What storage technologies will meet rising performance and cost demands?
Cray

Step one to solving your I/O and mixed workload storage challenge is understanding exactly what you’re up against. Regardless of your field of research or the nature of your organization, if you’re using high-performance computing (HPC) then you need high-performance storage. Storage matters. A lot.

Traditionally, performance and cost alone have driven the choice of HPC storage platform. And traditionally, organizations have responded to the performance demand by building systems with enough spindles to deliver the required throughput. It’s an approach that has often resulted in more capacity than necessary.

But data — and the demands it puts on HPC and storage — is changing. Capacity requirements are growing exponentially. Some large research institutes are planning storage solutions in the 50–250 petabyte (PB) range. In the commercial world, artificial intelligence (AI) research and development is expected to generate and use data of even larger magnitude. Keeping up with data growth means the cost of storage is quickly overtaking compute.