rpi_logo
Making Milkyway@home Better
Making Milkyway@home Better
log in

Advanced search

Message boards : News : Making Milkyway@home Better

Author Message
Matthew
Volunteer moderator
Project developer
Project scientist
Send message
Joined: 6 May 09
Posts: 217
Credit: 6,856,375
RAC: 0

Message 51396 - Posted: 12 Oct 2011, 21:45:36 UTC

Recently Milkway@home has seen some major crashes, and there have been stability issues popping up. There have been plenty of rumors running around the message boards as to what the issue is. There has been discussion, from both Milkyway crunchers and staff, of buying a new server. The Milkyway staff has been discussing this issue for awhile, and we've recently come to some conclusions.

There's a strong consensus that the Milkyway server hardware is not to blame. It appears that the OS is buggy, being fairly old and forced to run modern database software (among other things).

In order to update the OS (and other software), we would need a new server so that the old one could continue to run during the install and upgrade process. We have access to funding for a new server, but the issue is that we do not have the staff to oversee the installation of BOINC/MW and the eventual migration. Especially since our computing support staff has seen major cutbacks in the recent past.

The donations from Milkyway crunchers have already funded one student's work on Milkyway@home during the last summer. In addition to funding that student, donations have just recently funded a Milkyway development machine which will soon be capable of compiling all of the Milkyway client software across multiple architectures simultaneously. We take your donated money seriously, and will ONLY use it to fund students or Milkyway-related hardware purchases.

So while we appreciate the recent drive by our users to donate or raise money for new hardware, it is actually support staff that we need the most. Direct monetary donations to Milkyway will be the most useful to us, and can be made through our donation page.

We are looking to hire a part-time computer scientist to clean up the server, work on server stability, and create a better experience for our crunchers. This person might also assist in migrating us over to a new server, when the time comes. We only need to raise a few hundred more dollars to meet the salary goal. We also need to raise money for next summer's salaries.

ALSO, we will be creating a hardware donation page similar to SETI@home's. Matt A. needs a few GPU's to continue his dev work, and he is also looking to build a twin to the new Milkyway dev machine. I'll let him provide the details, and all of this will be posted on the new page. Look for it in the next few days.

Donations of old GPUs will always be useful and welcome to Milkyway@home. When the hardware donation page is up, we will provide an address to which these may be sent. I believe that all donations will be tax-deductible.

Thank you for providing your valuable cycles to us, as they provide us with the ability to crunch enormous amounts of data and make Milkyway@home one of the top BOINC projects. This is all we really need to keep Milkyway@home running and successful. But if you would like to help make Milkyway@home BETTER, please consider making a small donation to help support our staff.

Cheers,
Matthew N

Andrew McFarland
Avatar
Send message
Joined: 21 Sep 11
Posts: 41
Credit: 185,458,736
RAC: 0

Message 51399 - Posted: 12 Oct 2011, 22:32:23 UTC - in response to Message 51396.

There's a strong consensus that the Milkyway server hardware is not to blame. It appears that the OS is buggy, being fairly old and forced to run modern database software (among other things).

Perhaps. I'm not a DB specialist, but methinks that that modern database may want more than -->8GB<--.

We have access to funding for a new server, but
...[snip]...
We only need to raise a few hundred more dollars to meet the salary goal."

Hmmm. You have 3 fresh years of funding and enough cash to turn your nose up at a $3k modern server, but you need only hundreds to pay the temp? Call me a skeptic, but people have been donating cash for a while now, and money doesn't seem to be the problem. I'm concerned that the goodwill may be wearing out. Is it really going to be different this time? Maybe it would help if you gave us a concrete dollar figure to work towards, instead of the usual donate-into-a-black-hole.

Matt A. needs a few GPU's to continue his dev work...


Matt, check you inbox.


Ok. I'm breathing deep now. Sorry. I only get wound up because I like this project and want to help. Can I at least interest you in just a little more memory for the tired old Xeon ;-)

Keep up the good work.

Matthew
Volunteer moderator
Project developer
Project scientist
Send message
Joined: 6 May 09
Posts: 217
Credit: 6,856,375
RAC: 0

Message 51403 - Posted: 13 Oct 2011, 2:49:34 UTC

Perhaps. I'm not a DB specialist, but methinks that that modern database may want more than -->8GB<--.


I'd like more memory, too, but we've rarely seen it full; as of right now MW is using 5.5GB.

You have 3 fresh years of funding [...]


I'm not sure what you're referring to? Do you mean the grant that Heidi was awarded a while ago? Government grants always come with strict stipulations for spending, and allow little or no flexibility (X dollars for Y number of grad students, Z dollars for conference fees, B dollars for hardware, etc.). So all of that funding is technically allocated already.

The problem is that our support staff is provided by RPI, and not the grant, and RPI has been downsizing the support staff. Since the grant is already set in stone, we do not have the ability to reallocate the funds to provide the man-power we wish we had.

However, funds donated to Milkyway are not subject to strict oversight, and can therefore be allocated to what we need with fairly short warning. I hope that explains our predicament clearly.

[...]concrete dollar figure to work towards[...]


I'll ask around and see if I can get one.

Cheers,
Matthew Newby

Jimmy Gondek
Send message
Joined: 28 Sep 11
Posts: 60
Credit: 22,764,173
RAC: 0

Message 51516 - Posted: 26 Oct 2011, 14:21:00 UTC

I'll plead complete ignorance as to how computational project allocations are determined at RPI but it would seem to me that a "tiny" project like MWAH might fit rather insignificantly, cycle-wise, into the infrastructure mentioned here...?

HPCwire: Rensselaer Orders Up Blue Gene/Q for Exascale and Data-Intensive Research:
http://www.hpcwire.com/hpcwire/2011-10-25/rensselaer_orders_up_blue_gene_q_for_exascale_and_data-intensive_research.html?featured=top

...It looks like there is going to be a lot of room available on the slightly older systems as the high-priority projects migrate to the newer systems? A possible solution here for these lingering/continuing server issues once and for all?

Just thinking big... :)
JG


Post to thread

Message boards : News : Making Milkyway@home Better


Main page · Your account · Message boards


Copyright © 2018 AstroInformatics Group