Resource share - can't get the balance right


log in

Advanced search

Message boards : Number crunching : Resource share - can't get the balance right

Author Message
David Lazarus
Send message
Joined: 8 Mar 09
Posts: 37
Credit: 6,003,917
RAC: 0
United Kingdom
Message 1346192 - Posted: 13 Mar 2013, 15:21:56 UTC

OK so the objective was to crunch only SETI units unless I had no work to process which as we know is a more frequent scenario than any of us would like....

So I configured the resource share as; SETI 300 / Orbit 1 / Einstein 1 figuring this would only allow Einstein to run where there was no SETI work available.

What seems to be happening is that the work is being processed according to deadline and irrespective of the project.

Having put my team Totally Inbound in the top 10 UK contributors to SETI by RAC I would like to keep us there so your help would be appreciated.

Thanks in advance guys for your help.

David

juan BFB
Volunteer tester
Avatar
Send message
Joined: 16 Mar 07
Posts: 4939
Credit: 269,956,684
RAC: 374,432
Brazil
Message 1346194 - Posted: 13 Mar 2013, 15:30:21 UTC
Last modified: 13 Mar 2013, 15:31:31 UTC

What you need to do is put the resource share of the backup project at 0 (zero) not 1 (one), by doing that your host will crunch that project only when out of work from SETI.
____________

Profile HAL9000
Volunteer tester
Avatar
Send message
Joined: 11 Sep 99
Posts: 3860
Credit: 107,083,606
RAC: 99,028
United States
Message 1346195 - Posted: 13 Mar 2013, 15:31:32 UTC

Use 0 for backup projects. Otherwise BOINC will run the projects normally in it's time% debt system.
____________
SETI@home classic workunits: 93,865 CPU time: 863,447 hours

Join the BP6/VP6 User Group today!

David Lazarus
Send message
Joined: 8 Mar 09
Posts: 37
Credit: 6,003,917
RAC: 0
United Kingdom
Message 1346466 - Posted: 14 Mar 2013, 8:57:34 UTC

Thanks Guys,

Doh! So simple when you know how :-)

David

MikeN
Send message
Joined: 24 Jan 11
Posts: 300
Credit: 29,548,721
RAC: 41,308
United Kingdom
Message 1346474 - Posted: 14 Mar 2013, 9:57:59 UTC

Just one problem, some projects (e.g. Cosmology) will not allow you to give it a resource share of 0.
____________

juan BFB
Volunteer tester
Avatar
Send message
Joined: 16 Mar 07
Posts: 4939
Credit: 269,956,684
RAC: 374,432
Brazil
Message 1346531 - Posted: 14 Mar 2013, 13:16:21 UTC
Last modified: 14 Mar 2013, 13:21:45 UTC

Not sure if works with Cosmology but you could tray to edit the project configuration file manualy, look in the E@H file (I don´t run cosmology so not know what is their file name but must be something similar):

account_einstein.phys.uwm.edu.xml or account_setiathome.berkeley.edu.xml for example edit and change the resource_share from 100 to the zero.

<project_preferences>
<resource_share>0</resource_share>

Don´t forget to restart the entire Boinc after that (not just the Boincmgr) and if you click the read configuration file buttom the settings will be override with the old values (100 or whatever you use).

Maybe when Richard awake he could give you another better way to do that.
____________

Richard Haselgrove
Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 8375
Credit: 46,738,380
RAC: 21,536
United Kingdom
Message 1346546 - Posted: 14 Mar 2013, 14:49:06 UTC - in response to Message 1346531.

Not sure if works with Cosmology but you could tray to edit the project configuration file manualy, look in the E@H file (I don´t run cosmology so not know what is their file name but must be something similar):

account_einstein.phys.uwm.edu.xml or account_setiathome.berkeley.edu.xml for example edit and change the resource_share from 100 to the zero.

