E6DATA QUERY ENGINE

More answers
per dollar.
Same performance.

Q: Do we have to migrate our data?

No. e6data reads the open tables you already have, in your object storage, through your catalog.

Q: Will our SQL run unchanged?

The measured share of queries that run unchanged will be published with the method. Where a dialect differs, the statement is rewritten and the rewrite is explained.

Q: Does our data leave our account?

Not with in-VPC deployment: compute runs next to your tables. For serverless, we document the exact data flow before you decide.

Q: How is this different from scaling our warehouse?

You pay per query, not for the size you picked, and it stays fast at high concurrency instead of queuing behind a coordinator.

Q: Your benchmarks look good. Why trust them?

Do not take them on faith. Ask us for a pilot on your own workload. We put the target savings in writing. No proof, no contract.

Q: Where does e6data not fit?

Workloads dominated by 10 TB+ batch reports, or long join-heavy finance reports. We will tell you up front.

Q: Can we run it alongside Databricks or Snowflake?

Yes. Move one workload at a time, on the same tables. Reverse it if you want.

Q: How do we buy?

AWS or Azure Marketplace, against existing commitments.

Q: Is it ready for AI agents?

Built for high-concurrency, short queries, which is what agents send. Production proof is 1,000+ QPS under 2s at p95; an agent-specific benchmark is in progress.

The agent-ready query engine with compute-storage separation for teams with ambitious data needs that charges for what you use, not a provisioned size. Sustains high velocity under high concurrency, on your existing tables, SQL, and platform governance.

Book a demo Read the docs

Up to 50% lower TCO, at the same or better performance.

WHY IT MATTERS

Stop paying for the warehouse size you guessed.

Query bills grow faster than analytics. Warehouses are sized for the peak and paid for while idle, so teams often ration dashboards, delay new use cases, and jump size tiers to scale. Agents asking thousands of questions an hour make it worse.

Up to 2–3x

more performance per dollar

Up to 50% lower TCO, at the same or better performance.

1,000+

queries per second, steady

Sustained in production, under 2s at p95. Low latency that holds as concurrency climbs.

migrations or rewrites

Same tables, same SQL, same governance. Point the engine at what you already have.

Results from customer production environments; conditions on request.

WHERE IT FITS

Drops into the stack you already run.

Compute you can deploy and run anywhere, in-VPC, on-prem, or any cloud in your lakehouse: no changes to data, queries or governance. Connect to your analytics layer and point to your existing tables and catalog, one workload at a time.

DEPLOYS IN YOUR ENVIRONMENT AND PERIMETER

YOUR ANALYTICS LAYERunchanged

BI dashboardsNotebooksdbtAppsAI agents

E6DATA QUERY ENGINEyour workloads with the highest ROI

Postgres wireNo new driver1 vCPU at a time

YOUR CURRENT ENGINEthe rest of your workloads

Databricks SQL

SnowflakeTrinoAthena

YOUR OPEN TABLES AND CATALOGin your object storage

Iceberg

Delta Lake

Parquet

Glue

Unity Catalog

Polaris

Mixed, multi-engine compute is a normal part of leading data platforms.

WHAT IT IS FOR

Built for mission critical workloads.

Each one a place where a warehouse is sized for the peak and paid for while idle.

Lakehouse speed inside your perimeter

Elastic, kubernetes-native compute to deploy in-VPC, sovereign, or air-gapped. Same catalog, security, and governance.

Runs in your accountsovereign and air-gapped environments: confirm the docs before launch

Internal BI and scheduled dashboards

Run frequent, high-volume scheduled analytics and rapid refresh cycles without warehouse cost overruns.

Up to 50% lower TCOat the same or better performance

Embedded analytics

Serve multi-tenant analytics inside your SaaS product at high concurrency, at a fraction of the cost.

15M+ queries a dayNasdaq-listed SaaS, 60% cheaper than Databricks, 80%+ cheaper than Snowflake

Interactive analytics

Real-time dashboards and exploration that stay fast while hundreds of people and tiles ask at once.

under 2s at p95at 1,000+ queries per second, in production

Always-on agent analytics

Serve agents that ask thousands of short, bursty questions an hour, at a cost per query that makes always-on monitoring practical.

