Summary

While we are often busy trying to learn more and more about the newest innovations in tech, sometimes it is important to take a step back and think of how much we truly know about the existing accomplishments. This week we take a look at how supercomputers, a very common term and widely used machine in big institutions function. The article offers a very simplified overview of how supercomputers work, in both hardware and software.

Technical Article

For anyone who’s studied computers or anything relevant to the field from a young age, they’ve certainly heard of super computers sooner or later. An ambiguous term used to describe an astonishingly capable and astonishingly gigantic machine. While we’re always fascinated by the things said machine can achieve, have we ever wondered what really is the detailed mechanism behind its work?

Because it’s surely not pixie dust or some sort of sorcery fueling its abilities. If you take a moment of your time and read below, you’ll be disappointed and remotely surprised to learn, a supercomputer is vastly composed of traditional not-super elements which come together under specialized conditions to curate a machine so powerful.

Compute nodes

Any point in a network that can create, receive or transmit data is called a computer node. A computer node can range from being any physical device in the network such as printer, router, scanner etc to different virtual components as well. In super computers almost all such nodes are each their own self sustained computers with their own designated CPU and block of RAM for storage

Processors

CPU is that one thing almost everyone has heard as early as 5th grade, it’s one of those first things we are taught about computers, “the brain” of the computer. While regular computers have one CPU each, supercomputers utilize thousands of these at a given time. Alongside this they also use GPUs, which are specifically used for graphics and animation display.

High speed interconnect network

The network of nodes utilizes a very civilized high bandwidth and low latency networking which follows the standards of Infiniband, allowing the different parts of the system to communicate spontaneously and efficiently in a matter of microseconds.

Memory and storage

They have RAM going as high up as petabytes to allow the vast amount of data being processed to be stored effectively as well.

Cooling system

A very essential element of these computers is their cooling system, which often uses liquid cooling as well to ensure the different parts of hardware don’t suffer malfunctioning due to overheating .

Specialized software

All computers use the Linux based operating system built around the original Linux core released in 1991, this operating system is the most scalable for computer work and is free to use for modification and running. It is commonly used in androids and regular computers. Super computers use highly specialized versions of this Linux based operating system, which are specifically designed to enable “ parallel programming”.

Parallel programming

I often keep saying the most important thing about supercomputers is “xyz…” because almost every single element of such computers is worth discussing. However, this time when I say “parallel programming” is truly the most important aspect of supercomputers, trust me, I mean it.

Supercomputers, as you can tell by now, are essentially a large assembly of many regular computers working together under highly sophisticated software systems.

However, the big reason why they’re so effective is due to their technique of parallel programming, which ensures every single piece of hardware is working on something different, while they’re all working for the same task and all working simultaneously, the software also tackles hundreds of different issues simultaneously.

This method of work division allows a supercomputer to complete a very complex task in much less time than any traditional computer.