Continual Project Stoppage

Panther
Panther
Joined: 23 Feb 18
Posts: 5
Credit: 880377
RAC: 0
Topic 214000

To Everyone:

    I'm a new client, and have noticed that most every time my system's screen goes down after tiime-ing out: einstein try's 2 start, goes back 2 desktop, then program close box. This happens almost everytime it runs. Have allowed 2 run overnight and found next morning having 2 close many "close box" notices some mornings there is over 15 boxes I have to close. I am also running 2 other units, SETI and Milky Way...have no problem with them. I have allotted 100 gigabites for all units to use. Also running a 4 core CPU. 

How may I remedy this annoying problem?

   Thank you,

            Panther    

 

 

 

 

Is there a way 2 allocate disc space thru this application ?

 

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5872
Credit: 117809348495
RAC: 34763306

SteveT_2 wrote:... How may I

SteveT_2 wrote:

...

How may I remedy this annoying problem?

   Thank you,

            Panther

Hi Panther,
Welcome to Einstein@Home.

Apart from mentioning a quad core processor, you haven't given much information to go on.  I presume by "screen goes down after time-ing out" you are referring to the energy saving features that shut off the display?  If so, this should have no effect on task crunching performance.

I don't understand what you mean by, "einstein try's 2 start, goes back 2 desktop, then program close box."  Are you talking about a task starting up on the tasks page of BOINC Manager, advanced view, and then everything immediately crashes back to the desktop?  If not, can you please explain exactly what information you are looking at and what happens in more detail.  What exactly is a "program close box". Is it a dialog box from Windows telling you that Windows had to close the Einstein application?  If so, it probably means that your system is not powerful enough to run the Einstein tasks without tripping some internal limit - temperature, voltage, memory, whatever.

I had a look at the details about your computer available on the website.  I don't know what a Phenom II P960 CPU is so I found some information here.  It doesn't make very encouraging reading for the purpose of an Einstein cruncher.  Einstein tasks need a *lot* of memory per task and put a high load on the system.  Windows itself occupies a lot of RAM so if all 4 cores are tying to crunch, tasks could be swapping on a regular basis particularly if you are using the machine for anything else in any way demanding.  When I look at the crunch times of your returned tasks, the big difference between elapsed time and CPU time is suggestive of things like swapping.  It's certainly an indication of a machine under stress.

I would be concerned in particular about heat generated.  Your machine appears to be a laptop and these don't cope well with the demands of crunching.  How confident are you that the cooling system can cope with the heat produced?  When was the last time you thoroughly cleaned out all the internal passageways and restrictions?  I wouldn't be surprised to find that things shutting down is heat related.

To see if that's true, you could use your BOINC preferences to allow BOINC to use 25% of cores only.  If the machine will work without stopping that way, change it to 50%, and so on until it starts to play up.  It's probably worthwhile to try to improve cooling by dismantling and cleaning as much as you can.

One other thing I noticed whilst looking at your tasks list.  You have a lot of Einstein tasks and I suspect many of those will miss the deadline.  Until your machine is running in a stable and predictable fashion, you should use your work cache settings to limit these large numbers of new tasks.  Set something very low until things are running well.  You should do a rough calculation of how many Einstein tasks can be done before the deadline and abort the rest.  Part of the problem might be that BOINC has already gone into 'panic' mode and is trying to complete as many as possible before the deadline.

If things I've mentioned are not the problem, can you please give more details about exactly what you see happening.  Good luck with sorting it out.

 

Cheers,
Gary.

Panther
Panther
Joined: 23 Feb 18
Posts: 5
Credit: 880377
RAC: 0

Hello Gary, Thank you for

Hello Gary,

Thank you for your info on possible remedies. Yeah I am referring 2 Windows "Dialog Box" that pops btwn each project...only Einstein is doing this "WinDBox" .  I have set limit at 4 project's...running...with screensaver times changing 15 min intervals.  My system has 6G memory, 2T storage, and 8T external.

My laptop is fairly new in that I just went thru everything...total breakdown to case...cleaned, upgraded motherboard, CPU, memory, and have multiple longer life batteries.  Higher output power transformer for electricity. Highest quality fan I could find for my system.

I will try your options. Currently not running any heavy demand processes...and while I am on....their using only 25% memory, and overnite higher % .

Btwn each change, large white box pops up once maybe twice before the WinDBox.  There have been times when Einstein continues with the next unit right after the white boxes, not going to worry 'bout it right now. Just thought it might be something you might have remedies I will try those you suggested.

Thank you Gary,

Panther

  

 

 

 

Is there a way 2 allocate disc space thru this application ?

 

Panther
Panther
Joined: 23 Feb 18
Posts: 5
Credit: 880377
RAC: 0

Hello Gary,     I want 2

Hello Gary,

    I want 2 thank you 4 suggestions on wu processing. Tho now may I suggest a process you attempt 2 observe with anyone you know running Einstein on their laptop...using a sleep timer ? {screen saver}  Allowing wu's 2 run over night...check 2 find any wu stoppage boxes.

  I am running...Einstein_S5R6_3.01_windows_intelx86.exe this being the one and only out there...I wonder if having four telescopes centralized  and only allowing 'star screen' to run at 20 frames per sec...and updates every 5 to 10 secs by each location...slowing down rotation of screen with each...NO wonder process crashes when running the 3 wu's I have going now.

