I'm working this week actively on abc, since it's necessary to make some changes, it hardly changed since it went up, and changes are really necessary.
If you have any wishes or ideas you'd like to
see not listed below, post it here. I think there was once a thread similar like this one, but let's redo it. What I'll try to complete at least is:
-abc 64bit windows (finally, hopefully, maybe cruncher can help when I fail :) )
-mac powerpc client
-better anti-cheat protection (though this might
take some time to get it incorporated)
-better homepage layout
-contact information
-statistics on abcathome
-full list of abc triples found so far which will
be periodically updated and downloadable
/ 
Forum updates: 1) Mark all threads as read, 2) private messages
Account management: Improved merge computer functionality.
All three of these are in a more current version of the server software.
Thanks!
P.S. Can you also upgrade the beta server too, for the merge computer functionality? I need to clean up a number of machines there.
P.P.S: Does this mean there will be testing activity on the beta server? Should we re-attach?
____________
Dublin, CA
Team SETI.USA
/ James Dorsey (admin over at APS@Home) asked me to relay the following offer:
I noticed a thread on ABC@home saying that the administrator was going to spend part of the rest of the week looking at various issues including:
-better anti-cheat protection (though this might take some time to get it incorporated)
-statistics on abcathome
So that I don't have to register with ABC to make the offer, please could you let the admin there know that we have a very reliable credit system (which I'm willing to help them out with if my implementation meets the needs of ABC), and that I have some code from Leiden Classical, which I adapted to add project stats to our server status page. I'm also happy to provide that PHP code as is, if it's any use to ABC.
He can be reached here:
http://www.apsathome.org/
And then clicking on the "Project personnel" link. His email address is there. I don't want to post it directly here to avoid the SPAM.
____________
Dublin, CA
Team SETI.USA
/ 
What I'll try to complete at least is:
-abc 64bit windows (finally, hopefully, maybe cruncher can help when I fail :) )
-give crunch3r the source code
-mac powerpc client
-better anti-cheat protection (though this might
take some time to get it incorporated)
-better homepage layout
-contact information
-statistics on abcathome
-full list of abc triples found so far which will
be periodically updated and downloadable
made a little addition to the wish list
____________
regards
Crunch3r
/ Ok, thanks. I will look at the credit system later what they did.
I read the post on aps@home and saw what you did Cruncher, nice, porting
to all these platforms.
/ Ok, thanks. I will look at the credit system later what they did.
I read the post on aps@home and saw what you did Cruncher, nice, porting
to all these platforms.
Yeah. It's fun. After building the x64 APS@Home for 64 bit windows, wich is allready 85% faster than the stock 32 bit app, i manages to increase performance one again by another 80%... So the 64 bit app is now roughly estimated 160-180% faster than the stock application.
Same could apply for the abc app.... ;)
____________
regards
Crunch3r
/ Ok, thanks. I will look at the credit system later what they did.
I read the post on aps@home and saw what you did Cruncher, nice, porting
to all these platforms.
Yeah. It's fun. After building the x64 APS@Home for 64 bit windows, wich is allready 85% faster than the stock 32 bit app, i manages to increase performance one again by another 80%... So the 64 bit app is now roughly estimated 160-180% faster than the stock application.
Same could apply for the abc app.... ;)
Yes, yes, give Crunch3r the source code!
/ 
-better homepage layout
Thank you for the new layout. :)
Suggestion:
instead of
Powered by
;)
____________
Proud member of SETI.Germany
/ -better homepage layout
Thank you for the new layout. :)
Suggestion:
instead of
Powered by
;)
thanks
/ How hard would it be to upgrade a version of the boinc server to support separate settings for Connect to network about every and Maintain enough work for an additional/Enforced by version 5.10+?
/ How hard would it be to upgrade a version of the boinc server to support separate settings for Connect to network about every and Maintain enough work for an additional/Enforced by version 5.10+?
These settings are only client-related, so no matter which version the server uses.
You can set them in your BOINC Manager's preferences dialog, or, since these settings are global, set them in BAM! or any project you participate in.
