This webinar makes the case for embedding an Iceberg catalog directly in object storage rather than running it as a separate service. Traditional Iceberg deployments require four layers: compute, catalog service, transactional metadata database, and object storage. AIStor Tables collapses that to two, storing all catalog metadata as objects inside AIStor with no separate database to provision or secure. The session covers the cost crossover point where on-premises AIStor becomes cheaper than public cloud for hot data at scale, the airgapped and zero-trust deployment options for regulated industries, and how to use Iceberg table columns to store S3 object paths alongside structured data, enabling SQL queries that return both structured records and direct references to unstructured files. A live PyIceberg demo shows warehouse and namespace setup, table creation, and querying both data types in a single Iceberg query.
.svg)
AIStor Tables stores all Iceberg catalog metadata as objects inside AIStor with no external database required, and supports views alongside full Iceberg REST catalog compliance.
At approximately five petabytes of hot data, on-premises AIStor storage cost falls clearly below public cloud storage cost even before compute is factored in, and the gap widens significantly once compute costs are included.
Storing S3 object paths (or vectors pointing to object paths) inside Iceberg table columns alongside structured fields enables a single SQL query to return both structured data and direct references to unstructured files, giving AI agents and human analysts unified discovery across an entire enterprise data estate.
Data architects, data engineers, and platform leads evaluating Iceberg lakehouse deployments on-premises or in hybrid environments, particularly those looking to reduce catalog infrastructure complexity or unify structured and unstructured data for AI workloads.
.svg)
.jpg)
