Hi again-
And I do think that Linux is a good move in some instances, but for
large scale, clustered rendering farms, it's a monetary move, I don't
really have to try and compare a 64 processor scalable Linux cluster with
an Onyx Reality Monster. that's not fair to Linux... If your focus is
clustering, and you want pure performance and you are not taking money into
consideration (unlikely), then SGi is it, they cluster and scale procs
beyond the imagination of current Linux. And if you want parallel, I'll
only point back to my argument that the NSA buys HUGE amounts of SGI, this
is for certain parallel operations that make much more sense to run on SGI
than other commodity platforms. All Black Helicopter and big spending
arguments aside, the NSA gets it's pick of tools, and it chooses what works
best, not like where I work, we choose based on business partnerships, that
is a sad fact in the private sector, and they picked SGI, no surprises
there.
My arguments here are why SGI is one of the best for clustered and
parallel computing, the white paper on ccNUMA arch. explains why their
parallelism is amazing, look for it on SGI's pages. And clustering? I
can't tell you how big the clustered systems get, Jeremy can answer that,
but suffice to say, they are far beyond what Linux is capable of.
Now, people are talking about Linux and it's development in 4 years,
well, SGI is not standing still, contrary to popular belief, they will
advance as well, and possibly faster in some areas. Linux is not the only
technology moving forward, there are many others working on expanding what
they do and their advantage may be that they are focused in a given area
(also a dis-advantage, to be sure).
mbutler
* mbutler2 at mmm.com <mbutler2 at mmm.com> [011121 11:07]:
> Linux tha tdoes what their SGIs can't, but I doubt that, I think it looks
> more like a monetarily motivated move, and that is fine. I also have
every
I think that Linux just happens to be the commitity os for companies who
are looking to move away from single system solutions to clustering (HA)
or parallel computing clustering. People just choosing the OS that
fits, then developing tools and infrastructure on top of that OS.