____________
Proud member of SETI.Germany
/ 
How hard would it be to upgrade a version of the boinc server to support separate settings for Connect to network about every and Maintain enough work for an additional/Enforced by version 5.10+?
These settings are only client-related, so no matter which version the server uses.
You can set them in your BOINC Manager's preferences dialog, or, since these settings are global, set them in BAM! or any project you participate in.
If you have a look around at other projects, you will see that it's not only something at the client end. Yes, it can be done in the latest clients, through BOINC Manager, but try changing muleiple hosts (including those that are running command-line only and need to be controlled from another host).
The server side option makes it easy to update all hosts at one venue.
____________
Join the #1 Aussie Alliance on ABC@Home
/ .
Let me express a wish for trying to better predict the length of ABC work units.
I run off-line. Meaning that when I connect, I need to download enough work to keep my system busy until the next time I connect.
Every workunit provides an estimated # of floating point operations to complete that workunit. When that workunit finishes, the BOINC client divides this estimate by the actual # of floating point operations it measured, and keeps the result as a 'correction factor' (to be applied by the client's work scheduler in calculating how long it will take to crunch "to be done next" work from this project).
As far as I can tell, the workunits supplied by ABC all carry the __same__ estimated # of floating point operations. But the actual amount of crunching needed is __not__ the same for all workunits. Once in a while a "joker" workunit comes along, which takes from five to more than ten times *more* crunching than the "average" ABC workunit. And when the "correction factor" as calculated from that "joker" workunit gets applied to any "to be done next" ABC workunits, *their* calculated crunching requirement gets grossly inflated (unless they too happen to be "joker" workunits, which is normally not the case). [Although the value of the composite 'correction factor' for the project is re-adjusted as each workunit finishes, that adjusted value "drops" at a much slower rate than it "rises".]
--------
Suppose it takes my system one hour to crunch a "normal" ABC workunit. When a "joker" ABC workunit comes along and takes six hours to crunch, the client's work scheduler will expect "to be done next" ABC workunits would *also* need closer to six hours to finish (since the estimate each provides is the same as the estimate that the "joker" provided).
If I now connect and request work, far *fewer* ABC workunits will be fetched than if the "joker" had not inflated the 'correction factor'. If those ABC workunits take only one hour each, my system can run out of work unless the subsequent connection (to again fetch work) is made sooner than I planned.
What I __wish__ is that the ABC project could *identify* "joker" workunits, and assign them larger estimates than those assigned to "normal" ABC workunits.
.
/ 
Wouldnt we all but in the past it was technically not feasible. That said the uber short Wu's are now filtered so can the uber long ones be identified the same way?
____________
/
Clownius, has anyone directly connected to the project told you the short ones are being identified and then filtered? Maybe I missed a post from an admin?
I gather they have a method of increasing the length of the work units but the method doesn't allow fine control. In other words they can make the work units longer but they can never know precisely how much longer they will be. I don't think they can actually identify short work units before they are crunched and if they cannot identify them then I doubt they can filter them. I could be wrong.
/
Clownius, has anyone directly connected to the project told you the short ones are being identified and then filtered? Maybe I missed a post from an admin?
I'm not Clownius, but I do recall seeing either a news item or post from Hendrik in which he talked about work units being tested before release to filter out the very short ones. If I can find the post or news item, I'll let you know where it is.
EDIT: Hmm, can't find it. Maybe we (Clownius and I) are mistaken and the testing was on another project?
____________
Join the #1 Aussie Alliance on ABC@Home
/
Clownius, has anyone directly connected to the project told you the short ones are being identified and then filtered? Maybe I missed a post from an admin?
I'm not Clownius, but I do recall seeing either a news item or post from Hendrik in which he talked about work units being tested before release to filter out the very short ones. If I can find the post or news item, I'll let you know where it is.
EDIT: Hmm, can't find it. Maybe we (Clownius and I) are mistaken and the testing was on another project?
I think you are right, or at least close to right. I now have a similar recollection but it's vague and I can't find the post I'm thinking of either. As I recall the discussion started here. Later, Verhoek posted that he had increased the WU length by 200% and that he was thinking of a way to identify/catch/filter the very short WUs? Does that ring a bell? I got the impression they are trying to devise a way to better predict WU length but don't actually have anything in place at this time.
