Question on BOINC versions


log in

Advanced search

Message boards : Number crunching : Question on BOINC versions

1 · 2 · Next
Author Message
Profile ccappel
Avatar
Send message
Joined: 27 Jan 00
Posts: 362
Credit: 1,516,412
RAC: 0
United States
Message 969246 - Posted: 8 Feb 2010, 15:15:55 UTC

I've been sitting on version 6.6.36 for months now and was wondering if an upgrade to 6.10.x would be to my benefit? I'm not running CUDA and I'm using all optimized apps. What features might I be missing out on?
____________
"Life is a tragedy for those who feel, and a comedy for those who think."

"I never get into an argument that I cannot win."

Profile dnolan
Avatar
Send message
Joined: 30 Aug 01
Posts: 1245
Credit: 43,761,815
RAC: 43,607
United States
Message 969247 - Posted: 8 Feb 2010, 15:24:16 UTC

You'll probably hear other opinions, but if I wasn't doing GPU processing, I'd be using 5.10.45 (as I am on my one non-GPU machine).
I see no reason why you need to update.

-Dave
____________

Profile HAL9000
Volunteer tester
Avatar
Send message
Joined: 11 Sep 99
Posts: 3571
Credit: 98,153,926
RAC: 76,632
United States
Message 969256 - Posted: 8 Feb 2010, 15:55:47 UTC - in response to Message 969246.

I've been sitting on version 6.6.36 for months now and was wondering if an upgrade to 6.10.x would be to my benefit? I'm not running CUDA and I'm using all optimized apps. What features might I be missing out on?

There is a libcurl.dll fix in one of the new versions. I don't recall if it was 6.10.18 or if it was 6.10.2x. Also in the tasks list there is an option to only display active tasks. I think 6.10.xx versions also have "don't pound the crap out of the servers" code in them.
____________
SETI@home classic workunits: 93,865 CPU time: 863,447 hours

Join the BP6/VP6 User Group today!

Michael Goetz
Avatar
Send message
Joined: 14 May 99
Posts: 55
Credit: 587,625
RAC: 0
United States
Message 969281 - Posted: 8 Feb 2010, 17:26:55 UTC - in response to Message 969246.

I've been sitting on version 6.6.36 for months now and was wondering if an upgrade to 6.10.x would be to my benefit? I'm not running CUDA and I'm using all optimized apps. What features might I be missing out on?


It's not a factor with SETI, but on some other projects it's helpful to know when the application is (or is not) doing checkpoints. 6.10.18 has a properties button that allows you to see when the last checkpoint occurred, amongst other information. Yeah, you can have checkpointing listed in the log if you want, but that's a LOT of log messages if you're checkpointing frequently.

Also, you can do multiple selections using shift-click or control-click. That's very handy for those "oops, I didn't really want to get those 500 WUs" types of situations.

Profile [B^S] madmac
Volunteer tester
Avatar
Send message
Joined: 9 Feb 04
Posts: 1133
Credit: 3,098,080
RAC: 3,400
United Kingdom
Message 969298 - Posted: 8 Feb 2010, 19:02:21 UTC

I have been using 6.6.36 for awhile now and I see no point of upgrading even though I am getting messages to upgrade. So I will wait until we have to upgrade with a new optimised version comes out which will only work with the new versions of BOINC
____________

1mp0£173
Volunteer tester
Send message
Joined: 3 Apr 99
Posts: 8423
Credit: 356,897
RAC: 0
United States
Message 969300 - Posted: 8 Feb 2010, 19:04:25 UTC - in response to Message 969256.
Last modified: 8 Feb 2010, 19:06:34 UTC

I've been sitting on version 6.6.36 for months now and was wondering if an upgrade to 6.10.x would be to my benefit? I'm not running CUDA and I'm using all optimized apps. What features might I be missing out on?

There is a libcurl.dll fix in one of the new versions. I don't recall if it was 6.10.18 or if it was 6.10.2x. Also in the tasks list there is an option to only display active tasks. I think 6.10.xx versions also have "don't pound the crap out of the servers" code in them.

