AIStor Tables

The Structured Data Layer for AI & Analytics

MinIO AIStor Tables embeds a full Apache Iceberg™ V3 REST Catalog directly into the data storage layer for easier, faster BI and AI. Structured tables live securely alongside unstructured objects and scale together on one data plane. No separate metastore. No bolt-on catalog. No infrastructure sprawl. Simpler. Faster. Secure.

Get a Demo

Get the Solution Brief

One Foundation for Storage and Analytics

Query, share, and analyze data where it lives, with table capabilities built directly into the storage layer your infrastructure team already runs.

Analyze everything in one place.

No jumping between systems, platforms, or tools.

Tables as Native Primitives. ACID guarantees, schema evolution, and time travel are built into every table. No external metastore, no rewrites, no lost reproducibility.

Integrated Iceberg Catalog. Point Spark, Trino, Flink, or PyIceberg directly at your tables. No new catalog to deploy, no client wrapper required.

Table Sharing. Give Databricks live, read-only access to on-premises Iceberg and Delta tables, queryable the moment they land. No pipeline, no replication.

Deploy one system.

Infrastructure sprawl isn't a tradeoff. It's a tax.

Warehouse Buckets as Native Primitives. Provision an isolated, multi-tenant space in seconds, not a ticket. Isolation is built in, not configured after the fact.

Hierarchical Namespace Support. Give every business unit its own secured slice of the platform, without standing up separate infrastructure for each.

Storage-Layer Access Enforcement. Read-only, scoped, expiring access enforced at the same layer that already holds your security and compliance policies. One control plane, not two.

First-Class Data Organization for Modern Lakehouse

Warehouse buckets, tables, and namespaces built directly into the storage layer for modern data architectures.

Warehouse Buckets as Native Primitives

Top-level containers built into the storage layer, eliminating external bucket management and enabling seamless multi-tenant data organization.

Tables as Native Primitives

Tables are built into the storage layer with ACID guarantees, schema evolution, and time travel, eliminating the need for external metastores or catalog infrastructure.

Hierarchical Namespace Support

Organize tables into multi-level namespaces with isolated access policies. Support for complex organizational structures enables enterprise taxonomy and multi-tenant isolation at massive scale.

Integrated Iceberg Catalog

Native Apache Iceberg REST Catalog implementation eliminates Hive, Glue, Nessie, and proprietary catalogs. Direct integration with Spark, Trino, Flink, and PyIceberg without intermediary services or client wrappers.

Table Sharing

With native OpenSharing, AIStor Table Sharing makes the storage layer a secure, governed layer for distributing structured Iceberg and Delta tables.

Lakehouse Architecture Without Operational Complexity

Traditional lakehouse deployments require separate object storage, metastore infrastructure, and catalog synchronization. MinIO AIStor unifies these layers by making tables and namespaces first-class storage primitives.

This architectural innovation enables AI teams to build training pipelines, real-time feature stores, and analytical workloads on a single exascale-capable platform.

Zero Infrastructure

No external metastore or catalog infrastructure required

ACID Transactions

ACID transactions guarantee consistency for concurrent AI workloads

Schema Evolution

Schema evolution supports iterative model development without rewrites

Time Travel

Time travel enables reproducible training runs and model versioning

Why AIStor is Different

Traditional data lakes force enterprises to manage separate object storage, metastores like Hive or Glue, and complex orchestration to maintain consistency. Warehouse buckets and tables are treated as external constructs layered on top of storage, creating fragmentation, latency, and operational overhead.

MinIO AIStor makes warehouse buckets, tables, and namespaces native storage primitives, delivering the unified lakehouse foundation required for workloads at exascale.

Unified Lakehouse Platform

Objects and tables are managed in a single system with full catalog metadata visibility, without separate infrastructure.

Zero External Dependency

No Nessie, Hive, Glue, or metastore infrastructure required for table management.

First-Class Primitives

Warehouse buckets, tables, and namespaces are native constructs with ACID guarantees and multi-table atomic transactions built into the platform.

‍

Business Impact

Unify Data Infrastructure

Consolidate warehouse bucket management, object storage, table operations, and catalog layers on a single platform, reducing operational complexity and infrastructure costs.

Simplify Multi-Tenancy

Hierarchical namespace isolation allows multiple teams to share infrastructure securely, with granular access control and resource quotas for production deployments.

Future-Proof AI Architecture

Native Apache Iceberg 3.0 compliance with REST Catalog specification ensures direct integration with any Iceberg-compatible tool without intermediary services.

Enable Exascale Analytics

Query petabytes of structured data with consistent performance, supporting real-time feature engineering and analytical workloads for modern applications.

Ready to See It in Action?

Get AIStor running in your environment in minutes.

Get Started

Additional Resources

Logos of MinIO AI'Stor and Apache Iceberg on a dark background with colorful smoke at the bottom.

Blog

Announcing General Availability of MinIO AIStor Tables

The first data store to build-in Apache Iceberg™ V3, unifying tables and objects for analytics and AI at scale

Blog

Iceberg Views in AIStor Tables

AIStor Tables gives self-hosted deployments the same view capabilities that the Iceberg specification defines, but few catalogs actually deliver.

Training

MinIO Academy

Access training, documentation, and hands-on resources in one place with role specific content for administrators, developers, data engineers, and partners.