1,000+ QPSproduction proof; no agent benchmark yet

Vector search

Semantic lookups on the same tables as your SQL, with built-in cosine similarity, so there is no separate vector store to sync, govern and pay for.

One engine, one set of tableshybrid query latency at concurrency: confirm before claiming

WHAT YOU GET

Do more with every dollar

Run better, at lower cost.

MEDIAN QUERY · TPC-DS 1 TB · DELTA LAKE ON S3 · WARM8 Oct 2026 · single stream

e6data · Delta1.3s

Databricks · Delta3.02s

EMR Trino · Delta4.62s

Starburst Galaxy · Delta6.02s

An atomic, distributed architecture removes the typical central coordinator, so parallel queries fan out independently to achieve high velocity sustained at high concurrency.

Median query on 1 TB TPC-DS, Delta Lake tables on S3: 1.3s8 Oct 2026 · Delta Lake on S3, AWS us-east-1 · same tables for every engine · one query at a time · warm · result caching off

Stay fast at high concurrency.

SUSTAINED QUERIES PER SECOND · PRODUCTIONfour customer deployments

Security analytics, F1001,000+

Market research300+

Customer engagement175+

Networking125+

1,000+ QPS held under 2s at p95

There is no coordinator queue in front of the engine. Each query gets its own slice of compute the moment it arrives, so p95 holds as the number of askers climbs.

1,000+ QPS sustained, under 2s at p95in production at an F100 security analytics team and a subscription management leader

Power high-velocity agent traffic.

AGENT QUERIES · LAST 60s · ONE CELL, ONE QUERYillustrative

agent:ops-watch

agent:pricing

agent:support

agent:forecast

short, parallel, burstyeach planned in ms, sized to itself

Agents do not ask one big question. They ask thousands of short, bursty ones an hour, in parallel. The engine treats each as its own small job, so one agent's burst never queues another's.

Built to handle short, bursty, parallel queries1,000+ QPS under 2s at p95

Keep your leverage

Embed in your stack.

TableauPower BINotebooksdbtApps

Open standards

e6data Query EnginePostgres wire · JDBC · ODBC · Arrow Flight

DatabricksUnity · Delta

SnowflakeIceberg tables

Self-managedIceberg · Hive

Works with Databricks, Snowflake and self-managed open-format environments through open standards.

JDBC, ODBC, Arrow Flight and RESTthe connectors your BI tools already have

Run serverless, or in your boundary

YOUR VPC · AWS OR AZURE

KUBERNETES · QUERY PODS23 of 40 vCPU in use

scales by the vCPU with each querydrops to zero when idle

your tables, catalog and policies, untouched

Kubernetes-native. Run it as a hosted service, or deploy it in your own VPC on AWS and Azure. Your data, catalog and policies stay where they are either way.

Same engine, same elasticity, in either modeon-prem, sovereign, air-gapped

No migration or rewrites.

DIALECT REWRITE · EXPLAINED

-- as written (Snowflake)

SELECT DATEADD(day, -7, CURRENT_DATE()), IFF(x > 0, 'up', 'down')

-- as run

SELECT CURRENT_DATE - INTERVAL '7' DAY, CASE WHEN x > 0 THEN 'up' ELSE 'down' END

-- 2 functions rewritten · same result · shown before it runs

Point the engine at the tables you already have and keep the SQL you already wrote. If a dialect differs, the SQL is rewritten and the rewrite is revealed to you before it runs.

Rewrites shown, never silenttranspiler coverage

Account for every action

Pay as you go.

Granularly scale at 1-vCPU-hr consumption with each query, drop to zero when idle, not to a warehouse capacity size you picked in advance.

Compute per benchmark run 36–48% below Databricks and 32–63% below Snowflake at list pricesSept–Oct 2026 · TPC-DS 1 TB · single-stream

Govern every query.