<project_preferences>
<resource_share>0</resource_share>

Don´t forget to restart the entire Boinc after that (not just the Boincmgr) and if you click the read configuration file buttom the settings will be override with the old values (100 or whatever you use).

Maybe when Richard awake he could give you another better way to do that.

Zzzzzzz :P

Remember to change it for whichever venue is active, too:

<venue name="home">
<resource_share>100</resource_share>

But I lifted that out of a sched_reply file, so the likelihood is that it would get reset to whatever the project website accepts the first time it's used.

Of course, we would hope that you don't need to call on a backup project for months at a time...

N9JFE David S
Volunteer tester
Avatar
Send message
Joined: 4 Oct 99
Posts: 10751
Credit: 13,470,513
RAC: 14,434
United States
Message 1346550 - Posted: 14 Mar 2013, 14:57:29 UTC - in response to Message 1346192.

OK so the objective was to crunch only SETI units unless I had no work to process which as we know is a more frequent scenario than any of us would like....

So I configured the resource share as; SETI 300 / Orbit 1 / Einstein 1 figuring this would only allow Einstein to run where there was no SETI work available.

What seems to be happening is that the work is being processed according to deadline and irrespective of the project.

Having put my team Totally Inbound in the top 10 UK contributors to SETI by RAC I would like to keep us there so your help would be appreciated.

Thanks in advance guys for your help.

David

You do understand that once your machine has downloaded work for a backup project, it will go ahead and process that work even if it subsequently gets new work for your main project.

____________
David
Sitting on my butt while others boldly go,
Waiting for a message from a small furry creature from Alpha Centauri.


GALAXY-VOYAGER
Avatar
Send message
Joined: 21 Oct 12
Posts: 85
Credit: 128,843
RAC: 32
Australia
Message 1346576 - Posted: 14 Mar 2013, 16:24:04 UTC

One thing that nobody has apparently mentioned.

orbit@home Is Not Active, and Has not been Active for a number of months. besides seti@home (which is what i first became involved with, because of my Interest in The Project, and which is my Favorite Project), I have found that einstein@home and rosetta@home seem to give me more work tha the others that i am involved with: However, this may just be coincidence.
____________
GALAXY-VOYAGER

Profile ignorance is no excuse
Avatar
Send message
Joined: 4 Oct 00
Posts: 9529
Credit: 44,433,274
RAC: 0
Korea, North
Message 1346580 - Posted: 14 Mar 2013, 16:42:55 UTC - in response to Message 1346576.

orbits hasnt been active in more than a couple of months. They havent had work for several years now. Interestingly, another project Asteroids@home is doing almost the exact same search as Orbits was
____________
In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

End terrorism by building a school

Profile William
Volunteer tester
Avatar
Send message
Joined: 14 Feb 13
Posts: 1572
Credit: 9,249,904
RAC: 12,744
Message 1346593 - Posted: 14 Mar 2013, 17:15:21 UTC - in response to Message 1346576.
Last modified: 14 Mar 2013, 17:17:01 UTC

One thing that nobody has apparently mentioned.

orbit@home Is Not Active, and Has not been Active for a number of months. besides seti@home (which is what i first became involved with, because of my Interest in The Project, and which is my Favorite Project), I have found that einstein@home and rosetta@home seem to give me more work tha the others that i am involved with: However, this may just be coincidence.

Orbit is currently considered a dead project and has been taken off the official list of active projects. It doesn't matter if people are still attached to it, as BOINC will only periodically poll the scheduler server and, if there are no replies, the master url - it keeps trying at such a low level, that you won't notice a thing.

I just checked with Rosetta and it still is one of the projects running such an old server version that it does not accept a resource share of 0, so doesn't work so well as a pure backup project - for a BOINC 7 setting a low share will result in minimal work while SETI has tasks and in a good supply when it doesn't. Older Boinc versions will try to keep at least one task of such a project on the rig at all times, so more (unwanted) work will get done.

