The HDF Group is ready to announce a new major release, HDF5 1.12.0. This will be the first release that allows our users to access HDF5 data stored on non-POSIX storage, such as Cloud or Object Store. The release will also include performance enhancements to the parallel and sequential HDF5 libraries.
In this talk we will give an overview of the new features, and will outline the HDF5 roadmap for the next year. The major new features include Virtual File Drivers (VFD) to access HDF5 files via Amazon Simple Storage Service (Amazon S3), on Hadoop Distributed File system (HDFS), and HDF5 library’s Virtual Object Layer (VOL) that enables different methods for accessing data and objects that conform to the HDF5 data model.
The release will also address a new requirement for referencing data that is stored in another HDF5 file. It will have extended capabilities for referencing HDF5 datasets, groups, attributes, and region references including simplified programming model and APIs for working with the HDF5 references.
A big portion of the talk will be dedicated to performance improvements done for the release and our work in progress.
In 2020 we plan to hold in-depth webinars on how to create VOL connectors and tutorials on tuning HDF5 applications performance including performance on HPC systems.
This webinar will be held December 6, 2019, at 11:00 a.m. CST. We will distribute a recording after the event but we’d love to have you attend live, and we will have a Q&A session at the end.