HPC Basics Series
Supercomputers are used for just about everything. What turns a regular computer into a supercomputer is clustering multiple highly-performance, optimized computers all configured to accomplish a specific type of task. This optimization will usually include finely tuned hardware, a specialized network, vast amounts of storage and so forth. Conversely, the workloads that require a supercomputer have usually two things in common, they either require computation on a huge amount of data, or they are very computationally intensive.
Supercomputers are not used for things like web service hosting, app hosting or general business computing that continually run. For example, you would not run your website or a spreadsheet on a supercomputer. You wouldn’t run these types of applications for a couple of reasons; the first is cost, supercomputers are comprised of highly optimized, expensive hardware and an off the shelf computer could handle this type of application without even breaking a sweat. The second is there would be no benefit to running commodity applications on such highly optimized hardware, again a standard computer would work just fine. The benefit is seen when the amount of data or the complexity of the calculation is such that commodity hardware cannot execute the job in a reasonable amount of time – (days, weeks, or months). Often these workloads need to be run in minutes our hours because the time to results matters.
Traditional uses for supercomputers have been in oil and gas exploration, converting seismic data into maps that indicate where to drill oil and/or gas wells. They are employed in engineering simulations such as fluid dynamics calculations on automobiles that look at drag and aerodynamic efficiency. Likewise, mechanical engineering simulations that calculate the loads across an entire structure to determine where weak points and stress may be located. A less traditional application can be seen in training models for artificial intelligence, where supercomputers are used to assemble sequenced genomes in a matter of hours as opposed to days.
Are supercomputers special beasts? Oh yes! They do specific types of computes very well, but also, they are not all that different than your average workstation – there are just many of them clustered together. They both probably run a version of Linux. You really wouldn’t know that you are dealing with a huge, highly optimized piece of hardware (when running your workloads on the Nimbix Cloud) until you launched a job that took your workstation overnight to finish and was completed by a Nimbix supercomputer in the time it took you to get a cup of coffee.