The libcurl fix does away with all the ugly mucking around with the "hosts" file when a project has more than one server listed in DNS but one of them is down.

... and the "don't pound the crap out of the servers" code will mean quicker recoveries once enough people get on board.
____________

Profile HAL9000
Volunteer tester
Avatar
Send message
Joined: 11 Sep 99
Posts: 3571
Credit: 98,153,926
RAC: 76,632
United States
Message 969301 - Posted: 8 Feb 2010, 19:05:10 UTC - in response to Message 969281.

I've been sitting on version 6.6.36 for months now and was wondering if an upgrade to 6.10.x would be to my benefit? I'm not running CUDA and I'm using all optimized apps. What features might I be missing out on?


It's not a factor with SETI, but on some other projects it's helpful to know when the application is (or is not) doing checkpoints. 6.10.18 has a properties button that allows you to see when the last checkpoint occurred, amongst other information. Yeah, you can have checkpointing listed in the log if you want, but that's a LOT of log messages if you're checkpointing frequently.

Also, you can do multiple selections using shift-click or control-click. That's very handy for those "oops, I didn't really want to get those 500 WUs" types of situations.

I've been able to do that since I started crunching again in July with 6.6.something .28 maybe?
____________
SETI@home classic workunits: 93,865 CPU time: 863,447 hours

Join the BP6/VP6 User Group today!

Sten-Arne
Volunteer tester
Send message
Joined: 1 Nov 08
Posts: 3307
Credit: 16,366,213
RAC: 16,254
Sweden
Message 969310 - Posted: 8 Feb 2010, 19:38:17 UTC - in response to Message 969301.

I've been sitting on version 6.6.36 for months now and was wondering if an upgrade to 6.10.x would be to my benefit? I'm not running CUDA and I'm using all optimized apps. What features might I be missing out on?


It's not a factor with SETI, but on some other projects it's helpful to know when the application is (or is not) doing checkpoints. 6.10.18 has a properties button that allows you to see when the last checkpoint occurred, amongst other information. Yeah, you can have checkpointing listed in the log if you want, but that's a LOT of log messages if you're checkpointing frequently.

Also, you can do multiple selections using shift-click or control-click. That's very handy for those "oops, I didn't really want to get those 500 WUs" types of situations.

I've been able to do that since I started crunching again in July with 6.6.something .28 maybe?


That was possible even when I had 6.2.19, by selecting "Grid View"

Sten-Arne

Aurora Borealis
Volunteer tester
Avatar
Send message
Joined: 14 Jan 01
Posts: 2975
Credit: 4,814,242
RAC: 1,285
Canada
Message 969348 - Posted: 9 Feb 2010, 0:49:13 UTC
Last modified: 9 Feb 2010, 0:54:15 UTC

A new feature in 6.10.32 allows for a true back up project. Setting a project Resource Share at 0 will tell Boinc to only ask that project for work if a CPU become idle. It's still to be tested in the real world since older server software won't let you enter 0 for RS although .001 may be read by Boinc as 0.

John McLeod VII
Volunteer developer
Volunteer tester
Avatar
Send message
Joined: 15 Jul 99
Posts: 23702
Credit: 493,921
RAC: 159
United States
Message 969397 - Posted: 9 Feb 2010, 4:46:40 UTC - in response to Message 969348.

A new feature in 6.10.32 allows for a true back up project. Setting a project Resource Share at 0 will tell Boinc to only ask that project for work if a CPU become idle. It's still to be tested in the real world since older server software won't let you enter 0 for RS although .001 may be read by Boinc as 0.

Since the resource share is a double, 0.0001 will be read as 0.0001 and not 0.
____________


BOINC WIKI

Aurora Borealis
Volunteer tester
Avatar
Send message
Joined: 14 Jan 01
Posts: 2975
Credit: 4,814,242
RAC: 1,285
Canada
Message 969403 - Posted: 9 Feb 2010, 5:22:20 UTC - in response to Message 969397.
Last modified: 9 Feb 2010, 5:30:09 UTC

A new feature in 6.10.32 allows for a true back up project. Setting a project Resource Share at 0 will tell Boinc to only ask that project for work if a CPU become idle. It's still to be tested in the real world since older server software won't let you enter 0 for RS although .001 may be read by Boinc as 0.

