Hey I'm watching my two slower machines crunching for BOINC, and I'm getting frustrated as they're single core machines, and when theyre attached to more than one project it seems like theyre SO slow to finish WU's. I'd rather have the slower machines do an entire WU at a time than have them bounce back n forth every 60 min. It seems so slow bouncing back and forth, and I'd rather have one accounted for than have two at 50%. A bird in the hand is worth two in the bush, I guess...

As far as I can tell, the only way to really do this is to have the host only attached to a single project at a time.. But if that project runs out of WU's or goes off for maintenance or something then the host would be sitting there doing nothing, so for that reason, having only one project per host would be a stupid idea, woudlnt it?

Or is there some other way to have the host crunch for more than one project, but only in full WU's at a time? Or perhaps some way to have it crunch WU's for one project, but with a backup in case the one is unreachable?

PS.,
There are no stupid questions, just a lot of inquisitive idiots.

You could use an account manager and from that swiftly and easily change settings like those.

ie:
you could set up Seti (Main) to run 90% of the time and Cosmology at 10% (Backup).
That way if Seti goes into a long maintenance and you finish all your WU, itll automaticaly transfer the now "free" 90% over to Cosmolgy. When Seti comes back up, itll resume to its previous settings.

That might not be perfect, but its better than having it at 50/50

Streen,

I usually run only one project on a machine at a time. I have a couple of projects that interest me and I often run the POTM here or participate in a race.
If you tell BOINC to keep a couple day's work in the cache then you should be fine.

BB

The other things you could do are change the preferences for those PCs either by putting them in a different profile - work vs home, or manually (advanced --> preferences in BOINC manager). Increase the time per task to whatever you feel like (4, 8, 10, etc hours). As well, there may be issues if you're keeping applications suspended in memory as they won't release resources to additional projects - especially true if they're running RAM intensive projects (e.g. Cosmology!). Beware if changing this, however, as several projects won't save the work you've already done unless it's left in memory (or the checkpoints are spaced ridiculouly far apart and they might as well not be saving anything...) - check the individual project forums for more info on this (search for check points).

If I'm running a 50/50 resource share between two projects, wouldnt that still be leaving unfinished WU's for hours at a time?

For now none of this is really a problem, as I'm interested enough on how they're doing that I check up on BOINC on each system often. My fear though is that if they leave WU's unfinished for ages then some time down the road when I'm not checking up on them, they might leave WU's unfinished past the Report Deadline.

Is there any way to remotely check up on a host to see how far through WU's they are?
ie. log in somewhere and see that my host "streen-desktop' is 38% through a Cosmology WU, 28% thru SETI, 45% through Einstein@Home, and has 1 SETI WU ready to report.. and is currently crunching on Cosm and SETI at this point in time.

All I know of that I COULD make work, would be to connect thru SSH or VNC or something, but I was thinking more along the lines of some kind of BOINC stats site or Account Manager.

Yes, but if you're crunching projects that have 2 hour long wus and you're constantly switching every hour you'll be losing several CPU cycles due to the way the wus do their checkpointing (some are worse than others). You also lose CPU cycles during the switchover. If you keep the "switch between applications" time very large you'll have a fighting chance at completing several wus without any switching and it'll boost your productivity slightly. Keeping that in mind, this was just an additional thing to do above and beyond changing resource share and increasing resource share which will both give far superior results (as well as removing applications from memory when not active - but keep in mind my previously mentioned warnings regarding this).



BOINC keeps track of Deadline times and will start running any application that is nearing it's deadline time in "high-priority" mode. It's not fool-proof as it depends on estimated finishing times that may be off, but at least it'll try to finish things off in time.



BOINC manager allows for remote managment of BOINC installations on other PCs (so does BOINCView - just Google it - I've had problems with it in the past so I just stick with the "official" manager). Check out http://boinc.berkeley.edu/wiki/Controlling_BOINC_remotely. I do it all the time in LAN situations, but I'm sure it can be done over the internet - I just don't remember what port it uses for the firewall config (I think it's TCP 1043 or something similar...). Keep in mind the various security issues when opening up ports - every open port in a firewall is a new entry point for bad guys!

Good luck!

Aha, now it makes sense.. I'll keep that in mind, thanks


Okay THIS is something I never realized.. thanks for the info

As far as BOINC View is concerned, it appears as if the "official" site for it has been taken down, and its listed on the BOINC website as being a Windows app so that doesnt really help me on Linux, but thanks anyways.

You can always try BOINC-PHP-GUI which can be downloaded from http://einstein.phys.uwm.edu/download_network.php. You'd have to get Apache and PHP working on your boxes, but it does give you another option. I've never tried it myself as all my Linux boxes were connected to a KVM or accessed via VNC; unfortunately all my old hardware for my Linux builds is dead so I can't try it out for you at the moment. Maybe I'll get around to building a virtualized linux build again...