01
Elastic Compute
Query engines spin up in seconds and scale per workload. No idle clusters, no over-provisioned reservations, no surprise bills.
Cortex SQL · Data warehouseGENEDATA / 01
More answers.Fewer moving parts.
Bring analytical workloads to your data with a warehouse that connects storage, compute, and governance.
Shared context · Lineage · Governance
Connected
Your sources
Open table storage
Operational data
Event streams
Connected intelligenceCortex SQL
Governance
Business impactAnalytical workloads
Shared contextLineageGovernance
02
Open Storage
Native support for Apache Iceberg, Delta, and Parquet — your warehouse data is yours, queryable from any external engine.
03
Query Acceleration
Adaptive caching, materialized views, and result-set re-use deliver sub-second response on dashboards backed by petabyte-scale fact tables.
Storage and Compute
Storage you own. Compute you rent by the second.
Tables live in open formats in your own object storage. Compute attaches to them on demand, sized per workload, and detaches when the query finishes — which is why an idle warehouse costs nothing and a thousand-user Monday costs the same per query as a quiet Wednesday.
How It Works
Land, organise, accelerate, serve.
The warehouse optimises itself against real query patterns rather than requiring a DBA to predict them in advance.
Land in open formats
Data is written as Iceberg, Delta, or Parquet in your own object storage. No proprietary container, no export tax to leave.
Organise automatically
Partitioning, clustering, and file compaction are maintained by the engine against observed access patterns.
Accelerate hot paths
Frequently repeated queries are served from adaptive caches and incrementally maintained materialised views.
Serve every consumer
Dashboards, notebooks, and third-party engines read the same tables, so there is one copy and one definition.
Capabilities
Built for the workloads that break warehouses.
Concurrency spikes, enormous fact tables, and unpredictable ad-hoc queries are the normal case here, not the exception.
Separation of storage and compute
Scale either independently. Storage grows without paying for compute you are not using.
Workload isolation
Give BI, data science, and ETL their own compute so a runaway query cannot slow the executive dashboard.
Columnar compression
8–12× typical compression on analytical tables, reducing both storage cost and scan volume.
Incremental materialisation
Materialised views refresh only the partitions that changed rather than rebuilding from scratch.
Time travel
Query any table as of a past timestamp or snapshot, and restore from it without a backup pipeline.
Per-query cost attribution
Every query is attributed to a team, project, and cost centre, so chargeback is a report rather than an estimate.
In Practice
Who this is for.
The same tables, read very differently depending on who is asking.
Data Analyst
Explore without asking permission
Run ad-hoc queries against full-fidelity history on isolated compute, without worrying about affecting production dashboards.
Exploration that cannot page the on-call engineer.
Finance / FinOps
Make warehouse spend legible
See cost per query, per team, and per dashboard, and set budgets that alert before they are exceeded rather than after.
Spend attributable to the team that caused it.
Platform / SRE
Stop capacity planning for peaks
Compute scales to the workload in seconds, so quarter-end concurrency does not require a reserved cluster running all year.
No idle capacity bought for four days a year.
What Changes
What changes with open storage and elastic compute.
Traditional warehouses couple your data to their compute, and price accordingly.
| Dimension | Before Genedata | With Genedata |
|---|---|---|
| Data ownership | Proprietary format inside the vendor's storage | Open Iceberg, Delta, or Parquet in your own bucket |
| Scaling | Resize the cluster, then wait | Compute attaches per query in seconds |
| Idle cost | Reserved capacity billed around the clock | An idle warehouse bills for storage only |
| Tuning | Manual partitioning and vacuum schedules | Maintained automatically against real query patterns |
| Exit | A migration project measured in quarters | Your tables are already in an open format |
IcebergStorage Format
Zstd + dictionaryCompression
Plan-scopedConcurrent Queries
Per-queryResult Cache
The next step
Petabyte scale. Predictable cost.
Genedata Data Warehousing decouples storage from compute, lets you bring your own table format, and delivers consistent performance whether you're running a single board report or a thousand-user analytics product.
FAQ
Cortex SQL, answered.
The questions that decide a warehouse migration.
What is Cortex SQL?
Cortex SQL is the analytics engine on the Genedata platform — elastic compute over open table formats, the semantic layer, and the query surface behind dashboards and ad-hoc exploration. It reads the tables GeneFlow produces under GeneCatalog policy.
Who owns the data?
You do, literally. Tables are written as Iceberg, Delta, or Parquet in your own object storage. There is no proprietary container and no export tax to leave — an exit is a matter of pointing another engine at buckets you already control.
How does elastic compute affect cost?
Compute attaches per query and detaches when it finishes, so an idle warehouse bills for storage only. You stop buying peak capacity for the four days a year that actually need it.
Do I still have to tune partitions?
No. Partitioning, clustering, and file compaction are maintained by the engine against observed access patterns rather than predicted ones, which is where hand-tuned schemes usually drift.
Can other engines read our tables?
Yes. Because storage is open-format in your own account, external engines can query the same tables directly — Cortex SQL is not a gatekeeper on your own data.
How is warehouse spend attributed?
Every query is attributed to a team, project, and cost centre, so chargeback is a report rather than an allocation formula argued over quarterly.
Related content
Benchmark
TPC-DS Benchmark Report
Performance and cost-per-query against comparable warehouses.
GuideOpen Table Formats
Choosing between Iceberg, Delta, and Parquet on the platform.
GuideWorkload Isolation
Giving BI, data science, and ETL independent compute.
RelatedMigration
Moving off a legacy warehouse without a cutover weekend.
Take the next step
Benchmark it against what you run today.
Bring a representative workload to a working session, or start with the benchmark report and cost analysis.