Einstein almost always has work and will gladly fill up your cache if you let it. BOINC 7 workfetch algorithm means that it will try to keep the balance - when SETI workflow is stable, you'll see the host stablise in running equal shares (if you are set up for that). Older boinc versions again may not be so good at that.
____________
A person who won't read has no advantage over one who can't read. (Mark Twain)

Richard Haselgrove
Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 8375
Credit: 46,738,380
RAC: 21,536
United Kingdom
Message 1346595 - Posted: 14 Mar 2013, 17:22:23 UTC - in response to Message 1346593.

One thing that nobody has apparently mentioned.

orbit@home Is Not Active, and Has not been Active for a number of months. besides seti@home (which is what i first became involved with, because of my Interest in The Project, and which is my Favorite Project), I have found that einstein@home and rosetta@home seem to give me more work tha the others that i am involved with: However, this may just be coincidence.

Orbit is currently considered a dead project and has been taken off the official list of active projects. It doesn't matter if people are still attached to it, as BOINC will only periodically poll the scheduler server and, if there are no replies, the master url - it keeps trying at such a low level, that you won't notice a thing.

I just checked with Rosetta and it still is one of the projects running such an old server version that it does not accept a resource share of 0, so doesn't work so well as a pure backup project - for a BOINC 7 setting a low share will result in minimal work while SETI has tasks and in a good supply when it doesn't. Older Boinc versions will try to keep at least one task of such a project on the rig at all times, so more (unwanted) work will get done.

Einstein almost always has work and will gladly fill up your cache if you let it. BOINC 7 workfetch algorithm means that it will try to keep the balance - when SETI workflow is stable, you'll see the host stablise in running equal shares (if you are set up for that). Older boinc versions again may not be so good at that.

I'd just say that BOINC v7 seems to take an awful long time to settle back down into equilibrium if a new project is introduced to the mix, whether by user choice or by way of a backup fetch. Allow about a week...

BOINC v6 is better, but you're still probably counting in days.

GALAXY-VOYAGER
Avatar
Send message
Joined: 21 Oct 12
Posts: 85
Credit: 128,843
RAC: 32
Australia
Message 1346611 - Posted: 14 Mar 2013, 18:12:45 UTC - in response to Message 1346593.

yes, i still have it Attached, but i have Suspended it and about once every 2 or 3 weeks, I resume it for a day or two and see if anything has changed.
____________
GALAXY-VOYAGER

Profile Fred E.
Volunteer tester
Send message
Joined: 22 Jul 99
Posts: 768
Credit: 24,136,341
RAC: 2,611
United States
Message 1346614 - Posted: 14 Mar 2013, 18:14:54 UTC

I'd just say that BOINC v7 seems to take an awful long time to settle back down into equilibrium if a new project is introduced to the mix, whether by user choice or by way of a backup fetch. Allow about a week...

I'm using the option
<rec_half_life_days>1</rec_half_life_days>
with v7 and it will settle after a day. The defau1t is 10 days.

Client configuration

I still have to use a low cache setting and no-new-tasks to keep things in balance when running more than one project.
____________
Another Fred
Support SETI@home when you search the Web with GoodSearch or shop online with GoodShop.

rob smith
Volunteer tester
Send message
Joined: 7 Mar 03
Posts: 8140
Credit: 52,749,889
RAC: 75,123
United Kingdom
Message 1346644 - Posted: 14 Mar 2013, 19:23:04 UTC

By suspending and re-starting you are actually making it very difficult for BOINC to sort its life out. Either leave everything running, or drop the project.
____________
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

N9JFE David S
Volunteer tester
Avatar
Send message
Joined: 4 Oct 99
Posts: 10751
Credit: 13,470,513
RAC: 14,434
United States
Message 1346826 - Posted: 15 Mar 2013, 13:22:47 UTC - in response to Message 1346593.

