Issue - BOINC, Rosetta or ??

Message boards : Number crunching : Issue - BOINC, Rosetta or ??

To post messages, you must log in.

AuthorMessage
Profile The_Saint_(LDS)

Send message
Joined: 12 Aug 10
Posts: 6
Credit: 10,076,132
RAC: 0
Message 73073 - Posted: 16 May 2012, 14:45:29 UTC

Hey all,

One of my small machines (Atom powered Netbook, 1GB RAM, 120GB HD) has been running BOINC on Linux for quite some time (it is a Linux native netbook). Less than a week ago, it went to pot...lots of no progress units, lots of 1 unit active but "waiting to run" (nothing else was running on the machine and I have no restrictions on CPU use).

I tried an upgrade to Ubuntu 12.04 and re-installing the BOINC client and Rosetta. No dice...more of the same. I know it isn't a boatload of power to help, but every little bit helps, right? Just wondering if anyone has any bright ideas...otherwise I'm pulling Rosetta from it.

Thanks.
ID: 73073 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Link
Avatar

Send message
Joined: 4 May 07
Posts: 356
Credit: 382,349
RAC: 0
Message 73074 - Posted: 16 May 2012, 15:24:14 UTC

Look in the BOINC log, it should be written in there, why BOINC suspended those tasks. If I should guess: not enough RAM. But it might be something else, like one of those annoying BOINC v7 issues, see the other threads about it. Unless you really need this version, I'd recommend using one of the later v6 for now (6.12.34, 6.10.60 or even 6.10.18, the last one being IMO the least annoying, but it also depends on which features you need).
.
ID: 73074 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 73077 - Posted: 16 May 2012, 22:50:42 UTC

Yes, I'd tend to agree. What do your BOINC preferences say as to the amount of memory that BOINC is allowed to use? Also the % of page file space.

Some recent R@h tasks are using much more memory than the past. So BOINC may be suspending them, presumably hoping that you allow more memory to be used when the machine is idle, otherwise it won't ever be able to restart the task. I guess the other confusion for BOINC is that on a multi-core system, a task might complete on another core which would then free up memory. I'm not certain how smart BOINC is at determining which tasks will never be able to run in the configured memory preference. As "Link" says, it probably also depends upon BOINC version.

With two CPUs, and 1GB of memory, it seems fairly likely that the combination of a task needing large amounts of memory was unable to proceed.
Rosetta Moderator: Mod.Sense
ID: 73077 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile robertmiles

Send message
Joined: 16 Jun 08
Posts: 1234
Credit: 14,338,560
RAC: 1,227
Message 73087 - Posted: 17 May 2012, 20:37:06 UTC - in response to Message 73077.  

Yes, I'd tend to agree. What do your BOINC preferences say as to the amount of memory that BOINC is allowed to use? Also the % of page file space.

Some recent R@h tasks are using much more memory than the past. So BOINC may be suspending them, presumably hoping that you allow more memory to be used when the machine is idle, otherwise it won't ever be able to restart the task. I guess the other confusion for BOINC is that on a multi-core system, a task might complete on another core which would then free up memory. I'm not certain how smart BOINC is at determining which tasks will never be able to run in the configured memory preference. As "Link" says, it probably also depends upon BOINC version.

With two CPUs, and 1GB of memory, it seems fairly likely that the combination of a task needing large amounts of memory was unable to proceed.


I've seen a recent workunit using more than 1 GB of memory. If the workunits require that none of the memory they use can be paged out, I would expect all workunits of that type to fail on your machine. Even worse if it was trying to run two workunits at once.
ID: 73087 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 73089 - Posted: 17 May 2012, 21:57:09 UTC

Whether pages are swapped out or not is up to the operating system. But the application (R@h is this case) will run more efficiently if enough memory is available. And over time, such a state would lead to excessive disk use.
Rosetta Moderator: Mod.Sense
ID: 73089 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile The_Saint_(LDS)

Send message
Joined: 12 Aug 10
Posts: 6
Credit: 10,076,132
RAC: 0
Message 73096 - Posted: 18 May 2012, 15:06:28 UTC

I had no restriction on the amount of memory BOINC / Rosetta could use while it was running...the machine is a secondary one I use for conferences and travel, so when I'm not out and about it is free to do BOINC / R@H work with everything it's got.

Looking at some of the other units coming through on my other machines and how those other machines have been running them, it looks like my poor little netbook must have pulled every huge unit it possibly could...it simply doesn't have the resources to fill the work unit requirements for this spate of big units. The few bigger ones it could do, it did only run one at a time....but 1GB of RAM just isn't enough for some of these monsters.

Unfortunately, it looks like there is no mechanism in place to get these units back to R@H efficiently. If the work unit is sitting and won't let the machine do other units while that unit is waiting for resources that won't come, it appears it will continue to sit there until the big work unit hits deadline...and all it does is take the machine out of service.

I'll keep my eyes on the files I am getting on the other machines and if/when the big units subside, I'll throw the little machine back into the fray for the project again.
ID: 73096 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Link
Avatar

Send message
Joined: 4 May 07
Posts: 356
Credit: 382,349
RAC: 0
Message 73097 - Posted: 18 May 2012, 16:04:21 UTC - in response to Message 73096.  
Last modified: 18 May 2012, 16:07:36 UTC

I had no restriction on the amount of memory BOINC / Rosetta could use while it was running...

No restriction, i.e. BOINC is allowed to use 100%? Because standard setting is IIRC 60% when in use and 90% when not in use... or something like that.

Also, if you have set BOINC to leave tasks in memory when suspended, it might help to suspend all but one tasks and restart BOINC. If that helps, limit CPU usage to 50% of CPUs (not 50% of time). Since your Atom has just one physical core, that shouldn't decrease your throughput too much.

BTW, have you found the reason in the log why BOINC is suspending those tasks?
.
ID: 73097 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
mikey
Avatar

Send message
Joined: 5 Jan 06
Posts: 1895
Credit: 9,217,610
RAC: 1,154
Message 73108 - Posted: 19 May 2012, 11:17:43 UTC - in response to Message 73096.  

I'll keep my eyes on the files I am getting on the other machines and if/when the big units subside, I'll throw the little machine back into the fray for the project again.


There are LOTS of Boinc Projects that would love your limited power:
http://www.distributedcomputing.info/projects.html

The Boinc ones are noted! Some are short and sweet, some are long and take a lot of memory, you may have to look to see which is best for you. I have 2 gig in this laptop and am crunching the A series units for Malaria, the B ones are bigger and take longer. I have also recently gone back to Eon, the units are short, at around 15 minutes or so and seem to not take alot of memory either.
ID: 73108 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile ]{LiK`RangerS`
Avatar

Send message
Joined: 27 Oct 08
Posts: 39
Credit: 6,552,652
RAC: 0
Message 73230 - Posted: 5 Jun 2012, 18:11:41 UTC - in response to Message 73108.  

world community grid is good for your computer. less ram usage.
ID: 73230 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Message boards : Number crunching : Issue - BOINC, Rosetta or ??



©2024 University of Washington
https://www.bakerlab.org