The JBGeorge Tech Blog

.

This is a duplicate of a high-performance computing blog I authored for Cray, originally published at the Cray Blog Site.

Let’s talk operating systems…

Take a step back and look at the systems you have in place to run applications and workloads that require high core counts to get the job done. (You know, jobs like fraud detection, seismic analysis, cancer research, patient data analysis, weather prediction, and more — jobs that can really take your organization to the next level, trying to solve seriously tough challenges.)

You’re likely running a cutting-edge processor, looking to take advantage of the latest and greatest innovations in compute processing and floating-point calculations. The models you’re running are complex, pushing the bounds of math and science, so you want the best you can get when it comes to processor power.

You’re probably looking at a high-performance file system to ensure that your gigantic swaths of data can be accessed, processed, and stored so your outcomes tell the full story.

You’re most likely using a network interconnect that can really move so that your system can hum, connecting data and processing in a seamless manner, with limited performance hit.

We can go on and on, in terms of drive types, memory choices, and even cabling decisions. The more optimized your system can be for a high-performance workload, the faster and better your results will be.

But what about your software stack, starting with your operating system?

Live tweet chat

Join us for a live online chat at 10 a.m. Tuesday, June 19 — “Do you need an HPC-optimized OS?” — to learn about optimizing your HPC workloads with an improved OS. You can participate using a Twitter, LinkedIn or Facebook account and the hashtag #GetHPCOS.

Sunny Sundstrom and I will discuss what tools to use to customize the environment for your user base, whether there’s value in packaging things together, and what the underlying OS really means for you.

Bring your questions or offer your own expertise.

Moment of truth time

Are you optimizing your system at the operating environment level, like you are across the rest of your system? Are you squeezing every bit of performance benefit out of your system, even when it comes to the OS?

Think about it — if you’re searching for an additional area to drive a performance edge, the operating system could be just the place to look:

• There are methods to allocate jobs to run on specific nodes tuned for specific workloads — driven at the OS level.
• You can deploy a lighter-weight OS instance on your working compute node to avoid unnecessarily bringing down the performance of the job, while ensuring your service nodes can adequately run and manage your system.
• You can decrease jitter of the overall system by using the OS as your control point.

All of this at the OS level.

Take a deeper look at your overall system, including the software stack, and specifically your operating system and operating environment.

You may be pleasantly surprised to find that you can drive even more performance in a seemingly unexpected place.

Until next time,

JBG
@jbgeorge

Add our 6/19 tweet chat to your calendar for further discussion.