Infrastructure for exabyte-scale data
At exabyte scale, operational overhead never stops. Crunch runs continuously on your data, wherever it lives, keeping it optimized, secure, and ready for AI. In production, data teams reduce their infrastructure costs by over 50%, accelerate computational workloads, and eliminate tedious operational overhead.
The Crunch platform
A super effecient large data process platform that supports features across table/data optimization, Iceberg/Delta table maintenance, query and Spark acceleration, and GDPR compliance.
20% to 50% immediate storage savings, 20%+ compute savings
World Class Compression
20% to 50% size reduction on top of standard ZSTD · No change to existing readers and writers · No pipeline changes
Policy Driven
Set and forget · Adapt as your data and schema changes · API enabled
Cost Effective
20%+ cheaper than all popular spark engines when compressing (and don't forget we also compress 20% to 50% more)
Fully automated table maintenance driven by policies.
One Stop Shop
Hive · Iceberg · Delta · Objects such as JSON, CSV, ORC, and more
Full Maintenance Coverage
Compaction · Delta vacuum · Iceberg snapshot expiration · Iceberg orphan file deletion · Partition deletion
Safty First
Safe handle of objects shared by tables · Extra TTL for added safty
Highly Scalable
Tens of thousands of tables · Tens of millions of paritions
Faster queries and ETL jobs.
20% to 50% Query Speed Improvement
Multi-column sorting · Z-ordering with Z-curve · Z-ordering with Hilbert curve · Stats improvements
Layout Intelligence
Insights and recommendations for clustering columns and algorithms · Adapt as data layouts and query patterns change
Keep your data full compliant.
Full Visiblity
Advance and effecient ML model for PII detection · 10% higher recall and precision than leading providers · 50% cheaper to run
Fast Mitigation
Data minimization · Data anonymization/masking · Pseudonymization · Row deletion
Our feature list keeps growing...
Building on top of the Crunch Platform
All new features will benefit from our effecient large data processing platform, and all results will be fully optimized - less to store and faster to query
20% to 50% immediate storage savings, 20%+ compute savings
World Class Compression
20% to 50% size reduction on top of standard ZSTD · No change to existing readers and writers · No pipeline changes
Policy Driven
Set and forget · Adapt as your data and schema changes · API enabled
Cost Effective
20%+ cheaper than all popular spark engines when compressing (and don't forget we also compress 20% to 50% more)
Supported formats
Crunch is built against the open formats. If your data is in Delta, Iceberg, Hive, or raw Parquet on AWS, GCP, or Azure, this catalog applies as written.
AWS
Amazon S3
- table
Delta Lake
Delta 3.2.x, 3.3.x, 4.1.x · Unity
- table
Apache Iceberg
Iceberg 1.6.1, 1.10.1 · HMS, Polaris, more catalogs coming
- file
Apache Parquet
Parquet 1.13.x, 1.15,x · 1.16.x coming soon
- table
Hive
External tables · Managed tables
GCP
Google Cloud Storage
- table
Delta Lake
Delta 3.2.x, 3.3.x, 4.1.x · Unity
- table
Apache Iceberg
Iceberg 1.6.1, 1.10.1 · HMS, Polaris, more catalogs coming
- file
Apache Parquet
Parquet 1.13.x, 1.15,x · 1.16.x coming soon
- table
Hive
External tables · Managed tables
Azure
Blob Storage · ADLS Gen2
- table
Delta Lake
Delta 3.2.x, 3.3.x, 4.1.x · Unity
- table
Apache Iceberg
Iceberg 1.6.1, 1.10.1 · HMS, Polaris, more catalogs coming
- file
Apache Parquet
Parquet 1.13.x, 1.15,x · 1.16.x coming soon
- table
Hive
External tables · Managed tables
Object stores
Non-ACID · PB-scale
- objects
Object maintenance
Non-ACID, PB-scale object stores
Any data, any catalog, any cloud, and data never leaves.
Lake tables or raw files, on whichever stack you already run, deployed inside your own VPC so bytes never leave.
Any data
Lake tables or raw files: Delta, Iceberg, Parquet; JSON, CSV, Avro, ORC. Crunch optimizes the files in place; your tables stay in their format.
Any catalog
Whichever catalog runs your tables: Glue, Unity, Hive Metastore, Polaris, BigLake. Crunch syncs and respects the source of truth.
Any cloud
Runs on AWS, GCP, or Azure. Single-tenant per customer.
Choose the best deployment for you
SaaS, Hybrid, or In-cloud on AWS, GCP, or Azure. Same API, same UI, same SLAs.
- POST/v1/crunch/jobs
- GET/v1/crunch/runs
- GET/v1/policies
- GET/v1/tables
Use any way you like
Browse, schedule, and inspect from the Console, or drive the same operations from the open REST API.
Data never leaves
Single-tenant per customer. Audit every operation. Bytes, keys, and metadata never cross your boundary.
Enterprise Grade Security Controls
Crunch natively supports SSO, RBAC, Auditing, Alerting, etc... It can run completely in your jurisdiction, taking advantage of the work you've already done.
Compliant by design
SOC 2, GDPR-ready, immutable audit trail. Even in SaaS, data resides in your domain.
Control Your Data
Crunch can deploy completely within your environment, or you can utilize a hybrid or SaaS model, where your data never leaves, but metadata uses our control plane.
FAQs
How does Crunch run against my data?
Crunch deploys into your cloud account. Compute reads and writes your bytes in your account; Granica never holds a copy. Every operation commits through the native format protocol. Delta stays Delta, Iceberg stays Iceberg.
Will it break my pipelines?
No. Readers see one consistent snapshot at all times; an in-flight compaction or vacuum is not visible until it commits. Schema, partitioning, manifests, and time-travel history all stay intact.
What kind of reduction should I expect?
Typically 20–50% over your current ZSTD baseline on Delta and Iceberg tables, depending on row structure and partitioning. We'll quote a defensible DRR after profiling one table.
Which clouds and formats do you cover?
Delta Lake (2.x / 3.x), Apache Iceberg (v1 / v2), Apache Parquet on raw prefixes, and Hive external tables. Across AWS (S3), GCP (GCS, BigLake), and Azure (Blob / ADLS Gen2). The Supported formats section above is the authoritative list.
How long does onboarding take?
Bring one table. We connect, profile the workload, and quote a DRR estimate within a few hours. First production policy lands in days, not weeks.
What about GDPR / right-to-be-forgotten?
Crunch rewrites every affected snapshot when you tombstone a row, preserving the Delta or Iceberg protocol so time-travel queries don't expose the deleted data. Audit log entries record every rewrite.
Three deployment models: how do I pick?
SaaS for fastest start (control plane + data plane Granica-hosted, your bytes never leave your bucket through us). Hybrid (recommended): control plane Granica-hosted, data plane in your VPC. In-cloud: both planes in your account, for tight security postures. Migrate between models without changing table configurations.
What's the pricing?
Volume-tiered against managed-byte under policy. Enterprise plans with custom SLAs, SSO, and dedicated support for larger lakehouses. Talk to sales for a quote against your DRR estimate.
What happens if I stop using Crunch?
Your tables stay where they are, in their native format, fully readable by your existing engines. We turn off the orchestration; the bytes Crunch wrote remain valid Delta / Iceberg / Parquet. No lock-in.
You select the data, we deliver the results.
Choose any data in a supported format and Crunch will create the workflow that improves efficiency, reduces cost, and eliminates overhead, with no impact to your current pipelines.