DAOS
DAOS is an open-source, software-defined storage system for AI and high-performance computing workloads.
What DAOS is
DAOS is an open-source, software-defined storage system for AI and high-performance computing workloads. Its core interface is a scalable key-value-array store exposed through libdaos, with POSIX namespaces and HDF5 datasets among the models built over it. Pools can span some or all storage engines, and containers represent datasets that can be snapshotted. Fine-grained access controls are available for pools and individual containers.
Objects can use erasure coding, replication, or no data protection. DAOS can initiate rebuilds if a storage engine becomes unresponsive, and it distributes filesystem data and metadata across engines. RDMA is available through libfabric or UCX when supported. The client ecosystem includes libdfs for POSIX namespaces; documentation also covers interfaces including MPI-IO, Python, PyTorch, Spark and Hadoop, and TensorFlow-IO. DAOS can be deployed on premises or in cloud environments. Version 2.8 is built for Linux on x86_64 and lists SLES 15 SP6/SP7, openSUSE Leap 15.6, and EL 9.6/9.7. That version does not support Ubuntu, Fedora, CentOS Stream, or RHEL 8 and its clones. DAOS is free and uses the BSD-2-Clause Plus Patent License.
Good
- Open-source and free to use
- Supports POSIX namespaces and HDF5 datasets
- Containers can be snapshotted with fine-grained access controls
- Offers replication, erasure coding, and rebuilds
- Can deploy on premises or in cloud environments
Watch out
- Version 2.8 does not support Ubuntu or Fedora
- Version 2.8 excludes CentOS Stream and RHEL 8 clones
- Built for Linux on x86_64
Who it suits
DAOS is aimed at teams building AI or HPC storage systems that need scalable datasets and configurable protection. It suits Linux deployments on the distributions listed for version 2.8.
Verdict
DAOS provides a storage foundation for AI and HPC, with several data models and protection options. Check its version 2.8 Linux distribution requirements before planning a deployment.
The facts
Each one read from DAOS's own pages, with the date it was checked.
What it does
- Access control
- Fine-grained ACLs can be set on storage pools and individual containers.checked 3 Oct 2026
- Audience
- The site presents DAOS as an open-source storage platform for AI and HPC workloads.checked 3 Oct 2026
- Data model
- Its core interface is a scalable key-value-array store, with data models including POSIX namespaces and HDF5 datasets built over it.checked 2 Oct 2026
- Data models
- The overview lists POSIX namespaces, HDF5 datasets, and Python among data models built over libdaos.checked 3 Oct 2026
- Data protection
- Objects can use erasure coding, replication, or no data protection; DAOS can trigger self-healing rebuilds when an engine becomes unresponsive.checked 2 Oct 2026
- Datasets and snapshots
- DAOS containers represent datasets, can be snapshotted, and support fine-grained access control.checked 2 Oct 2026
- Deployments
- DAOS can be deployed on premises or in cloud environments.checked 3 Oct 2026
- Framework support
- The DAOS v2.8 documentation lists support topics for MPI-IO, HDF5, Python, PyTorch, Spark and Hadoop, and TensorFlow-IO.checked 3 Oct 2026
- Integrations
- The DAOS client ecosystem includes libdfs for POSIX namespaces, and its user guide documents a PyTorch interface for loading data from DAOS.checked 2 Oct 2026
- License
- The project repository displays the BSD-2-Clause Plus Patent License.checked 3 Oct 2026
- Network support
- DAOS can use libfabric or UCX and lists Ethernet, RoCE, InfiniBand, Omni-Path, and Slingshot fabrics.checked 3 Oct 2026
- Notable limits
- DAOS 2.8 does not support Ubuntu, Fedora, CentOS Stream, or RHEL 8 and its clones.checked 2 Oct 2026
- Performance
- The site describes distributed metadata, ultra-high bandwidth and IOPS, and time-to-first-byte below 20μs when using RDMA where available.checked 3 Oct 2026
- Protection and recovery
- Objects can use erasure coding or replication, and DAOS can trigger self-healing rebuilds when an engine becomes unresponsive.checked 3 Oct 2026
- Purpose
- DAOS is an open-source, software-defined storage system for AI and HPC workloads.checked 3 Oct 2026
- Release
- The DAOS documentation identifies its current documentation set as version 2.8.checked 3 Oct 2026
- Security
- DAOS supports access control lists on pools and individual containers.checked 2 Oct 2026
- Snapshots and scale
- DAOS containers are user-managed datasets that can be snapshotted; the site says a POSIX container can hold trillions of files or directories and petabytes of data.checked 3 Oct 2026
- Storage interface
- Its baseline interface is a high-performance, scalable key-value-array store exposed through libdaos.checked 3 Oct 2026
- Storage pools
- Pools can span all or some storage engines, and pool access can be controlled with fine-grained ACLs.checked 2 Oct 2026
- Support
- The DAOS site lists member support by email at [email protected] and links to a Linux Foundation support page.checked 3 Oct 2026
Where it runs
- Supported platform
- DAOS 2.8 is built for Linux on x86_64; its supported server and client distributions include SLES 15 SP6/SP7, openSUSE Leap 15.6, and EL 9.6/9.7.checked 2 Oct 2026
Alternatives to DAOS
- 1
JuiceFS
Free planOpen sourceAPILinuxmacOSSelf-hosted7.9From $0.02 a month - 2
MinIO AIStor
Free planWebAPILinuxmacOS7.3Free plan - 3
StarWind Virtual SAN
Free planWebAPILinuxSelf-hosted7.2Free plan - 4
MooseFS
Free planWebAPILinuxmacOS7.1€9 one-time - 5
SeaweedFS
Free planAPILinuxmacOSSelf-hosted7.0From $20 a year - 6
Ceph
Free planAPILinuxSelf-hostedWindows6.6Free plan