BOINC cache management

Questions and Answers : Wish list : BOINC cache management
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile greencreeper

Send message
Joined: 22 Jun 03
Posts: 49
Credit: 447,066
RAC: 0
United Kingdom
Message 5499 - Posted: 9 Jul 2004, 1:35:08 UTC

I'm a member of the SETI and Predictor projects. If I set, for example, the following cache parameters in my account pages:

Predictor - LWM: 0.5, HWM: 1
SETI - LWM: 3, HWM: 5

These settings are effectively ignored by BOINC. When it communicates with the SETI scheduler and the project is down, BOINC requests more work from Predictor instead. So instead of having upto a days worth of Predictor WUs and 5 days of SETI WUs, I actually have more than 5 days of Predictor WUs. Some won't be returned before their deadline. I know what BOINC is doing - it's ensuring the cache is full according to the greatest LWM - but this isn't right. A project that is frequently down, like SETI is at the moment, will be starved of processing time because BOINC is grabbing WUs from other projects that aren't down. It's pot luck whether the LWM will be hit when a project is up - everytime my BOINC has requested WUs from SETI the project has been down. BOINC hasn't processed a SETI WU in what must be nearly a week!


j
ID: 5499 · Report as offensive
Profile Ranz
Volunteer tester

Send message
Joined: 19 May 02
Posts: 34
Credit: 338,178
RAC: 0
United States
Message 5540 - Posted: 9 Jul 2004, 3:44:40 UTC

You can't set different LWM for different projects. When you go to the General Preferences and set your desired # of days work between xx-yy days, that preference is used for all projects, not just that one you set it on. In your example, you have the following settings:

Predictor - LWM: 0.5, HWM: 1
SETI - LWM: 3, HWM: 5


When you manually 'Update' or the next time your BOINC program contacts the server to get your preferences, it will use the 'General Preferences' of the most recent change. If, for example, you had done changes to your SETI project last, then those settings will be used. At this point, your BOINC cache should request more work to have minimum 3 days worth up to 5 days max worth of total WU's combined from both projects. Here's the note on the 'General Preferences' page:

General preferences

These apply to all BOINC projects in which you participate.
On computers attached to multiple projects, the most recently modified preferences will be used.



Ranz


ID: 5540 · Report as offensive
Profile greencreeper

Send message
Joined: 22 Jun 03
Posts: 49
Credit: 447,066
RAC: 0
United Kingdom
Message 5826 - Posted: 9 Jul 2004, 22:18:09 UTC - in response to Message 5540.  

> You can't set different LWM for different projects. When you go to the
> General Preferences and set your desired # of days work between xx-yy days,
> that preference is used for all projects, not just that one you set it on. In
> your example, you have the following settings:
>
> Predictor - LWM: 0.5, HWM: 1
> SETI - LWM: 3, HWM: 5
>
>
> When you manually 'Update' or the next time your BOINC program contacts the
> server to get your preferences, it will use the 'General Preferences' of the
> most recent change. If, for example, you had done changes to your SETI
> project last, then those settings will be used. At this point, your BOINC
> cache should request more work to have minimum 3 days worth up to 5 days max
> worth of total WU's combined from both projects. Here's the note on the
> 'General Preferences' page:
>
> General preferences
>
> These apply to all BOINC projects in which you participate.
> On computers attached to multiple projects, the most recently modified
> preferences will be used.
>

Yeah - just discovered that this morning. So that's another problem then :) Porjects are dependant on other projects - I didn't think BOINC was designed like this?? It doesn't change my original point though. If a project is frequently down, say because it's an alpha, then it's just not going to get processing time. I've had to reset Predictor to give SETI a chance.

At the moment if BOINC can't get work it immediately downloads work from the next project. It would be better if it waited a while - say another 4 attempts - before moving to the next project.


j
ID: 5826 · Report as offensive
Heffed
Volunteer tester

Send message
Joined: 19 Mar 02
Posts: 1856
Credit: 40,736
RAC: 0
United States
Message 5854 - Posted: 9 Jul 2004, 23:29:53 UTC - in response to Message 5826.  

