Einstein takes longer than the
6 Mar 2005 10:34:53 UTC
Topic 188309
(moderation:
I'm running Einstein/Seti2/Predictor and CPDN clients in Boinc. I've set up to receive up to 2 days worth of WU's on each and have CPDN and Einstein each set at 40% and Predictor and Seti at 10% each.
Einstein consistently downloads more WU's than the others and invariably goes past the deadline as a result on at least 1, sometimes 2 of the WU's.
What am I doing wrong here?
Peter
Toronto, Canada
Einstein takes longer than the
)
If you have set 2 days and are attached to 4 projects the actual queue length can be as much as 16 days. 2 days * 4 projects (* 2 if running an older client). Of course CPDN blows queues out of the water anyway but the math still applies to the other projects. The client will adjust some for the time the computer is off but at it starts at 24/7 and lowers with experience. The big problem is the client downloads the full queue from each project instead of reducing it based on how many projects are attached.
Reduce your queue to not more than 1.7 days to meet the 7 day deadline with 4 projects, preferably 0.75 days or lower.
[edit] It's not really your fault btw, just a flaw in how BOINC currently works.
BOINC WIKI
BOINCing since 2002/12/8
> What am I doing wrong
)
> What am I doing wrong here?
Nothing really :). The client is lousy at working out how much to request, particularly when you have multiple projects and/or don't run 24/7. By "2 days" does that mean you set your "Connect to Network" general pref to 2.0 days? If so, your smartest action would be to reduce it considerably to something like 0.2 days and let your cache size reduce to manageable values. Then you can gradually sneak it up to whatever you find doesn't leave you with a risk of expiring work. In the meantime temporarily bump up your E@H share to clear the backlog and then put it back to what you want later. Each time you change prefs you can do an immediate update by right clicking and "update.." on the E@H entry of the "Projects" tab of the Boinc GUI.
Cheers,
Gary.
Don't worry, this has
)
Don't worry, this has happened to quite a few people. It seems your cache is too large.
Try changing the "Connect to network about every" setting in your general preferences to something between 0.1 and 0.5 days.
This way you will eventually end up with less WU in your cache and will be able to complete them in time.
In the meantime, to stop you loosing more units, try temporarily increasing the share given to Einstein to flush out the units you have. If you do this, be careful not to download a whole new cache of WU before you reduce its size.
Good luck,
Richard
Thanks all for the advice.
)
Thanks all for the advice. I'll do some adjusting!! Seti2 is down for the weekend so it'll get more of a chance to catch up anyway.
If I set the "contact server" to say 0.5 days, that would affect all clients, right?
Peter
Toronto, Canada
> If I set the "contact
)
> If I set the "contact server" to say 0.5 days, that would affect all clients,
> right?
Yes it does but it's only temporary as you will increase it again when your backlog clears. To minimise your chance of getting more unneeded work, set it low until your caches reduce and then start bumping it up a bit at a time until you find an optimum value.
Cheers,
Gary.
Thanks!!
)
Thanks!!
Peter
Toronto, Canada
> Thanks!! You're welcome.
)
> Thanks!!
You're welcome. Remember to do a manual update as soon as you adjust your prefs on the website. In the Boinc GUI, right-click E@H under the projects tab.
Cheers,
Gary.
Already done...would you
)
Already done...would you believe Einstein downloaded another WU just before I did it....dang!!
Peter
Toronto, Canada