Intel’s DAOS 2.4 Storage Engine Released: Enhancing Open-Source Object Storage
Phoronix Premium offers an ad-free experience and exclusive access to multi-page articles on a single page for just under $4 per month. By supporting Phoronix, users enable the publication to produce more Linux hardware reviews and content. In the realm of storage, Intel’s divestment from its storage business and discontinuation of Intel Optane has led to interesting open-source software projects, including DAOS (Distributed Asynchronous Object Storage) – a software-defined object store designed for high-speed storage. The recent release of DAOS 2.4 brings several significant improvements and new features.
Essential Takeaways
DAOS 2.4 introduces key enhancements and additions:
- Stabilized support for UCX on InfiniBand fabrics
- Added support for Slingshot fabrics
- Introduced a technology preview level feature of metadata-on-SSD support
- Enabled support for multiple SCM namespaces per CPU socket when used on servers with Intel Optane persistent memory
- Implemented EC parity rotation for erasure coding
- Added multi-user DFUSE support
- Various other changes and improvements
For a more comprehensive list of changes, interested users can refer to the release notes of this open-source project update. Those who prefer building DAOS 2.4 from source can find the new release on GitHub. Initially scheduled for release in H1’2023, DAOS 2.4 arrived ahead of schedule. The roadmap for future releases includes stabilizing OmniPath Express support, expanding non-persistent memory support, implementing checksum scrubbing, providing Intel VMD hot-plug support, and introducing other new features.
Looking further ahead, DAOS 3.0, set for release next year, aims to bring multi-homed network support, DAOS storage tiering, Lustre integration, and catastrophic recovery support, among other additions. The development team is also exploring future features such as CXL SSD support, inline compression and encryption via Intel QAT, and more.
Overall, DAOS 2.4 represents a significant milestone in the evolution of the open-source DAOS project, offering improved performance, expanded capabilities, and a glimpse into the future of storage technology.
