Suspend Einstein.

John Bowman
John Bowman
Joined: 15 Nov 05
Posts: 8
Credit: 220,006
RAC: 0
Topic 190172

I use BOINC 5.2.7 crunching SETI and Climate Prediction data.
Einstein looked fun so I attached it today. It downloaded and appears to run fine but when I 'Suspend' it in BOINC Manager another version of the data set appears in the BOINC Work tab and starts to run. Any ideas how to stop this?

John B

Jord
Joined: 26 Jan 05
Posts: 2,952
Credit: 5,874,644
RAC: 7,826

Suspend Einstein.

Where do you suspend it? In the Work tab or in the Projects tab?

If you suspend it in the Work tab, only that work unit will be suspended. It will download a new work unit for the project itself.

So if you go to the Work tab of BOINC Manager and select the Einstein@Home project (leftclick once on the EAH entry) then click the Suspend button on the left, the whole EAH project will suspend, including any work units under the Work tab.

John Bowman
John Bowman
Joined: 15 Nov 05
Posts: 8
Credit: 220,006
RAC: 0

Jord, That was fast! I

Message 19418 in response to message 19417

Jord,

That was fast!

I suspended it using the 'Work'tab. However, the work unit that it then started to process had the same ID as the one I'd suspended.

I will try suspending in the 'Project' tab to see how that works.

Thanks,

John B

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5,870
Credit: 115,756,276,017
RAC: 35,111,893

RE: I suspended it using

Message 19419 in response to message 19418

Quote:
I suspended it using the 'Work'tab. However, the work unit that it then started to process had the same ID as the one I'd suspended.

Hi John B, welcome to the Einstein project.

As Jord said, suspemding in the "Work" tab suspends only that particular result and another will be downloaded in its place. If you check closely it will have a very similar name but it wont be exactly the same. Suspending the project under the "Projects" tab will suspend all work in progress from the Einstein project.

Your list of computers on the website shows five entries. I suspect you don't actually have 5 physical boxes. One of those shows three results with one aborted soon after it started crunching. If you would like some help with getting this sorted out please just ask.

Cheers,
Gary.

John Bowman
John Bowman
Joined: 15 Nov 05
Posts: 8
Credit: 220,006
RAC: 0

Gary, Thanks for your

Message 19420 in response to message 19419

Gary,

Thanks for your reply and the welcome.

I'm afraid my list of computers IS wrong though I wish it wasn't! Now that I've finished crunching my first WU and uploaded the result I'd be pleased if the erroneous items could be deleted.

Thanks,

John B.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5,870
Credit: 115,756,276,017
RAC: 35,111,893

RE: I'm afraid my list of

Message 19421 in response to message 19420

Quote:
I'm afraid my list of computers IS wrong though I wish it wasn't! Now that I've finished crunching my first WU and uploaded the result I'd be pleased if the erroneous items could be deleted.

Hi John B,

Here are some instructions which should allow you to consolidate your different computer identities (CPUIDs) into just one. I'm assumimg you only have one real machine. Let me know if that assumption is incorrect.

Before we do this, I note you mentioned that you were adding Einstein to the existing mix of Seti and CPDN. I assume you have set up the resource shares that you would like each project to have. If you haven't set up what you want, you need to do this by editing your preferences on the website. If you have left the defaults, each project will have its resource share set at 100 so each will get 33.33% of your computer's available time. This is probably a good way to go while you are learning.

A good lesson to remember is that BOINC is designed to follow your preferences. You should give BOINC as much time as it needs to do that without trying to force it. BOINC will get it right. Hence your previous actions of needing to suspend a project shouldn't really be necessary. If you think you would like a particular project to run more often, just increase the resource share rather than wielding the big stick and suspending the others. Another thing to remember while getting the feel of things is to keep a preference called "connect to network every X days" at a fairly small value and don't make sudden big changes. With a mix of three projects I would recommend the default value of 0.1 days as this will allow BOINC to settle into its routine as quickly as possible. You can then increase it gradually if you wish to see a bit more work kept on hand.

When changing preferences, many people get impatient and try to force the issue. Once you make a change on the website of one project, it might take a day or two to filter through to all the projects. You can speed things up by changing and manually "updating" for every single website if you really want the change to happen quickly. The effect of some changes is counter-intuitive and there are many examples of people stuffing themselves up by excessive "micromanaging". With BOINC, patience is an extreme virtue.

Once you have confirmed that you are happy with your preferences, here's what you can do to fix your CPUIDs. Check the ones you have on the website. You have these values (5 all told):-

  • * 440919
    * 441008
    * 441023
    * 441027
    * 441031

#441031 is the CPUID that has a validated result so it would be the "real" machine. Start with your oldest duplicate which is #440919. Click it and on the page that comes up you will see the "merge" option right down the bottom. If you click on "merge" you will be offered a choice of "compatible" CPUIDs to merge it to. You need to choose your "real" machine #441031, so select that one. All you need to do is keep repeating this sequence until all your excess CPUIDs have been amalgamated into the current one.

If any of this is unclear, please ask before doing something you are not sure about. One final point, some of the CPUIDs have 0 results in their results list. You can actually delete those rather than merging if you wish. You cannot delete a machine with results showing - just merge it. As a result of the merge, you may acquire a couple of extra results in your final results list. That is a good thing as they would otherwise be wasted and the system would have to wait 14 days before recycling them to someone else. This way you will be helping to clean things up more quickly.