Since the resource share is a double, 0.0001 will be read as 0.0001 and not 0.

Your are correct larger values didn't work, although showing 0 in the projects page, but testing on Spinhenge, SZTAKI and LHC with a value of .0000001 gave me <resource_share>0.000000</resource_share> in the client_state.xml
____________
Questions? Answers are in the "Unofficial" BOINC Wiki.

Boinc V7.0.27
Win7 i5 3.33G 4GB, GTX470

Profile X-Files 27
Avatar
Send message
Joined: 17 May 99
Posts: 100
Credit: 107,862,964
RAC: 0
Canada
Message 969532 - Posted: 10 Feb 2010, 5:13:03 UTC - in response to Message 969397.

A new feature in 6.10.32 allows for a true back up project. Setting a project Resource Share at 0 will tell Boinc to only ask that project for work if a CPU become idle. It's still to be tested in the real world since older server software won't let you enter 0 for RS although .001 may be read by Boinc as 0.

Since the resource share is a double, 0.0001 will be read as 0.0001 and not 0.

Even though its read as 0.0001 and not 0, the backup project function as it should. Like today for example, i ran out of seti wus, gpugrid as my backup project kicks in.
____________

Aurora Borealis
Volunteer tester
Avatar
Send message
Joined: 14 Jan 01
Posts: 2975
Credit: 4,814,242
RAC: 1,285
Canada
Message 969546 - Posted: 10 Feb 2010, 8:36:59 UTC - in response to Message 969532.

A new feature in 6.10.32 allows for a true back up project. Setting a project Resource Share at 0 will tell Boinc to only ask that project for work if a CPU become idle. It's still to be tested in the real world since older server software won't let you enter 0 for RS although .001 may be read by Boinc as 0.

Since the resource share is a double, 0.0001 will be read as 0.0001 and not 0.

Even though its read as 0.0001 and not 0, the backup project function as it should. Like today for example, i ran out of seti wus, gpugrid as my backup project kicks in.

True, as was argued on the mailing list as well as many forums, a large differential in RS acts as a work around, but it isn't as effective as a true back up project. You can still end up with more work than desired and that can delay getting work from your main project(s). If the new algorithm works properly, you should get only enough work to keep the CPU(s) warm and the BU project wont hog the system until all it's WUs have been processed.

Matthew S. McCleary
Avatar
Send message
Joined: 9 Sep 99
Posts: 121
Credit: 2,288,242
RAC: 0
United States
Message 969571 - Posted: 10 Feb 2010, 14:19:58 UTC

Is there any one version which stands out for particularly good performance, in terms of work done for CPU/GPU time expended?
____________

1mp0£173
Volunteer tester
Send message
Joined: 3 Apr 99
Posts: 8423
Credit: 356,897
RAC: 0
United States
Message 969595 - Posted: 10 Feb 2010, 16:59:20 UTC - in response to Message 969571.

Is there any one version which stands out for particularly good performance, in terms of work done for CPU/GPU time expended?

There shouldn't be any difference in terms of work done between any version of BOINC with "reasonable" settings (or even fairly unreasonable ones).

Some of the older versions had trouble following resource shares if you're doing more than one project and had deadlines and/or carried a big cache. These versions put too much priority on keeping the cache full, and a project that always had work would get more than their allocated resource share.

Some work is being done to avoid starting a bunch of work units for a given project when there is deadline pressure -- this is mostly cosmetic. Mostly.

Later versions are easier on the servers -- some think that being slower to report is a huge issue, but it cuts down on retries and wasted resources at the server end.

... and the libcurl fix takes care of the continuous failed uploads or downloads when one of the servers is out of service but another one can handle the request.
____________

Profile ccappel
Avatar
Send message
Joined: 27 Jan 00
Posts: 362
Credit: 1,516,412
RAC: 0
United States
Message 969781 - Posted: 11 Feb 2010, 15:15:53 UTC

Thanks all for the (as usual) knowledgeable information. I think I will go ahead and upgrade for the libcurl fix for myself and the "don't pound the servers" for the good of the project.
____________
"Life is a tragedy for those who feel, and a comedy for those who think."

