Great Big Linux Does Great Big Science
|
|
|
Comments (3)
Another interesting aspect about this is the democratization of computing. Most supercomputers today rely on being massively parallel. For example the No 1 on that list has 224,000 cores. As a user of that machine it would be difficult to access all those cores given a decent scheduler. Thus the ordinary user makes do typically with effectively far fewer cores. But you can buy a cluster yourself with that number of effective cores for a rather reasonable price.
As an example I purchased a 32 core (=4 nodes each with 8 cpus) cluster for around $8K (3 years ago) and am the sole user. I run it like I run my linux box at home and write ssh scripts to fire jobs off to all the nodes. You then let it run for a few days and it is rather surprising what you can achieve. I used to require access to a supercomputer but no more.
Of course if you run a gigantic numerical model such as a cosmology or global climate simulation you need all those cores for a day or so and so a supercomputer is vital. But if you run small models a personal cluster (running linux) is a very convenient way to go.
Of course if you run a gigantic numerical model such as a cosmology or global climate simulation you need all those cores for a day or so and so a supercomputer is vital. But if you run small models a personal cluster (running linux) is a very convenient way to go.
Post a comment
A free account is needed to post. Write your comment here first: your text is kept in this browser while you sign up or log in, and is waiting for you when you come back.
Your message:Markdown: **bold**, *italic*, [text](https://link), - lists, > quotes, $x^2$ for math. Formatting help
Sign up and post (it is free) or Log in and post

Drumroll........
and the percentage using Windows: 0.2% i.e. Only one out of 500. Talk about a token effort.
http://i.top500.org/stats