, GCC, GCC-High, and DoD IL5

Data Warehouse parity

What Fabric does

Fabric Warehouse runs on the Polaris distributed SQL engine — stateless, horizontally elastic, full T-SQL DML (INSERT / UPDATE / DELETE / MERGE) over Delta on OneLake. Claims up to 1024 concurrent queries; sub-second query SLA on warm data. Cross-warehouse and cross-lakehouse queries via 3-/4-part naming. AI Functions in T-SQL (sentiment, classification, translation, summarization) ship in March 2026.

CSA Loom parity design

Two surfaces by boundary:

Boundary Primary Why
Commercial / GCC Databricks SQL Warehouse (Serverless / Pro / Classic) Closest to Polaris's T-SQL DML over Delta + horizontal elasticity; Photon engine; aggressive query cache
GCC-High / IL4 / IL5 Synapse Serverless SQL Pool Databricks SQL Warehouse not available in usgovaz/usgovva; Synapse Serverless is always-on, no cold-start cost

Databricks SQL Warehouse (Commercial)

  • Photon engine; serverless tier scales automatically
  • Full T-SQL-equivalent on Delta: INSERT / UPDATE / DELETE / MERGE
  • Cross-table ACID via Delta transactions
  • Cross-catalog queries via three-level naming (UC) — equivalent to Fabric's 3-part naming
  • Disk + memory + result caches
  • Loom Console "Warehouse" pane embeds the Databricks SQL editor

Synapse Serverless SQL (Gov)

  • Read-only T-SQL over Delta and Parquet
  • External tables + views over ADLS Gen2 paths
  • Cross-warehouse queries via DEFINE / OPENROWSET patterns
  • No write path through Serverless (writes happen through Databricks classic Spark → Delta)
  • Loom Console "Warehouse" pane uses the existing Synapse Studio query editor embed

Materialized views

  • Commercial: Databricks SQL materialized views (auto-refresh on schedule or commit)
  • Gov: Synapse Serverless EXTERNAL TABLE over pre-aggregated Delta tables maintained by Databricks notebooks

AI Functions in SQL

Capability Commercial Gov
sentiment / classify / translate / summarize Databricks SQL ai_query() Databricks notebook UDF calling AOAI direct (gpt-4o in usgovvirginia)

ai_query is Commercial-only per research/02-gov-boundary- availability.md §7.1.

Copilot for Warehouse

NL→SQL via Loom Data Agents (see Data Agents parity); embedded in the Console Warehouse pane.

Per-boundary behavior

Boundary T-SQL DML over Delta Cross-warehouse joins ai_query
Commercial ✅ Databricks SQL Warehouse ✅ UC 3-level names ✅
GCC ✅ Databricks SQL Warehouse ✅ UC 3-level names ✅
GCC-High / IL4 ❌ (Synapse Serverless read-only; writes via Databricks Spark) ⚠ via external tables ❌ (notebook + AOAI direct)
IL5 (v1.1) Same as IL4 Same Same

Honest gaps

  • No sub-second cold-start in Synapse Serverless comparable to Polaris's claimed sub-second. Synapse Serverless has cold-start of seconds; Polaris claims sub-second.
  • ai_query in Gov: not available. Substitute via Databricks notebook + AOAI direct (slower; harder to invoke from a SQL query).
  • Cross-warehouse joins through 4-part names: Synapse Serverless supports cross-database joins inside one workspace; cross-workspace requires Synapse Link or explicit external table setup. Fabric's 4-part is more transparent.

Forward migration

  • Warehouse schemas migrate via T-SQL DDL → Fabric Warehouse CREATE TABLE
  • Delta data via OneLake shortcut (zero data movement)
  • dbt models — zero rewrite (dbt-fabric adapter mature)