IT Connect

Information technology tools and resources at the UW

HPC by the Numbers

Hyak, UW’s shared supercomputer cluster, delivers high performance computing speed and power to UW researchers efficiently and cost-effectively. It’s fast, nimble, and scalable—and it continues to grow to meet Big Data and other research challenges. Here’s a look at Hyak by the numbers.

Hyak Growth

Growth since Hyak launched in 2010:

  • Total cores: Almost 300 percent growth – from 3,596 to 10,784 cores
  • Nodes: More than doubled – from 412 to 828 nodes
  • Monthly core hours used by researchers: Over 12 times more hours – from 574,000 to over 7 million hours
  • CPU hours used: 202 million hours

Hyak Computing Power

  • 2 petabytes of data crunched by UW researchers using Hyak (that’s nearly 30 years of HDTV content)
  • Data transmission via 100G Internet bandwidth
  • Hyak currently stands at 90 percent overall utilization

Elastic Supercomputing

Hyak’s “elastic supercomputing” resource model enables users to access additional computing power by retrieving unused compute cycles. This elasticity has allowed many research groups to use more computing resources than they have bought (allocated), as illustrated in the graph and pie charts below.

This means that researchers do not need to build and operate separate, small-scale systems that are expensive to maintain and consume more energy.

Groups with low allocation rates have been able to use Hyak’s compute power intensively when needed, using more than their allocation of this shared resource. By pooling resources, UW researchers are effectively and efficiently making use of Hyak’s elastic supercomputing power to advance their work.