Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
![]() |
World Community Grid Forums
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 6
|
![]() |
Author |
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello,
any 'trick' on how to predict the length (duration) of a WU? I mean something like HCMD2, with the parents, the children and the grandchildren? ![]() As everyone of you, I have WUs running for 3 hrs, many around 5 hrs, and some as long as 11 hrs. Planning on buffering for FAAH isn't so easy as for C4CW! Thanks and keep crunching! |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Switch on checkpoint logging and start counting the number per task. They tend to be constant in number and time and fairly linearly progressing from there, so if you have the time for 3 checkpoints, and you know there are in the batch 18, multiply CPU and Elapsed times.
How to switch on checkpoint logging?... The trick is there to visit the Start Here FAQ index, stickied at top of all forums for the relevant topic. Crunch On. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Sekerob,
I was only able to find this one on checkpoints: http://www.worldcommunitygrid.org/forums/wcg/viewthread?thread=11332 Not exactly what you where indicating I guess. Thanks for the support. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Yes, that's the one. Since the docking can only be short term predicted in duration, non-deterministic calculations and varying per batch, it's a question of counting and correlating. In longer past history we had a thread going on how many there were in a batch and a table... we even had batches that had only 1-2 checkpoints, but also 32 and more.
The interest waned, just keeping tabs on when a checkpoint is made for the best time to boot and of course with a 4-6-8-12 core device, what's the best moment to boot? Nowadays I only look when CEP2 is running, to decide on a time to do a patch update related boot, and then since discovering Ksplice on Linux, no booting is needed unless the kernel is replaced/upgraded.. --//-- |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Very interesting, although useful if the aim is to identify the best moment for rebooting (with 8 threads anyway kind of difficult).
My question was more: how I can predict how long the FAAH wu that now is there 'ready to start' will run? Maybe I missed something. Sorry if it's the case. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
You need to be counting the checkpoints and ''assume'' that the batch will be fairly even in run times or look at the FAAH Crunch Chart's average run times for the past day. The below are those for the past 10 days:
8.04818 13th 7.96126 14th 7.77835 15th 7.69051 16th 7.49098 17th 7.35540 18th 7.35333 19th 7.21433 20th 7.11593 21st 7.36973 22nd Non-deterministic calculation are near impossible to predict accurately, so you'll have to watch and since computers are not all the same, nor OS, it can differ from host to host. Maybe other members have hands on pointers for the present experiment... not crunching FAAH for longer (double sapphire ;>). ciao |
||
|
|
![]() |