[Discuss] Linux has 100% of Market Share.

[richard.pieri at gmail.com (Richard Pieri)]

On 6/18/2018 6:21 PM, Dan Ritter wrote:
> If all you need is a large number of processors working on
> different chunks of data, you're absolutely right, Marco.

Such as the kind of high throughput compute that HTCondor was designed
to manage. You don't see this kind of compute cluster listed in the Top
500 even though some of the largest high throughput compute grids out
there would easily be in the Top 10 by TFlop/s counts. You can get a lot
of compute together by grid connecting HT pools.

For what it's worth, about five years ago the HTCondor people set up a
4000-core single pool on AWS as a proof of concept.

> If you need to solve physical simulations and models that 
> require lots of interprocessor communications, no, you can't
> just run out to Amazon and say "Give me a data center full
> of machines for 24 hours".

Such as what you would throw at a high performance cluster like a
Beowulf or pretty much anything ever in the Top 500.

-- 
Rich P.