One thing that nobody has apparently mentioned.

orbit@home Is Not Active, and Has not been Active for a number of months. besides seti@home (which is what i first became involved with, because of my Interest in The Project, and which is my Favorite Project), I have found that einstein@home and rosetta@home seem to give me more work tha the others that i am involved with: However, this may just be coincidence.

Orbit is currently considered a dead project and has been taken off the official list of active projects. It doesn't matter if people are still attached to it, as BOINC will only periodically poll the scheduler server and, if there are no replies, the master url - it keeps trying at such a low level, that you won't notice a thing.

Orbit has actually shut down its Boinc servers. Only the web site is still up and running. I'll take a look at asteroids.

Einstein almost always has work and will gladly fill up your cache if you let it.

That's for dang sure!

BOINC 7 workfetch algorithm means that it will try to keep the balance - when SETI workflow is stable, you'll see the host stablise in running equal shares (if you are set up for that). Older boinc versions again may not be so good at that.

Are 6.10.58 and 6.10.60 okay?


____________
David
Sitting on my butt while others boldly go,
Waiting for a message from a small furry creature from Alpha Centauri.


Horacio
Send message
Joined: 14 Jan 00
Posts: 536
Credit: 69,362,536
RAC: 97,829
Argentina
Message 1346894 - Posted: 15 Mar 2013, 16:30:07 UTC - in response to Message 1346826.

Im using 6.10.60 and it doesnt try really hard to keep the balance. At least not as hard as I wish...
In that version the cache is always filled with the project that can supply work. As SETI is never able to give as much work as requested then it falls behind.
In the loooooong run, as the "debts" go highly unbalanced it will try a bit harder but anyway, it's not ussually enough to put the shares were they should be...

____________

rob smith
Volunteer tester
Send message
Joined: 7 Mar 03
Posts: 8140
Credit: 52,749,889
RAC: 75,123
United Kingdom
Message 1346982 - Posted: 15 Mar 2013, 19:45:15 UTC

Two of my crunchers are running 6.10.x BOINC. They have taken a long time (months, not hours or days) to become stable, and remain so.
You have to be careful in how you set the resource share for each project. I have found that by setting the main project to 1000, and the other projects to a significantly lower level, in the range 10-50 each I get the desired split. I also run a number of resource-zero projects.
The averaging appears to take several time the longest deadline to reach stability, and S@H has about the longest deadline for "normal" tasks, at about a month, so you can expect to wait say three or four months before any sort of stability is reached.

Oh, and don't be tempted to fiddle during that time, as fiddling always upsets things in an unpredictable manner....
____________
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

Profile HAL9000
Volunteer tester
Avatar
Send message
Joined: 11 Sep 99
Posts: 3860
Credit: 107,083,606
RAC: 99,028
United States
Message 1346995 - Posted: 15 Mar 2013, 19:58:02 UTC - in response to Message 1346982.
Last modified: 15 Mar 2013, 19:58:36 UTC

Two of my crunchers are running 6.10.x BOINC. They have taken a long time (months, not hours or days) to become stable, and remain so.
You have to be careful in how you set the resource share for each project. I have found that by setting the main project to 1000, and the other projects to a significantly lower level, in the range 10-50 each I get the desired split. I also run a number of resource-zero projects.
The averaging appears to take several time the longest deadline to reach stability, and S@H has about the longest deadline for "normal" tasks, at about a month, so you can expect to wait say three or four months before any sort of stability is reached.

Oh, and don't be tempted to fiddle during that time, as fiddling always upsets things in an unpredictable manner....

A new feature starting with 7.0.40 is the use of a file app_config.xml to control the number of tasks can run for a given app/project. Which can keep BOINC from putting all of the host resources into one project.
Like the machine where I run SETI@Home beta. I have it limited to running 1 task for beta. So when it does get work from beta it doesn't take over the whole machine for weeks.
____________
SETI@home classic workunits: 93,865 CPU time: 863,447 hours

Join the BP6/VP6 User Group today!

Message boards : Number crunching : Resource share - can't get the balance right

Copyright © 2014 University of California