If I may, suggest Einstein's programmer's look to how Seti set up their process on updating...so as 2 hopefully alleviate wu crashes and longer running times...ie the 3 I have running have stated times of 36 hrs and have expanded to almost 48.  These wu are "3cpu" units as stated when dwnld'd. How's that happen ?

  Hopefully those in the know...will help the many out there having this same problem.

Thank you for your time,

  Panther

 

 

 

 

Is there a way 2 allocate disc space thru this application ?

 

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5872
Credit: 117809348495
RAC: 34763306

Quote:SteveT_2 wrote:I want 2

Quote:

SteveT_2 wrote:
I want 2 thank you 4 suggestions on wu processing. Tho now may I suggest a process you attempt 2 observe with anyone you know running Einstein on their laptop...using a sleep timer ? {screen saver}  Allowing wu's 2 run over night...check 2 find any wu stoppage boxes.

Hi Panther,

I was wondering how you were getting on.

I'm not sure why you think running a screensaver might be helpful.  That program is really just 'eye candy' that attracts people into running the project but doesn't play any part in the crunching of tasks and the production of useful scientific results.  To be brutally honest, all it really does is draw pretty pictures on your screen, waste some of your precious CPU cycles, and lower your actual scientific output.  If you want to run it that's fine.  Just realise that it does take away from your potential science output.  Most people who want to maximise their scientific output end up disabling the screensaver.

Quote:
I am running...Einstein_S5R6_3.01_windows_intelx86.exe this being the one and only out there...I wonder if having four telescopes centralized  and only allowing 'star screen' to run at 20 frames per sec...and updates every 5 to 10 secs by each location...slowing down rotation of screen with each...NO wonder process crashes when running the 3 wu's I have going now.

That is not the application that does the science calculations.  I'm not familiar with the Windows apps since I only use Linux.  However I'm certain that anything with S5R6 in the name has to be the graphics app (screensaver).  If I look at your tasks list, I see 27 tasks for the current Gravitational Wave (GW) search where the name is listed as "Continuous Gravitational Wave search O2 All-Sky v1.01 windows_x86_64".  The actual app name is something like einstein_O2AS20-500_1.01_windows_x86_64.exe.

'O2' stands for Observation run #2 (the second observation run with the advanced LIGO detectors), 'AS' stands for All-Sky (the source of GW could be from anywhere rather than from some particular 'target' like a supernova remnant), and the '20-500' stands for the full spin frequency range of any massive object that might be emitting GW.  If you have three tasks running on three individual CPU cores, you should be able to see three instances of this app running and accumulating CPU time.  If you run BOINC Manager (Advanced view) you will be able to see more detail about each running task.

You should open BOINC Manager and select Advanced view.  Then you should go to the tasks tab where you can see what is 'running' and what is 'waiting' to run.  If you look at a running task it will show you the current elapsed time.  Let's say it was 15 hours.  The Progress column will tell you how much has been completed.  Let's say it was 25%.  If you divide the time by the progress (as a decimal fraction) it will tell you how long the task will ultimately take.  For the example it will be 15/0.25=60hrs. This would be a much closer estimate than what you could guess from the remaining time column.

When you complete your first task BOINC will fully correct the original estimate (since it was way less than the true crunch time) and all other tasks will then show a much better estimate.  Unfortunately, BOINC will probably go into 'panic mode' immediately, because it will suddenly realise it can't complete the remaining tasks within the deadline.  If you have tasks for other projects, this is going to play havoc with orderly crunching of work.

Your GW tasks were received around 2.5 days ago but none have been crunched and returned so far.  As I mentioned in my previous message, you should set a smaller work cache setting in your preferences so that you don't get so many.  If none have been completed and returned in 2.5 days, it seems like an impossible task to complete 27 in the remaining 11.5 days before deadline.  You must control the number of tasks you request by lowering your work cache settings.

Quote:
If I may, suggest Einstein's programmer's look to how Seti set up their process on updating...so as 2 hopefully alleviate wu crashes and longer running times...ie the 3 I have running have stated times of 36 hrs and have expanded to almost 48.  These wu are "3cpu" units as stated when dwnld'd. How's that happen ?

There are no "3cpu" tasks coming from Einstein.  Each task requires a single CPU core.  You have settings that can be used to limit the number of CPU cores that can be used and hence the number of tasks that can run simultaneously.  You are the only one that has control of your settings.

The Einstein project staff have already made the tasks as 'small' as possible.  The signals being searched for are computationally difficult to find.  It is not possible to make the tasks use less memory or run faster than they currently do.  A modern CPU running at full speed might be able to crunch these tasks in something like 12-24 hours - even faster for the very latest.

In my previous message, I did warn you that your CPU was unsuited to the crunching load and that you should try limiting the number of cores allowed to crunch.  Did you actually try running only a single task to see if it would run any faster?  Any tasks from that period are long gone from the online database so the only way to know is if you happened to record any of the crunch times from that period.

For GW tasks, the only way to crunch faster is to use a machine with a better CPU and running at full frequency.  Your problem isn't memory or disk space.  Laptop CPUs, particularly for the generation that yours belongs to, are throttled back in frequency to limit heat production.  They are quite unsuited to the rigours of crunching and will always have much longer crunch times than something able to run at full speed.

 

 

 

 

 

Cheers,
Gary.

Comment viewing options

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