"I never get into an argument that I cannot win."

Richard Haselgrove
Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 8275
Credit: 44,981,343
RAC: 13,875
United Kingdom
Message 969785 - Posted: 11 Feb 2010, 15:46:26 UTC - in response to Message 969781.

Thanks all for the (as usual) knowledgeable information. I think I will go ahead and upgrade for the libcurl fix for myself and the "don't pound the servers" for the good of the project.

Remember that the libcurl fix first appeared in v6.10.22 - but that version was a no-no for other reasons.

If the purpose of upgrading is libcurl, you're pretty much limited to v6.10.25 or v6.10.32 - and there's a general bug in the entire v6.10.xx range, fixed but not yet deployed, so v6.10.32 won't be the end of the development line. (Just don't use <ncpus>0</ncpus> - that's the only problem I know of).

Profile HAL9000
Volunteer tester
Avatar
Send message
Joined: 11 Sep 99
Posts: 3571
Credit: 98,153,926
RAC: 76,632
United States
Message 969790 - Posted: 11 Feb 2010, 16:22:46 UTC - in response to Message 969785.

Thanks all for the (as usual) knowledgeable information. I think I will go ahead and upgrade for the libcurl fix for myself and the "don't pound the servers" for the good of the project.

Remember that the libcurl fix first appeared in v6.10.22 - but that version was a no-no for other reasons.

If the purpose of upgrading is libcurl, you're pretty much limited to v6.10.25 or v6.10.32 - and there's a general bug in the entire v6.10.xx range, fixed but not yet deployed, so v6.10.32 won't be the end of the development line. (Just don't use <ncpus>0</ncpus> - that's the only problem I know of).

I've had v6.10.25 on most of my machines for a while. It is a pretty stable release IMO.
____________
SETI@home classic workunits: 93,865 CPU time: 863,447 hours

Join the BP6/VP6 User Group today!

Profile Dr. Bob
Avatar
Send message
Joined: 1 Apr 03
Posts: 78
Credit: 430,831
RAC: 873
United States
Message 969799 - Posted: 11 Feb 2010, 17:12:00 UTC

Good morning all...

There is an email coming to many Boinc users. A survey but they want your email and passwords...DON'T GIVE THEM THAT---DIRECTOR DR. ANDERSON NEEDS TO LOOK AT THAT; HE SIGNED IT...I AM SUSPICIOUS...I WILL NOT GIVE THAT INFO; SETI ALREADY HAS THAT INFO...

SOMEONE NEEDS TO ADDRESS THIS RIGHT AWAY BEFORE PEOPLE FALL FOR IT. IF IT IS REAL, IT IS DONE VERY POORLY!

Best,

mizzou
____________
Robert L. Hanson, Ed.D.

Profile HAL9000
Volunteer tester
Avatar
Send message
Joined: 11 Sep 99
Posts: 3571
Credit: 98,153,926
RAC: 76,632
United States
Message 969802 - Posted: 11 Feb 2010, 17:19:28 UTC - in response to Message 969799.

Good morning all...

There is an email coming to many Boinc users. A survey but they want your email and passwords...DON'T GIVE THEM THAT---DIRECTOR DR. ANDERSON NEEDS TO LOOK AT THAT; HE SIGNED IT...I AM SUSPICIOUS...I WILL NOT GIVE THAT INFO; SETI ALREADY HAS THAT INFO...

SOMEONE NEEDS TO ADDRESS THIS RIGHT AWAY BEFORE PEOPLE FALL FOR IT. IF IT IS REAL, IT IS DONE VERY POORLY!

Best,

mizzou

I would think something of this nature might be worthy of it's own post, rather then a comment in an unrelated thread.
____________
SETI@home classic workunits: 93,865 CPU time: 863,447 hours

Join the BP6/VP6 User Group today!

1 · 2 · Next

Message boards : Number crunching : Question on BOINC versions

Copyright © 2014 University of California