4.35 is in early alpha testing and has a new download and CPU scheduler on the client that will do a better job of limiting work on the computer and scheduling the computer so that deadlines are met.
> > Dump the pathetically slow Celeron.
> >
> >
>
> I don't think there's any call for talk like that. Obviously Chiana is
> interested in putting the ol' machine to constructive use and that's
> admirable. Chiana, do you crunch for SETI or any other projects for which the
> Celeron might be better suited? Perhaps dedicate that machine solely to the
> project(s) with longer deadlines and shorter processing time. I keep my
> Celeron here on Einstein because it crunches a unit pretty quick, but my old
> K6-2 will definitely be doing SETI work only.
>
> Happy Crunching ;)
>
> Dig
Yes, I did just that, I moved the E@H calculations from that mendocino into a more "normal" celeron @ 600 MHz, that one does the WU's on time...
> The client side CPU and download schedulers are being re-worked slightly to
> take deadlines more seriously. There are two modes for the new CPU scheduler,
> the normal mode that works like the one we have now, and the panic mode
> scheduler that crunches on the earliest deadline WU in an attempt to make it
> in on time (or make it less late). The download scheduler will prevent
> downloading from any project if it is determined that there is already too
> much work for the CPU scheduler to handle comfortably.
>
> This code is still in development although it looks very promissing.
Yeah, that sounds good, I think I've put something like that on the "wish list" at seti...
You guys are missing his point - irregardless of celeron or not.
I'm running a P4 2.4GZ with 512MB - not an old dog by any measure. However, I frequently don't finish them "on time" because the check in time is too brief. SETI@Home gives what, 2 weeks to return results? Considering folks may not run their PC for a couple of days when they are out of town, the 1 week window E@Home uses it too short.
They'd get more results from me if they bumped it out to two weeks.
'nuf said.
I recently joined Einstein in addition to SETI, which I've been running (Classic and now Boinc) for over a year on my G4 450mHz Mac. However, due to the short processing time allowed, I've had to detach the project, having managed to finish just one WU in time but only by suspending work on SETI. I wouldn't mind giving SETI a 'rest' now and again but the screensaver doesn't work yet on Macs for Einstein either, which was one reason for trying it out.
BOINC 4.4x should be able to meet deadlines if the host can meet the deadline on its own. If you are using 4.45 or later, and are missing deadlines. Could you help me figure out why the deadlines are being missed?
I am running 4.45, connect every 0.04
I just discovered that i missed the deadline and have 2H left to crunch. Usual time 30H / WU.
There are 2 reasons why i missed the deadline.
I attached to SETI Beta, and have spent 2H crunching that project.
The computer have been on 8H less this week.
Because i have an old computer i wish to only have one WU at a time on my computer. That is no problem then i am only attached to one project. Now since you introduced the new scheduling in the client i have been adding more projects to my computer to see if the scheduler can handle it. Doing this i have found a few things you might want to hear about.
To have the margins on my side i only want to have one WU at a time on my computer. For that reason i set my connect time to 0.04 so i get a new WU just before i finish the current one. so downloading a new WU from another project if more then 0.04 days remains on the current one would be against my wishes. Having CPDN in mind i only see two options to ensure that this happens.
Have an advanced option that say "Don't use round robin scheduling" or make the scheduler not download more work if more time remains on the WU then the "connect to" time, if "connect to" less then 0.05 or something more appropriate.
BTW, just uploaded the result and got credit. Have 15H left on the SETI Beta WU ant the client immediately try to download new work from Orbit and then LHC. If it had downloaded work from Einstein i would have been in real trouble.
Now if my memory serves me correctly, i found this to happened. If the client trays to download more work and are not able to get some, it will try to download more work immediately then i power on the computer next time, even if it 10 sec later will say Suspending work fetch because computer is over committed.
BTW, BOINC manager cant connect to local host if i have Gmail notifier running.
Then you're really interested in a subject, there is no way to avoid it. You have to read the Manual.
I write for a lot of user of "L'Alliance Francophone" team
I think taht 2weeks should increase the number of users. For some people , it's impossible to crunch in 1 week ( a lot of project , holydays , no intenret connection)
For exemple seti@home 2.5hours (deadline 14 days)
einstein 8hours (deadline 7days)
Units are not really "linked" so I don't understand that small deadline .
