Navigating HPC for Competitive Advantage in 2018

So far, we’ve only scratched the surface of what’s possible

istock 583997004

The most successful organizations in 2018 will use even more computing power to excel. To help our own organizations be more successful, we all need to champion better use of high performance computing techniques. It’s not hard to get the feeling that we’ve only scratched the surface of what is possible despite the use of powerful computing being a given in so many organizations. Here are my thoughts for digging deeper.

The first, most critical step is for an individual or group to see the possibilities, benefits, and ultimately the necessity of using HPC. Generally this means introducing HPC into the design process. This is, admittedly, a daunting task for those who haven’t actively pursued it before.

I advise starting with four things:

  1. Engage user communities (including user group meetings and software-specific conferences) to learn about software that appears to be most helpful for your work.
  2. Engage vendors. Ask them to educate you. Naturally, they’re on a mission to sell what they have, so apply some skepticism and critical thinking.
  3. Engage internal stakeholders. Connect the expertise of HPC systems experts, IT experts, and cloud advocates.
  4. Engage independent advice. Network with experts from National Labs, academic HPC centers, and specialist HPC consultants (e.g., NAG).

Reduce any internal divides

We commonly see companies with internal divides. Make a New Year’s resolution to figure out how to connect the dots for your own needs, and beat the external competition by reducing internal competition across any divides that exist in your organization.

Manage HPC as a value generator; HPC is not IT!

Don’t let the similarities between HPC and IT (they both use computers, right?) mask what should be different. Fundamentally, IT is an important capability that is generally operated as an accepted cost. HPC is a value generator, and needs to be managed as such. But please avoid creating a great divide between them. Cooperation and understanding are highly encouraged. Nevertheless, the goals against which HPC is managed need to be different.

Use cloud computing for some HPC

Enterprise organizations are finding enormous value in moving a range of IT operations to the cloud. The selection of cloud services for HPC is more complex than enterprise IT, because the role of parallelism (fine grain and global) and communications (bandwidth and latency) between compute units is much more important than non-HPC experts generally grasp. Forcing all HPC into the cloud means the flexibility and cost-effectiveness of local capabilities can be overlooked. The Total Cost of Ownership (TCO) models to evaluate cloud for HPC are also non-trivial. While HPC in the cloud should be taken quite seriously, it’s still a more complex choice than general enterprise IT use of the cloud.

The smartest organizations are very active in evaluating all technologies, but more conservative in buying them for on-site usage. Cloud providers offer great opportunities for exploring, while enabling you to be more conservative at home for the “non-burst needs” of your organization.

Winning in 2018

The longest journey begins with the first steps. Winning in 2018 starts by making sure you take the four steps mentioned above and, agreeing that we can all use more HPC in our endeavors to beat the competition.

Get Intel® Parallel Studio XE and download now 


Copyright © 2018 IDG Communications, Inc.