MilkyWay@Home: News

Separation updated to 1.00

Wed, 08 Feb 2012 17:36:48 GMT

I've updated all of the separation applications to 1.00. For changes people might care about,

- The old CAL version is gone; it's replaced with the OpenCL application. On AMD/ATI GPUs (older than 79xx) it is using some hackery to use the same IL kernel as before so it should be as fast. However this also means the Radeon 38xx cards aren't supported with new stuff. - Radeon 79xx stuff should work - The occasional validate errors from empty / truncated stderr should stop - AVX will be used if available on Linux and Windows (64 bit only for Windows) - I've increased the default GPU target frequency so GPU stuff should make things less laggy on average. You can now also configure this with the web preferences now so you don't need to use app_info stuff if you want to play with that. - Partial workaround with high CPU usage with recent Nvidia drivers.*

As usual post problems you run into here. * It should cut down on the CPU usage a bit while not sacrificing too much. I would recommend not using it unless you are very unhappy with the CPU usage on Nvidia. There are options to change the polling mode if you want to lower the CPU usage further while not slowing it down. (--gpu-wait-factor (default = 0.75) and --gpu-polling-mode (default = - 2) work similarly to how they did with the old CAL one, but slightly different). With the default of -2 it will use mode -1 unless it is an Nvidia driver newer than the one that introduced the high CPU issue, where it will use mode 0. Mode -1 uses the correct waiting method, mode 0 use the correct waiting method with an initial sleep based on time estimates, and modes > 0 are a polling period in milliseconds. The wait factor is a sort of correction of the time estimate used for the initial wait. The default is 0.75, to wait for 75% of the estimated time before trying to poll.

Remainder of the Double Credits: A Valentine's Day Present

Sun, 05 Feb 2012 03:25:39 GMT

A few weeks ago, we ran double credits in order to make up for a database error that occurred earlier in 2011. Unfortunately, the server crashed with just 12 hours left until the make-up credits were finished. Now that the new server is up and running (and stable! very stable!), and everyone here is back from travels and winter break, we are ready to finish running those last 12 hours of double make-up credits. If there are no significant objections, we'll run the double credits on St. Valentine's Day (February 14), from 10:00 am until 10:00 pm, US Eastern time (EST). This corresponds to 3pm (Feb14)-3am (Feb15) UTC. Happy crunching!

Raised GPU work limits

Sat, 07 Jan 2012 20:03:17 GMT

I've raised the work limit to 20 tasks per GPU to see what happens.

New server test post

Thu, 05 Jan 2012 08:10:11 GMT

Everything should be running on the new server now.

Moving to a new server

Tue, 27 Dec 2011 19:39:05 GMT

Hi Everyone, The new server is in, and we'll be migrating everything over to it this week. We expect to have everything done Tuesday or Wednesday next week. We are expecting to be able to move the entire database over, so we don't expect that you'll lose any work or credit during the transition. Sometime along the way we'll be shutting down the old server for a period of time while we migrate the database over, and redirect milkyway.cs.rpi.edu over to the new machine. We'll be trying to make it as seamless and painless as possible for everyone. Hopefully this will solve a lot of our crashing issues as well. --Travis

N-body updated to 0.80

Fri, 09 Dec 2011 03:27:35 GMT

I've updated all the N-body applications to 0.80. These add a new kind of likelihood calculation (and soon validation method) we're probably going to switch to using that will allow adding GPU versions sometime in the near future (the GPU version is pretty much ready to go, but we need to fiddle with these kinds of issues first). Since the n-body is statistical, for any simulation the result can be anywhere in a distribution. So far for validation we have relied on the results from any 2 system being identical, but this is more problematic when you try to include GPUs. The new results should be "fuzzier" and more resistant to some other types potential problems the old likelihood calculation could have.

New Server

Fri, 02 Dec 2011 21:24:21 GMT

A new server has been ordered. Because delivery times are slow now (due to flooding in Thailand producing slow delivery times for disks), the estimated ship date is December 29, which means we do not expect to get the new server running until the middle to end of January 2012. We will keep watching the current server for repeats of previous software glitches, and hopefully will keep that server running smoothly for the next two months. Thanks, Heidi

Milkyway@home back up

Wed, 30 Nov 2011 21:57:32 GMT

Matthew A was able to get the database stabilized, and it looks like we are running smoothly again. It appears that the SQL query threads were not timing out properly, and were jamming the CPU with worthless tasks. This has been remedied. While we expect to run smoothly for now, other database issues may pop up - the FreeBSD operating system that is currently running Milkyway@home is old, and notoriously incompatible with modern SQL. Of course, we hope to not have more problems. We are ordering a new server, and should transition over to it in mid-December. We'll post the specs once we finalize them, and keep you posted. Cheers, Matthew Newby

Recent Problems with the Database

Tue, 29 Nov 2011 21:08:38 GMT

Excuse us for a while... It appears that the server became unresponsive shortly after everyone left for Thanksgiving break. We've determined that the SQL database is melting down and monopolizing system resources, and we are currently pursuing a solution. If everything goes well, Matthew A. will have us back up tonight. If not, it could be another day or 2 before things stabilize. We are in the process of ordering a new server - we hope to transition over the winter break. We'll keep you posted. Cheers, Matthew N

Separation Status and New Runs

Thu, 17 Nov 2011 00:49:48 GMT

The separation runs that analyse the northern galactic cap (Sloan Digital Sky Survey, or SDSS Stripes 9-27) have nearly finished - soon we'll be able to wrap all of that data together, and with data from Nathan Cole's PhD thesis, we'll have a nice scientific paper out in the next few months. We'll keep you notified. That doesn't mean that the separation code will be retired - far from it. We're starting new separation runs (the ones with "mix" in the name) that are testing the robustness of the separation code. We'll be running simulated data sets along side real data to test several aspects of the stream-fitting process. The main questions are: What would Milkyway@home do if something different from our model exists in the data? And how different would that something have to be in order to modify our results? Also, the release of SDSS Data Release 8 earlier this year gives us access to several stripes of data in the southern galactic cap. We are currently processing this data (It's not as continuous as the northern data, so we have to cut out areas with spotty data), and it will eventually run on Milkway@home. We are also looking to improve the code on Milkyway@home. My brand new paper studies the distribution function of stars in the Milky Way halo, and provides a new convolution kernel that should make Milkyway@home more effective. There are a few other results from that paper, and inside work, that we would like to implement in the Milkyway@home searches. When we update these functions, we will run stripes over again and look for differences. Long story short: We're almost done with one part of Milkyway@home's mission, but there's still plenty to do. Happy Crunching! Cheers - Matthew N