FILE
EXAScaler, DDN’s Lustre-based parallel file system, will be able to run on industry-standard storage servers.
Up until now EXAScaler has been a combined hardware and software system, running on DDN’s A1400X storage appliances or appliances from qualified DDN partners. The upcoming EXAScaler 7 release will allow it to run on qualified industry-standard servers, starting with DDN’s AI2200, with qualified third-party appliances to follow. Together with DDN’s object storage-based Infinia, which is already available in a software-defined format, the introduction will make both of DDN's data engines available in software-defined form.
DDN’s Chief Product Officer, Omer Asad, said: “AI infrastructure is becoming more diverse, and customers increasingly want the freedom to build environments around their specific requirements. By making EXAScaler available as software, we’re extending that flexibility across our portfolio. Customers can choose DDN’s reference hardware design-based systems or deploy our software on qualified third-party hardware, with the core capabilities they depend on from EXAScaler.”
The company suggested that its expanded deployment model can support organizations bringing AI and HPC workflows together. In life sciences, for example, researchers train AI surrogate models on molecular dynamics simulation outputs to help predict protein behavior without repeating every computationally intensive simulation. With software-defined EXAScaler on qualified hardware, simulation and AI training can run on the same data without copying it between environments.
The need for DDN hardware qualification should limit third-party hardware to high-performance offerings and not let EXAScaler run on poorly-performing servers.
HPE ClusterStor, its Cray supercomputer storage system, is a main EXAScaler competitor in the on-premises HPC, supercomputing market. It is not available in SW-defined form. SW-defined EXAScaler could (should) lead to lower appliance prices and enable it to compete more effectively against Everpure, NetApp, VAST Data, WEKA and others supplying high-performance storage systems for AI workloads.
DDN will continue developing purpose-built systems for customers who prefer the performance, reliability, and ease of integration of a complete DDN offering. It’s also developing enhancements to EXAScaler on multi-tenancy, performance and data protection. Validated performance data will be shared as it becomes available
A limited beta of SW-defined ExaScaler is planned for early 2027, with the new EXAScaler 7 capabilities targeted for rollout in the first half of 2027.