I also recall thinking... Who is Verhoek? Isn't Hendrik's last name Verhoek? And now all of Verhoek's posts have mysteriously disappeared? Huh?
/ Hmm im lost i do recall something being said but cant find anything now either. But all the sub 10 second ones are gone so something happened.
____________
/
Clownius, has anyone directly connected to the project told you the short ones are being identified and then filtered? Maybe I missed a post from an admin?
I'm not Clownius, but I do recall seeing either a news item or post from Hendrik in which he talked about work units being tested before release to filter out the very short ones. If I can find the post or news item, I'll let you know where it is.
EDIT: Hmm, can't find it. Maybe we (Clownius and I) are mistaken and the testing was on another project?
Hmm...
I wonder if you have confused the ABC short-WU problem with the RCN short-WU problem. At RCN there is a check for seconds-long-WU by performing a short computation.
____________
/ Nope that solution is very old i used to crunch a lot there back then. Im certain i saw it here somewhere but... cant find it now and an in depth search is beyond me with the shifts im doing at work at the moment.
____________
/ 
... Im certain i saw it here somewhere but... cant find it now and an in depth search is beyond me with the shifts im doing at work at the moment.
Maybe you were looking for this one? ;-)
____________
Lovely greetings from Cori
/ Maybe you were looking for this one? ;-)
Nope.
At any rate, it would appear that the duration of work units is still quite variable but something has been done about the very short (under a minute) work units. Most of the other work units have increased in duration as well (a few hours on a fast computer is quite common now).
/ It would be nice to allow WU validation/awarding of credits for late results... especially the monster WUs.
re: http://abcathome.com/result.php?resultid=10642061
One of my unattended boxes suffered the results of the area's lone high line (power feed) failure - a number of the poles replaced after the hurricane toppled in a recent thunderstorm. The box's UPS does not allow for 12+ hr runtime.
If the credits cannot be awarded, that's Life...
A
____________
/ 
Clownius, has anyone directly connected to the project told you the short ones are being identified and then filtered? Maybe I missed a post from an admin?
I'm not Clownius, but I do recall seeing either a news item or post from Hendrik in which he talked about work units being tested before release to filter out the very short ones. If I can find the post or news item, I'll let you know where it is.
EDIT: Hmm, can't find it. Maybe we (Clownius and I) are mistaken and the testing was on another project?
Hmm...
I wonder if you have confused the ABC short-WU problem with the RCN short-WU problem. At RCN there is a check for seconds-long-WU by performing a short computation.
RCN is the only project I can recall that precomputes wu's - if they complete within xx seconds then all well and good the internal servers/crunchers claim the glory - else they are let loose on a hungry crunching population.
I think Hendrik posted a while back that he had managed to increase the minimum crunchtime of a wu - but it wasn't due to precrunching - just tweaking the parameters.
____________
/ It would be nice to allow WU validation/awarding of credits for late results... especially the monster WUs.
wouldn't that in affect remove dead line for work units ?
____________
/ It would be nice to allow WU validation/awarding of credits for late results... especially the monster WUs.
wouldn't that in affect remove dead line for work units ?
Yes... and it might encourage folks to start/continue running their boxes unattended.
/ Also would lead to people running even bigger caches. Somwe poeple keep 1k or more Wu's on hand as is lets not make them hold more or our pending credits will never get granted.
If its returned late but before the computer that crunches the reissued WU it still gets credit but if the computer that gets the reissue crunches it first the WU isnt credited. The idea of DC is to complete the work faster not slower so deadlines are needed to make sure incomplete work is reissued and completed.
____________
/ ...The idea of DC is to complete the work faster not slower so deadlines are needed to make sure incomplete work is reissued and completed.Agreed... but if folks w/unattended boxes were to stop running them while unattended to avoid crunching for nothing/kicks for free in the case of mitigating factors as I've previously described, the net result would be?
Cache sizes can be controlled by the projects.
If memory serves me right, allowances can, and have been made on WU deadlines...
____________
/