[ILUG] trialling Sun hardware

[Colm MacCarthaigh]

On Fri, May 19, 2006 at 10:30:19AM +0100, ollie at eillo.org wrote:
> Cool - any indication of price difference between the itanium and
> Niagara? Been on HEAnet for four years now - thats four happy years of
> websucking :) One more snapshot of the Arch Linux repo before i finish
> up for good :)

We got the Itanium box in an odd way, we actually paid less for it than
the price of just the memory on crucial, I think we paid cost for the
memory and got the chassis for free. At the time, it was around 30k.

The number I've been given for the T2000 is about 10k-15k depending on
what you put in it, but I think there is downward room in that price.
Noone pays list.

I had some behind-the-scenes time at JavaOne and got to speak to some
people about the Ubuntu/Sun thing, and to me it seems like it's going to
be really cool. Canonical havn't announced anything yet, but it seems
like there is goign to be reciprocal support. Sun will sell you boxes
with Ubuntu on them, and offer support for it, while Canonical will back
Sun up for support and Sparc will become an official Ubuntu distro
(though Canonical really only have as much say in that as the whole
community process allows), though that's really just some
slightly-informed speculation.

It'd work out well for people like me, as Sun could become a source of
ubuntu-supported Opteron kit for example.

Sun, and Jonathon Schwartz seemed really keen to emphasise that Ubuntu
would run on Niagara, Mark Shuttleworth seemed more cagey about it. I'm
not sure if it might be a bit too soon for that though, so I mailed
ubuntu-devel about it;

 https://lists.ubuntu.com/archives/ubuntu-devel/2006-May/018072.html

Still, it will be really cool. 

Interestingly, I heard some Sun people giving out Google as an example
Ubuntu user. 

If you don't have time to read my mail to u-d, one thing I havn't put
on my blog, because a lot of people can mis-interpret these things
is the libmicro comparison between system call return times on 
Solaris express (the fastest version I could test) and Linux;

	http://www.stdlib.net/~colmmacc/mv-res.html

Basically, most system calls seem to return faster on Linux. Also
notably, epoll() is not tested, which is more performant that Solaris'
poll().

-- 
Colm MacCárthaigh                        Public Key: colm+pgp at stdlib.net