Index  | Recent Threads  | Unanswered Threads  | Who's Active  | Guidelines  | Search
 

Quick Go »
No member browsing this thread
Thread Status: Active
Total posts in this thread: 7
[ Jump to Last Post ]
Post new Thread
Author
Previous Thread This topic has been viewed 2560 times and has 6 replies Next Thread
RTS48
Veteran Cruncher
Bolivia
Joined: Aug 2, 2009
Post Count: 1353
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Very variable run times for SCC

I run 3 Mac computers (an iMac quad, a MacBook Pro quad and an old MacBook pro Dual). Generally the WU times are constant at around 2h 30m for the Quads and around 4h for the dual core. Recently, despite the estimated number of Gigaflops being the same, some WUs are suddenly taking ~ 7 hours which is playing havoc with my WU scheduling and requiring manual intervention to prevent 'Too Late". In addition the points yield per hour for these long run WUs drops to around 20 compared with 25 - 30 for the shorter WUs.
How are the run time estimates done and can they be improved so that there are no nasty surprises?

PS Just had 2 WUs that have come in at over 12 hours even though the GFlops are estimated the same as 2h 30m WUs.
----------------------------------------
Rod Peel
Santa Cruz
Bolivia
South America

,
,
----------------------------------------
[Edit 1 times, last edit by RTS48 at Jun 7, 2018 11:28:32 PM]
[Jun 7, 2018 11:39:46 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Speedy51
Veteran Cruncher
New Zealand
Joined: Nov 4, 2005
Post Count: 1326
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Very variable run times for SCC

Have you tried doing a computer restart? Also if you could provide it with unit names this will help people to be able to provide you with assistance. I doubt I can provide you with further assistance as I am not a Mac person
----------------------------------------

[Jun 15, 2018 9:54:35 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: Very variable run times for SCC

I don't know whether the GFLOPS estimates are updated dynamically or manually, but most of the calculations run in these projects are not deterministic, so there will always be some variation. With some projects it is possible to make a reasonable guess based on some of the parameters to the calculations but, again, I don't know about specific projects.

Nevertheless, I was intrigued by the level of variation that you have seen, so I ran a few calculations on the couple of hundred recent results that were still displayed as I run quite a lot of SCC. What I saw was this (all for the same machine, details irrelevant) for CPU time (hours):

Average: 1.48
Std Dev: 0.19
Max: 2.06
Min: 0.98

That's not perfect, but it doesn't seem too bad to me.

I think you were just unlucky. sad

[Edit: Typo in Average, it's 1.48, not 1.44]
----------------------------------------
[Edit 1 times, last edit by Former Member at Jun 15, 2018 11:25:31 PM]
[Jun 15, 2018 11:20:43 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Sgt.Joe
Ace Cruncher
USA
Joined: Jul 4, 2006
Post Count: 7846
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Very variable run times for SCC

I did the same thing as Apis Tintinabulator to see if I could arrive at similar results. I used the last 300 units I completed today which should be a big enough sample to see if there are any outliers.
CPU Time Elapsed Time
Average 2.517 2.518
Median 2.480 2.480
Minimum 1.650 1.650
Maximum 3.580 3.580
Std Dev 0.347 0.347

These were run across a variety of machines, all running 24/7, all running Linux. None of them are MACS. I don't see anything which looks like an outlier. They seem to be pretty steady in their run times. As you can see, my machines are somewhat slower than Apis's machines. Hope this helps.
Cheers
----------------------------------------
Sgt. Joe
*Minnesota Crunchers*
[Jun 16, 2018 1:25:09 AM]   Link   Report threatening or abusive post: please login first  Go to top 
ca05065
Senior Cruncher
Joined: Dec 4, 2007
Post Count: 328
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Very variable run times for SCC

The majority of my work units run for between 1:55 and 2:30. A few weeks ago I had one which ran for over 9 hours. I had been suspicious of it because of the indicated time to completion. I attempted to correct it by LAIM off, suspend work unit, wait a few minutes, resume work unit, LAIM on. It continued to completion, validated and provided a full quota of points.
[Jun 16, 2018 7:03:49 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Speedy51
Veteran Cruncher
New Zealand
Joined: Nov 4, 2005
Post Count: 1326
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Very variable run times for SCC

The majority of my work units run for between 1:55 and 2:30. A few weeks ago I had one which ran for over 9 hours. I had been suspicious of it because of the indicated time to completion. I attempted to correct it by LAIM off, suspend work unit, wait a few minutes, resume work unit, LAIM on. It continued to completion, validated and provided a full quota of points.

I am just curious it did the work unit finish quicker after you suspended and re-enabled the the task?
----------------------------------------

[Jun 22, 2018 10:43:50 PM]   Link   Report threatening or abusive post: please login first  Go to top 
ca05065
Senior Cruncher
Joined: Dec 4, 2007
Post Count: 328
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Very variable run times for SCC

I expected the work unit to speed up after resuming. I assumed it had tied itself in knots and would have a 'fresh start'. However it continued in its slow steady progress to completion.
[Jun 22, 2018 11:05:21 PM]   Link   Report threatening or abusive post: please login first  Go to top 
[ Jump to Last Post ]
Post new Thread