Ok, I've subscribed (hopefully list volume won't kill me :-)
I'm covering several things in this message since I didn't receive the
prior messages in the thread
first off these benchamrks are not being sponsered by my employer, they
need the machines burned in and so I'm going to use them for the tests
while burning them in. I can spend a little official time on this,
justifying it as learning the proper config/tuneing settings for our
project, but not too much. and I'm deliberatly not useing my work e-mail
and am not mentioning the company name, so please respect this and keep
the two seperate (some of you have dealt with us formally, others will
over the next few months)
this means no remote access for people (but I am willing to run tests and
send configs around). in fact the machines will not have Internet access
for the duration of the tests.
it also means I'm doing almost all the configuration work for this on my
own time (nights and weekends). the machines will not be moved to
production for a couple of months. this should mean that we can go back
and forth with questions and answers (albeit somewhat slowly, with me
checking in every night) while whatever tests are done happen during the
day. once we get past the system tuneing and start doing different tests
it would probably be helpful if people can send me scripts to run that I
can just let loose.
I don't have any money to pay for benchmark suites, so if things like the
TPC benchmarks cost money to do I won't be able to do them
to clarify the hardware
I have 5 machines total to work with, this includes client machines to
make the queries (I may be able to get hold of 2-3 more, but they are
similar configs)
none of these have dual-core processors on them, the CPU's are 246 or 252
Opterons (I'll have to double check which is in which machine, I think the
large disk machine has 246's and the others 252's)
I have access to a gig-E switch that's on a fairly idle network to use to
connect these machines
the large-disk machine has 3ware 9500 series 8-port SATA controllers in
them with battery backup. in our official dealings with Greenplum we
attempted to do a set of benchmarks on that machine, but had horrible
timing with me being too busy when they worked with us on this and we
never did figure out the best setting to use for this machine.
Part of the reason I posted this to /. rather then just contacting you and
MySQL folks directly is that I would like to see a reasonable set of
benchmarks agreed to and have people with different hardware then I have
run the same sets of tests. I know the tuneing will be different for
different hardware, but if we can have a bunch of people run similar tests
we should learn a lot.
David Lang
