Learn about the Column-Oriented Storage Layout for HDF5 (HEP001) – Aleksandar Jelenak – Call the Doctor 5/19/26

Learn about the Column-Oriented Storage Layout for HDF5 (HEP001) – Aleksandar Jelenak – Call the Doctor 5/19/26

Join us for our next Call the Doctor session on Tuesday, May 19, 2026, at 12:20 p.m. CT, featuring Aleksandar Jelenak discussing HEP001: Column-Oriented Tabular Data in HDF5. In the spirit of modern analytical frameworks like Apache Parquet and Apache Arrow, this new ecosystem proposal introduces a native, column-oriented storage layout for tabular data within HDF5 groups. By decoupling each column's physical storage, HEP001 resolves long-standing row-oriented limitations—allowing for independent per-column chunking and filter pipelines, eliminating I/O amplification during column-specific queries, and making schema evolution a breeze. Come learn how this specification allows columnar tools and traditional multidimensional array tools to seamlessly meet in the middle, and stay for our usual open Q&A and troubleshooting session.

To join, just jump on the zoom:
Launch Meeting - Zoom
May 19, 2026,12:20 p.m. central time US/Canada

1 Like

We are excited to share the full “Call the Doctor” session featuring Aleksandar Jelenak, where he introduces HEP001: Column-Oriented Tabular Data in HDF5. This proposal lays the groundwork for a community-vetted standard for storing tabular data in HDF5 using a clean, columnar approach.

Relevant Links

Today I Learned

  • Columnar storage in HDF5 allows for specialized compression and chunking per column, unlike legacy compound data types.
  • The “storage-first” approach aims to allow multiple software tools to implement the standard rather than tying users to a single, monolithic library.
  • The project uses MyST for technical documentation to enable better cross-referencing and web-based access, moving away from static PDFs.

For a comprehensive breakdown of the proposal in this Call the Doctor session, including the rationale behind our design decisions, check out our latest blog post.

1 Like