I know this is a Seti cruncher board, but there are smart people here so I'm sure one of you can answer my silly question.

I have an Predictor account and have attatched that to Boinc. I set the proportion of Seti to Predictor to 99%:1%.

I want Seti to have a priority over Predictor, and Predictor to kick in only if Seti has no work (so my computer doesn't sit idle). Other than the 1% of the time where Predictor has priority over Seti (which is effectively negligent), am I correct in assuming that if Seti fails to get work units, Predictor will kick in and try, even though it only has priority 1% of the time?

your guess is as good as mine.
But I think I would go 90-10 instead of 99-1

Either way I would just like to be able to get Wu..

Till then i un-installed it and back to the old SETI...At least it gets stuff to work on !

[/url]

Please don't use this completely unrelated thread as a means to bitch about work units.

I think that is the way it will work after we get the volume of WUWUs to keep us busy. Predictor will kick in if Seti runs out of WUWUs......


Owner's manual

Website Owner's Manual


Bah!!! I am in it for the money.

Message from the Head ;o) of the
Horse Head Nebula Branch of Seti@Home,

Check is in the mail

Here's why I ask Pascal: Here is a screenie of Boinc after I installed Predictor on it:



As you can see, it tried to contact Seti for work units and got none. However, it didn't then try to contact Predictor for units - instead it just clammed up for 40 minutes.

I'm trying to set this up so that when one project fails, it tries for units from another, and not just clam up for awhile.

Currently I only run SETI, so I don't know for sure. I think throughout the various posts I've read, I've gotten the impression that when even when SETI is set up as a primary project it will default to other projects when it can't run the primary for one reason or another. Again, I can't confirm it, but that is the impression I get. If Predictor isn't doing anything, you might try the force update option for the Predictor project. Might snap it into gear. For now though that is the only thing I can think of.

Chris

> Here's why I ask Pascal: Here is a screenie of Boinc after I installed
> Predictor on it:
>
>
>
> As you can see, it tried to contact Seti for work units and got none.
> However, it didn't then try to contact Predictor for units - instead it just
> clammed up for 40 minutes.
>
> I'm trying to set this up so that when one project fails, it tries for units
> from another, and not just clam up for awhile.
>
>

> Currently I only run SETI, so I don't know for sure. I think throughout the
> various posts I've read, I've gotten the impression that when even when SETI
> is set up as a primary project it will default to other projects when it can't
> run the primary for one reason or another. Again, I can't confirm it, but that
> is the impression I get. If Predictor isn't doing anything, you might try the
> force update option for the Predictor project. Might snap it into gear. For
> now though that is the only thing I can think of.
>
> Chris
>
> > Here's why I ask Pascal: Here is a screenie of Boinc after I installed
> > Predictor on it:
> >
> >
> >
> > As you can see, it tried to contact Seti for work units and got none.
> > However, it didn't then try to contact Predictor for units - instead it
> just
> > clammed up for 40 minutes.
> >
> > I'm trying to set this up so that when one project fails, it tries for
> units
> > from another, and not just clam up for awhile.
> >
> >
>
>
Ok push comes to shove and you can use the update function under projects tab just right click on the project you want to manual update and then watch your message tab, it may or may not connect and always should give you an error message as to why.....like project down, or no work, or my fav from predictor was, work available but used for another platform.... Hope this helps, I am new at this BOINC myself and most of what I know I picked up from the forums and doing a little reading....

The last couple of days both predictor and Seti were or are having problems. I have my 2 boxes set at 90% seti and 10% predictor, so for a while there I was doing mostly predictor cause Seti was down, then seti came up and my faster machine pickup a few Seti and the slower one kept on crunching predictor, then both picked up some more seti and they had predictor on them also and they both crunched the WUWUs that had the earliest due date.
So the key to all this I think is to get enough WUWUs from both sites to keep the boxes busy and then BOINC can get some real history, but at this point I do not know if BOINC has any bugs in this area, so it is a wait and see thing for me. I am sure the Devs have it under control and as time goes by things will get better.....
Almost forgot, the low water mark will effect the download of WUWUs also, so if you have a que full of Seti it will not get any Predictor and if it is full of predictor, no seti and when the mark is hit it should use the % and the history of what has been processed to determine which to dl or at least that is what I understand. Got to do some more reading as I hate to give out bogus info...


Owner's manual

Website Owner's Manual
Bah!!! I am in it for the money.

Message from the Head ;o) of the
Horse Head Nebula Branch of Seti@Home,

Check is in the mail

> Here's why I ask Pascal: Here is a screenie of Boinc after I installed
> Predictor on it:
>
>
>
> As you can see, it tried to contact Seti for work units and got none.
> However, it didn't then try to contact Predictor for units - instead it just
> clammed up for 40 minutes.
>
> I'm trying to set this up so that when one project fails, it tries for units
> from another, and not just clam up for awhile.
>
BOINC uses a random exponential backoff. If a project has been down for a long time, BOINC will not have all of the thousands of hosts hit it at once this way. So each time that Predictor was asked for work, and replied that it was down, your BOINC client waited longer to get back to it. On most of my machines, the wait for Beta is over a week now.

I have mine set up 10000 SAH & 1 Predictor, that works out to a 99.99% SAH & 0.01% Predictor & it works that way. The BOINC GUI trys to connect to SAH and if it gets no work then it will try Predictor. It seems to try SAH more than Predictor most of the time...
JOIN TEAM

> BOINC uses a random exponential backoff. If a project has been down for a
> long time, BOINC will not have all of the thousands of hosts hit it at once
> this way. So each time that Predictor was asked for work, and replied that it
> was down, your BOINC client waited longer to get back to it. On most of my
> machines, the wait for Beta is over a week now.

John,

I would have added that right now, BOINC does not do preemtive multitasking. I have my ratios set fot 95/5 with the big end pointed to SETI. Just because SETI does have the biggest "slice" does not mean that it will do that at all times.

OVER time, the ratios will take place, work unit availability permitting, and within the deadline times for each work unit in the buffer.

So, because work for SETI is not continuouly available, and the seconday project has work and provides WU the computer is spending more time on the other project than desired. Work in the buffer, to the extent possible, is not abandoned but completed and returned to the project.

So, Matt and I are seeing the results of "starvation" because of external factors, primarily the imbalance of the WU availability between the two projects.

Matt,

Simply put, the ratio are a Guideline for BOINC (rule of thumb as it were), not a must command.