The presentation introduces Enakta Labs, a storage software company founded in 2023 with the goal of making DAOS (Distributed Asynchronous Object Storage) practical, usable, and supportable for enterprise and high-performance environments. Founded by Denis Nuja and Denis Barakhtanov - both long-time contributors to DAOS and members of the original DAOS Foundation team - Enakta Labs builds on years of hands-on experience deploying DAOS-based systems in demanding production and government environments.
Enakta Labs positions its platform as a bridge between the raw performance of DAOS and the operational realities of enterprise IT. While DAOS is widely recognized as one of the fastest storage engines available—particularly for HPC and AI workloads—it has traditionally been difficult to deploy, manage, and integrate. Enakta Labs addresses this gap by delivering a fully managed DAOS lifecycle platform that runs on bare metal, removes operational complexity, and exposes familiar enterprise interfaces.
At the core of the offering is Enakta Labs Platform v1.3, based on DAOS 2.6.4. The platform provides a highly available management framework, containerized data services, and a slim, immutable Linux OS image booted entirely via PXE. This approach allows entire clusters - from small systems to hundreds of nodes—to be deployed in hours rather than days or weeks. All services run in lightweight containers, and the system avoids legacy dependencies such as TFTP, favoring modern HTTPS-based PXE booting.
A major differentiator is usability and integration. Enakta Labs provides a fully featured, high-performance SMB interface, enabling DAOS to be consumed by traditional enterprise applications, alongside a limited-API S3 interface (currently in technical preview). Native PyTorch integration allows direct acceleration of AI and machine-learning workflows, while support for RoCE over 400-Gb Ethernet enables extreme throughput with low latency on modern networking hardware.
Performance and scalability are central themes of the presentation. Enakta Labs demonstrates real-world benchmarks from large GPU clusters, including participation in the IO-500 benchmark on Core42’s 10,000-GPU “Maximus-01” system. These results show the platform scaling linearly with hardware, achieving extremely high bandwidth and IOPS while approaching physical hardware limits. The company emphasizes that future gains will come automatically as faster NVMe drives and network adapters are introduced—without requiring architectural redesigns.
The platform is designed to eliminate common pain points in high-performance storage: GPU idle time caused by slow I/O, excessively long filesystem checks, over-provisioning to meet performance targets, and the need for deep DAOS expertise just to operate a system. Enakta Labs also highlights its customer-centric philosophy: transparent roadmaps, simple licensing, no mandatory professional services, and direct access to engineering support - even outside business hours.
Target use cases include HPC, AI/ML, fintech, and media and entertainment, particularly environments where storage performance directly impacts business or research outcomes. The go-to-market strategy is partner-first and hardware-agnostic, with simple node-based perpetual licensing and subscription options for service providers. Looking ahead, the roadmap includes broader S3 and NFS support, continued performance optimization, and deeper ecosystem partnerships, while maintaining a deliberate focus on product quality and customer satisfaction over aggressive VC-driven growth.
Overall, Enakta Labs presents itself as a pragmatic enabler of next-generation storage - unlocking DAOS performance for real-world enterprise and AI workloads without the traditional complexity barrier.
The coming months will be interesting as some vendors like HPE has clearly selected DAOS for HPC and AI storage already deployed at famous US labs.
0 commentaires:
Post a Comment