Advanced search

Message boards : General Topics : Time to withdraw from C@H

Author Message
grn
Send message
Joined: 25 Nov 14
Posts: 3
Credit: 1,403,557
RAC: 0
Message 20598 - Posted: 17 Nov 2015, 19:16:56 UTC

I've been subscribed to this project for a few months but feel like many that there is little information provided about what the contributed processor time is achieving so think it is time to let the current batch of WUs complete and withdraw. It is the only project I am subscribed to where is see 2, sometimes 3 WUs waiting for memory out of the 16 threads on my 16GB system. I have one WU that has been in this state for around 2 hours (either that or it made no progress when it did get memory). I understand that more memory might help performance but surely should be balanced against the resources available. Also, the estimated time for each WU is significanntly lower than the typical run time.. I have a WU that probably has 2 hours still to run and is already an hour longer than the estimate on an initial estimated time of 11 hours 51 minutes. ANd all of this for somewhere around 450 credits. I think I'd rather contribute my compute resources to projects where I get informed what is happening, do not waste the resource by overconsuming the available memory resource and at least award credits to some extent based on runtime as this is the only "benefit" this project seems to provide as it does not give any detail about how this is helping to contribute to the science behind the project.

Jim1348
Send message
Joined: 17 Nov 14
Posts: 48
Credit: 2,358,299
RAC: 0
Message 20599 - Posted: 17 Nov 2015, 19:30:16 UTC - in response to Message 20598.

You need to read the announcements.
http://www.cosmologyathome.org/forum_thread.php?id=7339

Do you have VirtualBox installed? It will automatically download the new Boinc2Docker work units, and they should use all your threads.

grn
Send message
Joined: 25 Nov 14
Posts: 3
Credit: 1,403,557
RAC: 0
Message 20600 - Posted: 17 Nov 2015, 20:29:36 UTC - in response to Message 20599.

Reading the announcements may provide information but I don't see why, like ,ost of the projects, annuncements with links aren't posted through Boinc Manager.

I don't see how Virtual Box helps in any way other than allwoing access to other WUs. I have no problems getting all 8 CPUs and 8 hyperthreads running WUs, the problem when it is running C@H on all 16 threads they are running out of memory and going into a wait for memory state. It is a virtual memory architecture, so this should not really happen (yeah it may get hit by more paging but it should be able to satisfy a memory request). And, of course, the wait sends the WU back to its last checkpoint (again I don't undertsand why that should happen as I've never seen C@H WUs drop back to a checkpoint while running and presumably thes are making memory requests). Nether will VB solve the significant overruns on the initial estimated time nor the fact that C@H somehow manages to hog the machine once it gets running despite preference settings. I've seen a WU finish with no more in the queue, yet managing to get another WU downloaded and running before any other project has had a chance to run.

Profile Ageless
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 15 Jun 07
Posts: 345
Credit: 50,500
RAC: 0
Message 20602 - Posted: 18 Nov 2015, 22:59:16 UTC - in response to Message 20600.

Reading the announcements may provide information but I don't see why, like ,ost of the projects, annuncements with links aren't posted through Boinc Manager.

Only items posted in the News forum will show up in the Notices of BOINC Manager, although this doesn't work that way on all projects. But by default, without admins editing their forum software by hand, News == Notices.
____________
Jord.

pmassey
Send message
Joined: 22 Dec 10
Posts: 1
Credit: 346,147
RAC: 0
Message 20742 - Posted: 20 Jan 2016, 5:46:55 UTC - in response to Message 20599.


Do you have VirtualBox installed? It will automatically download the new Boinc2Docker work units, and they should use all your threads.


Well, that leaves a lot of my systems out then - running more then one hypervisor at the same time is ill advised, plus some of my systems do not support Virtualization, and a few are virtual machines themselves (one such lives on my ESX box)
____________

Message boards : General Topics : Time to withdraw from C@H