HPC Storage Performance White Paper

Cray ClusterStor NXD Performance Comparison

Is your storage solution ready for the future?

High-performance computing (HPC) storage needs are changing with new areas of computing, such as artificial intelligence. While traditional parallel HDD-based file systems are capable of high sustained throughput and are good at large streaming reads and writes, the increasing number of small reads and writes, random seeks and unaligned writes can cause parallel file systems to falter.

In this white paper, Cray’s Torben Kling Petersen details storage options for high-performance computing and discusses why flash acceleration — and Cray’s ClusterStor™ NXD software stack — is a smart choice for your HPC storage requirements.

Learn about the benefits and constraints of various storage architectures:

Traditional
HDD-based storage

All-flash
storage solutions

A hybrid HPC
storage approach

In addition, the white paper covers:

  • The NXD I/O manager and how NXD delivers high IOPS without affecting streaming throughput
  • NXD I/O tools that allow the user to get a clear concept of HPC application throughput and to identify potential performance issues
  • The difference between the NXD software stack and other flash-caching solutions

Download for immediate access to this white paper.

Download

Explore related resources:

Related Resources

The Cryo-EM Researcher’s Guide to Breaking Data Bottlenecks

Research labs spend sometimes up to $10 million for an electron microscope that generates, on average, up to 2Tb of data every day. Without the infrastructure in place to transfer, process, store and analyze that data – to create those 3D models – it simply languishes.

Meet the Challenges of Implementing Cryo-EM Technology

The resources needed to take a drug treatment from concept to reality are extraordinary. And if a treatment is not going to work, the researchers need to know. Now.

The Critical Role of Storage in Autonomous Vehicle Development

Autonomous vehicle development presents a number of AI workflow challenges. Challenges that can slow your progress and erode your competitive advantage.

How to Streamline AI System Design for Autonomous Vehicles

No one will wait half a millennium for self-driving cars. So to speed development, the industry has turned to AI simulation, which brings its own set of challenges. Chief among those is the volume and variety of data ― and the implications for data storage.

See All Resources