Mark FernandesIn a previous blog, we discussed the coming of PetaFLOPS to the masses. In general, increased compute power at affordable prices is continuing to arrive each day. Along that path to increased compute power, there are a few bumps that need addressing. One is storage and the required I/O rates.

It is relatively easy and affordable to grow the compute power of a system these days. Most commercially available HPC systems scale well today. However, at some point the capabilities of Direct-Attach Storage (DAS) or Network-Attached Storage (NAS) components can become a prohibitive bottleneck. The need for a larger storage pool with an affordable fast parallel file system may become apparent.

The fear (real or otherwise), the costs (real or otherwise) and the support commitment (real or otherwise) of parallel file systems are often considered when addressing such growth. In contrast, large data centers, large HPC centers and other large installations have sufficient staff and expertise to support parallel file systems.

Dell’s HPC Team Manager, Tim Carroll called for his Team to address this space. In the space, users are outgrowing a NAS or have a “small” cluster and want to grow it, but are too small to afford the staff and support thought to be needed with larger storage pools and a parallel file system.

With today’s technologies, we at Dell’s HPC Team believe we can address this space by carefully and judiciously exploiting the scalability of both the storage hardware and the available file system software.

Also, by using a limited number of configurations, we at Dell believe we can deliver these fast, parallel file system solutions at affordable prices. More to follow.

-- Mark R. Fernandez, Ph.D.