Distributed SQL / Query federation

Query data.
Not copies.

A unified intelligence layer across warehouses, databases, lakes and streams. Trino plans the query, distributes the work and reaches the data where it already lives.

SELECT across 4 systems
DISTRIBUTED SQL
TRINO
PLAN · SCHEDULE · EXECUTE
Teradataenterprise warehouse
PostgreSQLoperational data
MinIO / Hivedata lake
Kafkastreaming data
01 / FEDERATEOne SQL surface

Query multiple catalogs in one statement.

02 / DISTRIBUTEParallel execution

Coordinator plans; workers execute stages.

03 / PUSH DOWNCompute near data

Reduce unnecessary movement to the query layer.

04 / GOVERNPolicy enforced

Ranger controls resources, rows and columns.

01 / Core Trino

One query.
Many systems.

Watch a federated query become a distributed execution plan. The point is not another data copy — it is one query layer over the estate.

QUERY / FEDERATED_01942LIVE · 3 CATALOGS · 4 WORKERS
Subscriber usage + complaints by segment
SELECT c.segment, COUNT(*) subscribers, SUM(u.usage_gb) usage
FROM teradata.crm.customers c
JOIN hive.usage.daily u ON c.subscriber_id = u.subscriber_id
JOIN postgres.care.complaints x ON c.subscriber_id = x.subscriber_id
GROUP BY c.segment;
TERADATA · crm.customers
HIVE · usage.daily
POSTGRES · care.complaints
Coordinator
Distributed plan
WORKER 01WORKER 02WORKER 03WORKER 04
02 / Teradata pushdown

Send the work.
Return the answer.

The custom connector story becomes visible here: supported predicates, aggregations, joins and window functions are pushed toward Teradata's AMP-parallel engine.

Teradata

AMP engine

Pushdown plan

Execute close to the data

WHEREGROUP BYSUM / COUNTJOINROW_NUMBERLAG / LEAD
ONLY RESULT PAYLOAD →
03 / Apache Ranger

Identity changes
the result.

Same table, same query, different governed view. Switch identities and watch masking and column access change immediately.

FA

Finance analyst

Revenue is visible. Subscriber identifiers are masked. Network-only attributes are restricted.

RESOURCE · catalog / schema / table / column
MASKING · user / group / role
ROW FILTER · policy injected
MSISDNREGIONREVENUECELL_ID
POLICY APPLIED · mask(msisdn) · allow(revenue) · restrict(cell_id)
04 / AI Intelligence

Ask naturally.
Execute properly.

AI sits on top of the query platform: prompt → context → SQL → Trino execution → governed result. It complements the engine; it does not replace the core Trino story.

Show active subscribers in Area X with more than 3 complaints in the last 90 days, grouped by plan type.
01 PROMPT02 CONTEXT03 SQL04 EXECUTE05 RESULT
-- intelligence layer ready

Generated SQL and execution state appear here.
05 / Connector estate

Connect the ecosystem.
Keep the data.

Live proof of concept

Your data estate.
One current.

Connect your own sources and demonstrate federation, distributed execution, pushdown, Ranger governance and natural-language querying as one platform.