Cheers,
Gary.

John Bowman
John Bowman
Joined: 15 Nov 05
Posts: 8
Credit: 220,006
RAC: 0

All done. Thanks,

Message 19422 in response to message 19421

All done. Thanks, Gary.

John

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5,870
Credit: 115,756,276,017
RAC: 35,111,893

Great!! That looks much

Great!! That looks much better and it has done exactly what it was supposed to do. You have acquired the extra results so nothing has been wasted. Good job!!

In looking at your results, I'm a little surprised that your first one took as long as it did to crunch. You have an Athlon 64 3200+ so it ought to be able to do a result in 20,000 to 21,000 seconds, I would imagine. If your results keep taking around 28,000 seconds, there must be something that needs tweaking. As a comparison I have an Athlon XP 2200+ which regularly does results in around 26,500 seconds and an XP 3200+ that takes around 21,000 seconds.

Also, as you mentioned that you crunch for Seti, what sort of times do you get there? As an example, I have an XP 2700+ that does Seti results in about 8,000 seconds (depends on angle range). It's running an optimised science app so that speeds it up a bit. Your Athlon 64 would have SSE2 instructions, I think, so should show a good speedup (better than Athlon XP) with a suitable optimised science app. Do you know about the optimised Seti science apps that are available? They are very easy to install.

Cheers,
Gary.

Michael Roycraft
Michael Roycraft
Joined: 10 Mar 05
Posts: 846
Credit: 157,718
RAC: 0

RE: RE: Great!! That

Message 19424 in response to message 19423

Quote:

Quote:

Great!! That looks much better and it has done exactly what it was supposed to do. You have acquired the extra results so nothing has been wasted. Good job!!

In looking at your results, I'm a little surprised that your first one took as long as it did to crunch. You have an Athlon 64 3200+ so it ought to be able to do a result in 20,000 to 21,000 seconds, I would imagine. If your results keep taking around 28,000 seconds, there must be something that needs tweaking. As a comparison I have an Athlon XP 2200+ which regularly does results in around 26,500 seconds and an XP 3200+ that takes around 21,000 seconds.

Also, as you mentioned that you crunch for Seti, what sort of times do you get there? As an example, I have an XP 2700+ that does Seti results in about 8,000 seconds (depends on angle range). It's running an optimised science app so that speeds it up a bit. Your Athlon 64 would have SSE2 instructions, I think, so should show a good speedup (better than Athlon XP) with a suitable optimised science app. Do you know about the optimised Seti science apps that are available? They are very easy to install.

I would expect that John's A64 3200+ will be doing Einstein WUs in the high 24000 sec bracket, since it clocks at 2.0 GHz. Einstein times seem to be directly proportionate (on any given platform) to raw CPU speeds. Keep in mind that he is using the 0.18 app, and the 8-10% time increase that Athlons see with it. Perhaps there is another process running, like AntiVirus or screensaver, that is taking CPU cycles away.

Michael


microcraft
"The arc of history is long, but it bends toward justice" - MLK

John Bowman
John Bowman
Joined: 15 Nov 05
Posts: 8
Credit: 220,006
RAC: 0

Gary, The Seti data is

Message 19426 in response to message 19423

Gary,

The Seti data is averaging 9200 secs to crunch with a best of 8400 secs.

I work from home and run Einstein and SETI in the background whilst using the PC for work or pleasure.

I run some memory intensive graphical programs and am ussually connected to broadband for music. Firewalls, anti-virus and SpywareBlaster also run in the background. Does this explain the relatively slow processing or is there something else I could try?

Cheers,

John

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5,870
Credit: 115,756,276,017
RAC: 35,111,893

RE: Einstein times seem to

Message 19427 in response to message 19424

Quote:

Einstein times seem to be directly proportionate (on any given platform) to raw CPU speeds.

Yes, you're right about that. I don't have any A64 boxes so I was thinking that a 3200+ should be about the same as an XP 3200+ - my mistake :).

Quote:
Keep in mind that he is using the 0.18 app, and the 8-10% time increase that Athlons see with it.

Yes, indeed.

Quote:
Perhaps there is another process running, like AntiVirus or screensaver, that is taking CPU cycles away.

Undoubtedly, and you could put the "real" work that the machine does in that same category but none of these should really affect crunch time. If a result took 6 hours of cpu to crunch and if the machine was 50% occupied with real work shouldn't the cpu time still report as 6 hours but be achieved after 12 hours of wall clock time? I still reckon there must be something that needs a bit of tweaking. After all if an XP2200+, not overclocked and with very average mobo and RAM can do a result in 26,500, the A64 has just got to be well under that.

Out of interest, I just went and found an A64 3200+ in the top computers list. So as not to get any inflated figures I headed down towards the lower half of the list and found this one. 22,200 seconds sounds about right to me. Add on 10% for the 0.18 beta app and 24,500 seconds should be the number to aim for. John B should be able to save about sn hour of crunch time per result.

EDIT:
Purely by chance I've just run into this Athlon 64 3200+ where the owner is having some "Client error" problems. The times are around 18,200 seconds so I'm guessing that the machine must be rather overclocked to be doing results that fast.

Cheers,
Gary.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.