Exo deep dive — Architecture

Control plane and data plane.

Exo runs the brain. Your cloud holds the data. A split-plane architecture that preserves sovereignty, eliminates egress, and deploys in days — not quarters.

Split-plane design

The brain and the body — separated by design.

Exo's control plane orchestrates from FalconDive infrastructure. Your data plane executes inside your cloud — data never crosses the boundary.

Exo split-plane architecture Control plane orchestrates from FalconDive infrastructure while the customer data plane executes inside their cloud account. EXO CONTROL PLANE · FalconDive-operated EXO UI Workspaces · Lakehouse · Compute · Pipelines · Govern · Observe API GATEWAY Authentication · Rate limiting · Request routing Query Catalog Cluster mgr Pipeline orch. RBAC Metrics SECURE BOUNDARY Orchestration commands Metadata + telemetry CUSTOMER DATA PLANE · Inside your cloud Compute clusters Spark / Trino workers in VPC Object storage S3 / GCS / Azure Data Lake Metadata store Iceberg / Hive metastore Identity provider OIDC / SAML directory
Why it matters

Four pillars of split-plane architecture.

Sovereignty

Your data never leaves your cloud. Satisfies GDPR, HIPAA, and RBI data-residency mandates by design — not by policy.

Economics

No egress charges, no surprise bills. Use your existing cloud discounts and reserved instances — Exo adds intelligence, not infrastructure.

Speed to value

Live in days, not quarters. A single Terraform template deploys the data plane — Exo connects, catalogs, and starts serving queries immediately.

Openness

Open formats, zero lock-in. Apache Iceberg tables, Hive-compatible metastore — any engine can read your data directly, no Exo dependency.

Exo brings the brain. You keep the data, the cloud, and the keys.

In practice

From SQL to result — six steps, two planes.

SELECT region, count(*), sum(total)
FROM prod.orders
WHERE region = 'APAC'
Life of an Exo query A SQL query flows through six steps across the control plane and data plane, from analyst input to result delivery.
  1. 01 Exo UI Analyst types SQL, hits Run Control plane
  2. 02 Catalog + RBAC Resolves table, checks column-level policy Control plane
  3. 03 Query service Builds execution plan, dispatches across boundary Control plane
  4. 04 Spark / Trino workers Execute in your VPC, push down filters Data plane
  5. 05 Iceberg on S3 / GCS / ADLS Stream Parquet from cloud buckets Data plane
  6. 06 Return Result rows return to the analyst's notebook Control plane
Product surfaces

One workbench, six surfaces.

Each module is a focused workspace — together they cover the full data lifecycle.

BUILD

Workspaces

For analysts

SQL editor, dashboards, collaborative scratchpads.

Lakehouse

For engineers

Tables, schemas, lineage, time-travel snapshots.

Compute

For platform teams

Cluster pools, autoscaling, spot-instance policies.

RUN & TRUST

Pipelines

For engineers

DAG builder, schedules, retries, backfills.

Govern

For security

Roles, row and column policies, audit trail.

Observe

For on-call teams

Query history, cost per job, SLA dashboards.

Ready to see the architecture in your cloud?