> Yeah - just discovered that this morning. So that's another problem then :)
> Porjects are dependant on other projects - I didn't think BOINC was designed
> like this??

This is where "resource share" comes in.

> It doesn't change my original point though. If a project is
> frequently down, say because it's an alpha, then it's just not going to get
> processing time. I've had to reset Predictor to give SETI a chance.

You are correct. If a project is frequently down, it's not going to get any processing time. That would seem blatently obvious... ;-)

> At the moment if BOINC can't get work it immediately downloads work from the
> next project. It would be better if it waited a while - say another 4 attempts
> - before moving to the next project.

But why? This would only delay processing in extreme instances. (say your queue is empty, so BOINC is sitting idle)

<a> [/url]
ID: 5854 · Report as offensive
Profile greencreeper

Send message
Joined: 22 Jun 03
Posts: 49
Credit: 447,066
RAC: 0
United Kingdom
Message 5905 - Posted: 10 Jul 2004, 1:43:15 UTC - in response to Message 5854.  

> > It doesn't change my original point though. If a project is
> > frequently down, say because it's an alpha, then it's just not going to
> get
> > processing time. I've had to reset Predictor to give SETI a chance.
>
> You are correct. If a project is frequently down, it's not going to get any
> processing time. That would seem blatently obvious... ;-)
>

How is it obvious? If I'd done nothing with BOINC, then it would be happily downloading and working on Predictor WUs. It's unlikely BOINC would have processed a SETI WU until the project is more stable - and how long will it be before that's the case? A down project can still get processing time - the project doesn't have to be up for BOINC to process the project's WUs.

> > At the moment if BOINC can't get work it immediately downloads work from
> the
> > next project. It would be better if it waited a while - say another 4
> attempts
> > - before moving to the next project.
>
> But why? This would only delay processing in extreme instances. (say your
> queue is empty, so BOINC is sitting idle)
>

All I am saying is that the way it works at the moment means that Alpha, and perhaps even Beta projects, aren't likely to stand much chance against projects that are always running and always have work. If alpha projects don't get processing time they're unlikely to progress beyond alpha.

Let's say that the LWM is 3 days - that means that when the water mark is hit there are 2.99 days of work remaining. A project isn't likely to be down or workless for that long - not even SETI ;) And the example I gave of "four tries and then move on to the next project" would take place over the course of a day anyway. It's unlikely that BOINC would run out of work, and it could always have a 1 WU critical water mark - an empty queue means download work from any responding project.

j
ID: 5905 · Report as offensive
Profile Keck_Komputers
Volunteer tester
Avatar

Send message
Joined: 4 Jul 99
Posts: 1575
Credit: 4,152,111
RAC: 1
United States
Message 5932 - Posted: 10 Jul 2004, 3:37:05 UTC

The smaller the range between high and low water marks the less amount of work you will get per connection and therefore the more likely you are to catch a specific project up.

John Keck
BOINCing since 2002/12/08
ID: 5932 · Report as offensive
Heffed
Volunteer tester

Send message
Joined: 19 Mar 02
Posts: 1856
Credit: 40,736
RAC: 0
United States
Message 5935 - Posted: 10 Jul 2004, 4:01:16 UTC - in response to Message 5905.  

> How is it obvious? If I'd done nothing with BOINC, then it would be happily
> downloading and working on Predictor WUs. It's unlikely BOINC would have
> processed a SETI WU until the project is more stable - and how long will it be
> before that's the case? A down project can still get processing time - the
> project doesn't have to be up for BOINC to process the project's WUs.

Sorry, bit of a joke there... If a project it down, it can't give you work. Therefore it cannot get any processing time. ;-)

Please excuse the levity.

<a> [/url]
ID: 5935 · Report as offensive
John McLeod VII
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 15 Jul 99
Posts: 24806
Credit: 790,712
RAC: 0
United States
Message 6368 - Posted: 11 Jul 2004, 4:40:09 UTC

BOINC has only one queue. It has one setting for high and low marks. The most recent of these will migrate from project to host to project until all are the same for that email address.

To control how much work is done by each project use the resource share.

ID: 6368 · Report as offensive

Questions and Answers : Wish list : BOINC cache management


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.