I am running 4.45, connect every 0.04
I just discovered that i missed the deadline and have 2H left to crunch. Usual time 30H / WU.
There are 2 reasons why i missed the deadline.
I attached to SETI Beta, and have spent 2H crunching that project.
The computer have been on 8H less this week.
Because i have an old computer i wish to only have one WU at a time on my computer. That is no problem then i am only attached to one project. Now since you introduced the new scheduling in the client i have been adding more projects to my computer to see if the scheduler can handle it. Doing this i have found a few things you might want to hear about.
To have the margins on my side i only want to have one WU at a time on my computer. For that reason i set my connect time to 0.04 so i get a new WU just before i finish the current one. so downloading a new WU from another project if more then 0.04 days remains on the current one would be against my wishes. Having CPDN in mind i only see two options to ensure that this happens.
Have an advanced option that say "Don't use round robin scheduling" or make the scheduler not download more work if more time remains on the WU then the "connect to" time, if "connect to" less then 0.05 or something more appropriate.
BTW, just uploaded the result and got credit. Have 15H left on the SETI Beta WU ant the client immediately try to download new work from Orbit and then LHC. If it had downloaded work from Einstein i would have been in real trouble.
Now if my memory serves me correctly, i found this to happened. If the client trays to download more work and are not able to get some, it will try to download more work immediately then i power on the computer next time, even if it 10 sec later will say Suspending work fetch because computer is over committed.
BTW, BOINC manager cant connect to local host if i have Gmail notifier running.
There is a server change comming later this summer that should help with this problem. However, it will take every project upgrading in order to make it fully effective. The client is currently sending back enough information about every work unit on your system so that the server can figure out if there is enough slack on your host to download work. Of course, the newly downloaded work may immediately drive the scheduler into EDF, but it should not download work that cannot be finished. The problem is that the client makes one call to the server to get a work description the return from that determines which if any files to download - the client does not have control at the point that the decision is made, and the server did not used to have enough information. Now all that is left is getting the server to act on the information.
I HAVE SUSPENDED MY EINSTEIN ACCOUNT UNTIL THINGS ARE PUT RIGHT!
having not been able to finish 2 of 3 packets for the reasons quoted here:-
either rethink the pathetically short schedule - speaking from a home users point of view - or make the work packets more reasonable in size...
How on earth can you expect to have approx 5 hrs of crunch time per week per pc!!! - Well easy if you're very cheeky or only want it run on dual cpu powerhouses etc etc (put your own gripe here)!
How on earth can you expect to have approx 5 hrs of crunch time per week per pc!!!
Not every project is going to be perfect for all users. It's really up to the project to decide what deadlines are acceptable for their project. I'm sure they would love you to run it, but if you can't make it happen, I'm sure they understand.
Quote:
(running 3Ghz AMD 1GB Ram etc etc)
Really? I wasn't aware there was a 3GHz AMD out yet. Out of curiosity, what CPU is it?
4.35 is in early alpha
)
4.35 is in early alpha testing and has a new download and CPU scheduler on the client that will do a better job of limiting work on the computer and scheduling the computer so that deadlines are met.
BOINC WIKI
> > Dump the pathetically
)
> > Dump the pathetically slow Celeron.
> >
> >
>
> I don't think there's any call for talk like that. Obviously Chiana is
> interested in putting the ol' machine to constructive use and that's
> admirable. Chiana, do you crunch for SETI or any other projects for which the
> Celeron might be better suited? Perhaps dedicate that machine solely to the
> project(s) with longer deadlines and shorter processing time. I keep my
> Celeron here on Einstein because it crunches a unit pretty quick, but my old
> K6-2 will definitely be doing SETI work only.
>
> Happy Crunching ;)
>
> Dig
Yes, I did just that, I moved the E@H calculations from that mendocino into a more "normal" celeron @ 600 MHz, that one does the WU's on time...
/Chi
> The client side CPU and
)
> The client side CPU and download schedulers are being re-worked slightly to
> take deadlines more seriously. There are two modes for the new CPU scheduler,
> the normal mode that works like the one we have now, and the panic mode
> scheduler that crunches on the earliest deadline WU in an attempt to make it
> in on time (or make it less late). The download scheduler will prevent
> downloading from any project if it is determined that there is already too
> much work for the CPU scheduler to handle comfortably.
