12/3 Are any of you CSUA alums working at Intel? Is it possible that we
might be able to hit Intel up for donated/partially donated (reduced
price) Core i7 Xeons when they come out? Who would be a good person
to contact about something like that? We're of course willing to put
out for them - perhaps we'll tattoo an Intel logo on toulouse if that's
what they want :-p -- steven
\_ fyi my wife recently started there and they eliminated the employee
discount on CPUs a few months ago
\_ I work at Oracle and I can get you 25% discount on our license
and bronze level support. How many CPUs are you thinking? With
a 25% discount, each CPU support is only $75,000/year.
\_ Why are you guys so obsessed getting an i7-based machine?
You look like salivating hardware dweebs, rather than people
with a plan.
\_ Well, this is the CPU-Salivating Undergrad Association,
isn't it?
\_ Drool
\_ I love Oracle and Larry Ellison is my master. We're trained
samurais and we'll obliterate anyone who makes us lose
face. Shut up and buy Oracle. You have no choice.
\_ Because they're shiny. What, are you saying we should buy old
stuff? Because, as best I can tell, buying ghetto/janky
equipment is most of the reason our stuff goes down so often.
Why so much hate? Do you want us to buy more secondhand or
outdated gear that can't stay up for more than five minutes
at a time? If you're so convinced we don't have a plan,
why don't you suggest one?
\_ stop having a "cpu centric view" instead of a
"what problem are you trying to solve" centric view.
what is the csua's hardware budget? is you want
an i7 to play games, that makes logical sense although
seems politically objectionle, but all this stuff about
virtual machines and zfs etc doesnt make sense.
you arent trying to replicate the ocf on a single
machine.
\_ I don't imagine we'd buy Core i7 Xeons for the
purpose of playing games.
--toulouse
\_ So what are you going to do? Linear algebra
homework?
\_ On the contrary, it's often the shiny stuff that doesn't
have all the bugs worked out. Old/outdated and shiny/new
are not the only options here. I sense a third choice...
\_ Seconded. I was around when we bought a shiny new dual-
Xeon for (the current) soda. FreeBSD didn't support
some of the HW, so now we run Linux and it breaks all
the time.
\_ Seriously. You can go out today and spend < $1000 and
get a brand new computer with quality timetested
hardware that will fulfill csua's hardware needs
10x over. Spend a little more and you can get some
lower end machines to play with/learn how to sysadmin.
\_ Ironically, right now the two machines with the most problems
are keg and scotch. Those are our two remaining FreeBSD
machines. Soda has a month of uptime, which is definitely
not great, but it's certainly not crashing all the time.
I do like FreeBSD but it's certainly not being the all-star
right now. I suspect that scotch has a hardware problem but
keg is unclear whether it's a hardware or software problem.
\_ Ironically, right now the two machines with the most
problems are keg and scotch. Those are our two remaining
FreeBSD machines. Soda has a month of uptime, which is
definitely not great, but it's certainly not crashing all
the time. I do like FreeBSD but it's certainly not being
the all-star right now. I suspect that scotch has a
hardware problem but keg is unclear whether it's a
hardware or software problem.
\- the freebsd machine with problems isnt fbsd7+multicore
is it? we are trying to solve a diffcult problem with
that combination and would be interested in other cases.
although i assume this is ancient hw with bugs. btw in
my fairly long experience linux does worse things on
buggy hw than freebsd. like corrupt data without
detection. this may have gotten better in linux-recent
[i admit linux-recent doesnt drive me insane daily
like it used to]. --psb
\_ I miss those days when soda was a 20 * 386DX-25 machine.
\_ I certainly learned a lot. I remember when it was a big
deal that I had soda up for 100 days in a row. (wow!)
But that setup was due to necessity; I agree that a big
honking hardware purchase isn't really necessary. Start
by trying to get experience developing and sizing a
scalable configuration--that'll be more valuable than
throwing a huge piece of hardware at the problem. -tom
\_ Agreed. Sysadm is NEVER about getting the equipment and
the people you need. You almost NEVER get the resources
you requested for. Sysadm is taking what you have at hand
and try to make the best out of it while not bringing down
the organization.
\_ Scotch - broken. Keg - overloaded, and 16 disks amounting
to about 5TB of data; Lifesaver - dead. Making do with
broken computers is not a viable plan. Buying a server now
means it will be obsolete in a few months, and will be a
waste of money.
waste of money. --toulouse
\- so is buying a $3k computer to do a job
better done by 2 $1k computers.
\_ High end brand new computer components depreciate a
hell of a lot faster than middle of the road workhorse
components. Cheap computer hardware is insanely fast
these days.
\_ Look, it's your network to run, I'm not going to tell
you how to do it. But I think you probably haven't
looked at where your bottlenecks are; does CPU really
matter to the services you're running? I always buy
servers with CPUs two and three generations removed
from the cutting edge, unless the application will
be CPU-bound, which it rarely is. A server doesn't
become obsolete just because someone came out with
a new chipset; it becomes obsolete when it no longer
works for your application. -tom
\_ I rarely agree with tom because he's such a
stubborn ass but this is one time I'll say...
+1 to tom's advice. -anon sysadm
\_ I have to agree with Tom on this. The CPUs and
memory are unlikely to be the bottleneck. --jwm
memory speed are unlikely to be the bottleneck.
A slightly older chip would is likely to be fine
and more stable. --jwm
\_ On the other hand, the CSUA is a place for people
to experiment and try out the new stuff, even if
it is to a certain extent at the cost of
some reliability. -ausman
\_ "we want to 'try out' a really fast
processor because it is cool" isnt a
serious project. this smacks of geekery
not leadership. now if you spent a modest
amount on say a ps3 to get exposure to the
cell, i could see that might be interesting
[eventhough more likely than not such a
purchase would be to "test" video games].
\_ Computers are always obsolete "in a few months".
\_ Intel doesn't release a shiny new architecture that
finally catches up to AMD in many respects every few
months.
\_ Have you been hiding the last few processor
cycles? AMD hasn't exactly been on top of
its game.
\_ What difference does it make? How would the
services be different on a new processor?
Are CSUA people writing assembly code? This
isn't like a new game console where the
supported applications are totally different;
the applications won't care what the underlying
processor architecture is. -tom |
http://csua.com/feed/