POLICIES IN FORCEillustrative · your catalog is the source

SCOPERULEWHOEFFECT

catalogfinance.*analysts, finance-agentsread

columncustomers.emailsupport onlymasked

roworders WHERE region = :user.regionevery rolefiltered

admissionmax 40 concurrent per workloadbi-prodqueued, not failed

timeout30s per clusteradhoccancelled

Set guardrails and thresholds per cluster. Admission control, workload isolation, and timeouts will log, alert, or cancel a query in real time before it costs you.

Your catalog is the source of truthadmission control, isolation and timeouts

Trace every query.

QUERY LOG · LAST 60sillustrative · exportable over OpenTelemetry

WHOFROM · QUERYGB READMSCOST

analyst:priyaTableau · cohort join31.51,240$0.0162

tenant:acmeembedded · 42 tiles18.2640$0.0094

agent:ops-watchAPI · checkout p955.0210$0.0031

svc:billingdbt · daily rollup9.6880$0.0049

analyst:weinotebook · funnel12.4405$0.0064

TOTAL5 statements76.7$0.0400

Every statement is logged with who ran it, from which tool, what it scanned, and how long it took. Export it over OpenTelemetry to your existing tools.

One record per statement, exportable

DIFFERENTIATION

Why not just…

You are considering	What it is good at	Where e6data fits
Scaling your Databricks SQL or Snowflake warehouse	Broad workloads, one platform	Your heaviest, high-concurrency workloads at a fraction of the cost, on the same tables at same or better performance
Trino, Presto, Athena, Starburst	Open, flexible engines	3–3.5x faster on total suite time than EMR Trino and Starburst Galaxy on the same Delta tables (TPC-DS 1 TB, 8 Oct 2026); every query finished in that run
A real-time database (e.g., ClickHouse, StarRocks)	Speed on data copied into its own format	Real-time speed on the lakehouse tables you already govern, without a copy to sync
Tuning queries and warehouses	Quick wins	Do it first. When tuning stops paying, swap the engine for that workload

PROOF

Proven on production workloads.

Up to 50%lower TCOsame or better performance

1,000+QPS under 2s at p95in production

$12M+saved over three yearsNasdaq-listed SaaS

200 → <10minutes to insightglobal media company

CASE STUDY · NASDAQ-LISTED B2B SAAS

15M+ customer-facing queries a day, 60% cheaper than Databricks

80%+ cheaper than Snowflake. 2x faster p95 than Databricks. $12M+ saved over three years. No changes to data, governance, applications or queries.

Read the case study

$12M+

saved over three years

QUERY ENGINE · SIX ENTERPRISES

Subscription management leader15x faster1,000+ QPS80% lower TCO

Global networking giant9x faster125+ QPS80% lower TCO

Global customer engagement platform6x faster175+ QPS60% lower TCO

Leading market research firm5x faster300+ QPS60% lower TCO

Iconic global media brand3x faster50% lower TCO

World's largest quick-commerce firm2x faster50% lower TCO

3–3.5xfaster than EMR Trino and Starburst Galaxy on total suite time

1.41xfaster than Databricks on total suite time

TPC-DS 1 TB8 Oct 2026 · Delta Lake on S3 · warm · single stream

Book a demo Read the docs

PRICING

Pay for the queries that run.

Compute scales with each query and drops to zero when idle. No warehouse size to pick, no idle hours on the bill.

See pricing

COMPATIBILITY

Works with the formats, catalogs, tools and clouds you already run.

Compatible with what you already run.

See full compatibilityShow less

SQL dialects and data platforms

Databricks

Snowflake

BigQuery

RedshiftAthena / TrinoSpark SQLANSI SQL

BI and connectors

Tableau

Power BI

AI agentsReporting toolsAPIsJDBCODBCPython

Catalogs and governance

Unity Catalog

Hive Metastore

AWS Glue

PolarisIceberg RESTRangerOPAPurview

Table formats

Delta Lake

Iceberg

Hudi

ParquetJSON

Storage

AWS S3

Azure ADLS

Google GCSOneLakeAny object storage

Deployment

Azure

Google CloudIn-VPCOn-premAir-gapped

Connects over open catalog APIs, Iceberg REST and Unity REST, the same way Apache Spark does. Zero query rewrites, no data movement, no governance changes.

Questions

Do we have to migrate our data?