>
> This code is still in development although it looks very promissing.
Yeah, that sounds good, I think I've put something like that on the "wish list" at seti...
/Chi
RE: You guys are missing
)
I recently joined Einstein in addition to SETI, which I've been running (Classic and now Boinc) for over a year on my G4 450mHz Mac. However, due to the short processing time allowed, I've had to detach the project, having managed to finish just one WU in time but only by suspending work on SETI. I wouldn't mind giving SETI a 'rest' now and again but the screensaver doesn't work yet on Macs for Einstein either, which was one reason for trying it out.
BOINC 4.4x should be able to
)
BOINC 4.4x should be able to meet deadlines if the host can meet the deadline on its own. If you are using 4.45 or later, and are missing deadlines. Could you help me figure out why the deadlines are being missed?
BOINC WIKI
I am running 4.45, connect
)
I am running 4.45, connect every 0.04
I just discovered that i missed the deadline and have 2H left to crunch. Usual time 30H / WU.
There are 2 reasons why i missed the deadline.
I attached to SETI Beta, and have spent 2H crunching that project.
The computer have been on 8H less this week.
Because i have an old computer i wish to only have one WU at a time on my computer. That is no problem then i am only attached to one project. Now since you introduced the new scheduling in the client i have been adding more projects to my computer to see if the scheduler can handle it. Doing this i have found a few things you might want to hear about.
To have the margins on my side i only want to have one WU at a time on my computer. For that reason i set my connect time to 0.04 so i get a new WU just before i finish the current one. so downloading a new WU from another project if more then 0.04 days remains on the current one would be against my wishes. Having CPDN in mind i only see two options to ensure that this happens.
Have an advanced option that say "Don't use round robin scheduling" or make the scheduler not download more work if more time remains on the WU then the "connect to" time, if "connect to" less then 0.05 or something more appropriate.
BTW, just uploaded the result and got credit. Have 15H left on the SETI Beta WU ant the client immediately try to download new work from Orbit and then LHC. If it had downloaded work from Einstein i would have been in real trouble.
Now if my memory serves me correctly, i found this to happened. If the client trays to download more work and are not able to get some, it will try to download more work immediately then i power on the computer next time, even if it 10 sec later will say Suspending work fetch because computer is over committed.
BTW, BOINC manager cant connect to local host if i have Gmail notifier running.
Then you're really interested in a subject, there is no way to avoid it. You have to read the Manual.
I write for a lot of user of
)
I write for a lot of user of "L'Alliance Francophone" team
I think taht 2weeks should increase the number of users. For some people , it's impossible to crunch in 1 week ( a lot of project , holydays , no intenret connection)
For exemple seti@home 2.5hours (deadline 14 days)
einstein 8hours (deadline 7days)
Units are not really "linked" so I don't understand that small deadline .
Thank you to einstein@home team
RE: I am running 4.45,
)
There is a server change comming later this summer that should help with this problem. However, it will take every project upgrading in order to make it fully effective. The client is currently sending back enough information about every work unit on your system so that the server can figure out if there is enough slack on your host to download work. Of course, the newly downloaded work may immediately drive the scheduler into EDF, but it should not download work that cannot be finished. The problem is that the client makes one call to the server to get a work description the return from that determines which if any files to download - the client does not have control at the point that the decision is made, and the server did not used to have enough information. Now all that is left is getting the server to act on the information.
BOINC WIKI
Exactly - I HAVE SUSPENDED
)
Exactly -
I HAVE SUSPENDED MY EINSTEIN ACCOUNT UNTIL THINGS ARE PUT RIGHT!
having not been able to finish 2 of 3 packets for the reasons quoted here:-
either rethink the pathetically short schedule - speaking from a home users point of view - or make the work packets more reasonable in size...
How on earth can you expect to have approx 5 hrs of crunch time per week per pc!!! - Well easy if you're very cheeky or only want it run on dual cpu powerhouses etc etc (put your own gripe here)!
(running 3Ghz AMD 1GB Ram etc etc)
RE: How on earth can you
)
Not every project is going to be perfect for all users. It's really up to the project to decide what deadlines are acceptable for their project. I'm sure they would love you to run it, but if you can't make it happen, I'm sure they understand.
Really? I wasn't aware there was a 3GHz AMD out yet. Out of curiosity, what CPU is it?