(Now back to "normal mode".) I think the granted credit/qourum process is pretty bogus (although it doesn't bother me enough to complain about). And, I am not too crazy about the claimed credit/benchmarhing thing either; but, again, it doesn't bother me. However, if you are "into" the credit counting "thing", why don't you start a thread on "Wish List" about the process. I might even throw in "my two cents worth".
Stick
Stick,
I'll take the silver-tongue plea under consideration. I haven't made a move yet, but I was pretty much waiting, "hanging in there", until my testing was concluded. Now that 5.2.2 has gone Prime-Time, the devs probably won't need any more of my feedback, especially that of the vehement, or as Gary so graciously called it "vigorous", sort. lol
As to the "normal mode" topic - I posted my thoughts on the credit thing in my Automatic Benchmarking" thread in Cruncher's corner a couple of months ago, when I was just getting my feet wet in the message boards. My feelings on that haven't changed one iota in the interim, but reality has set in. The bench/quorum method is patently unfair, and will always be so. While Einstein WUs invariably require a set amount of work, on some other projects, the amount of processing needed can vary widely from one WU to the next. That was sometimes the case with SetiClassic, sets of "broken" WUs distributed occasionally. The reality is that now, everything is "locked-in" by dependence upon Boinc for the purposes of cross-project compatibility in credit count, which is what draws the crunchers and keeps 'em crunching. I, for one, greatly appreciated that aspect of SetiClassic, that you did a day's (WU's) work and you received a day's (ahem) pay.
Michael
microcraft
"The arc of history is long, but it bends toward justice" - MLK
My Seti result was reported just as you explained it would be. (It's nice to have software that does what it's supposed to!) ;-) Anyway, I will remember your explanation (and this post) as, inevitably, the question will come up again as users begin to change over.
Stick
Quote:
Walt (and Michael),
Thank you for the response(s)! I guess I just came across a new "feature" and didn't recognize it as such. It's definitely a difference from the way 4.45 works.
Quote:
Quote:
I know this is not the place to report this (and if it doesn't correct itself soon, I'll post it on the BOINC site). But, having just installed 5.2.2, I have come across my first "strange experience" with it. It finished it's first Seti unit about an hour ago (which was probably started under 4.45). The unit is listed under the "Work" tab as "Ready to report". However, it is not listed under the "Transfers" tab. And, when I look under the "Messages" tab, it shows that, after three "Temporarily failed upload" sequences, it uploaded successfully on the fourth try. The Seti site stills shows the WU as "In Progess", however.
Stick,
Sending the results is a two step process. First the result data is uploaded to the upload server, and while thats happening you'll see it in the transfer tab. Afterwords the result will be "waiting to report", which it'll do the next time BOINC contacts the scheduler. Reporting is where it sends the administrative details about the result - exit status, CPU time, error messages, things like that.
Some earlier versions of BOINC had a bug where it reported results right after uploading. That was fixed in 5.2. Now it'll contact the scheduler to report if it reaches the deadline or 24 hours after the workunit finishes. If BOINC contacts the scheduler for some other reason, it'll report the results then. Like if you manually "update" the project or BOINC requests more work.
Idea is to reduce the times BOINC contacts the scheduler to report work or request it, and this way it combines several requests into one contacts. With a short "connect every" interval it'll report the previous result when it requests a new workunit. With a longer interval it can report several workunits either after 24 hours have passed or when it requests more work.
And even though a result gets uploaded, the server won't show it as complete until its reported. And it'll use that as the "completion" time, even though the result had already been uploaded.
My Seti result was reported just as you explained it would be. (It's nice to have software that does what it's supposed to!) ;-) Anyway, I will remember your explanation (and this post) as, inevitably, the question will come up again as users begin to change over.
Stick
Quote:
Walt (and Michael),
Thank you for the response(s)! I guess I just came across a new "feature" and didn't recognize it as such. It's definitely a difference from the way 4.45 works.
Quote:
Quote:
I know this is not the place to report this (and if it doesn't correct itself soon, I'll post it on the BOINC site). But, having just installed 5.2.2, I have come across my first "strange experience" with it. It finished it's first Seti unit about an hour ago (which was probably started under 4.45). The unit is listed under the "Work" tab as "Ready to report". However, it is not listed under the "Transfers" tab. And, when I look under the "Messages" tab, it shows that, after three "Temporarily failed upload" sequences, it uploaded successfully on the fourth try. The Seti site stills shows the WU as "In Progess", however.
Stick,
Sending the results is a two step process. First the result data is uploaded to the upload server, and while thats happening you'll see it in the transfer tab. Afterwords the result will be "waiting to report", which it'll do the next time BOINC contacts the scheduler. Reporting is where it sends the administrative details about the result - exit status, CPU time, error messages, things like that.
Some earlier versions of BOINC had a bug where it reported results right after uploading. That was fixed in 5.2. Now it'll contact the scheduler to report if it reaches the deadline or 24 hours after the workunit finishes. If BOINC contacts the scheduler for some other reason, it'll report the results then. Like if you manually "update" the project or BOINC requests more work.
Idea is to reduce the times BOINC contacts the scheduler to report work or request it, and this way it combines several requests into one contacts. With a short "connect every" interval it'll report the previous result when it requests a new workunit. With a longer interval it can report several workunits either after 24 hours have passed or when it requests more work.
And even though a result gets uploaded, the server won't show it as complete until its reported. And it'll use that as the "completion" time, even though the result had already been uploaded.
Walt
Hello All
Those last couple of posts came not a moment too soon: I was about to post a query about that behaviour (grin). Jusdging by the way some of the servers have struggled to keep up with their loads, I reckon this is actually quite a useful idea...
So if I go with 4.19 and it's inherent higher claim, granted credit will be increased on at least 63% of WUs I crunch.
Michael
Michael,
BTW: If you are really serious about cranking up your claimed credit you should look here for an optimized BOINC client (http://www.marisan.nl/seti/). I think I told you that I installed one a couple of weeks ago (when I also installed an optimized Seti app). The increased benchmarks I got increased my Einstein claimed credits by about 20%. (And, if you remember, at the time, I had no idea why my numbers were so much better than everybody else's in the quorum.)
Just a quick report back - as you know the "run based on preferences" is selected, in which I have said to change every hour - the idea being to rotate between the projects. Granted it is early days still and BOINC may still be settling in as it were, however Einstein is still grabbing total time. I admit that the deadlines for Einstein are a mere 11 days away, while Rosetta is only due mid November and Climate is due October 2006, so that might be the problem right there - BOINC is deciding what work units to do based purely on deadlines, rather than on preferences that I choose. I am not sure if this is by design or not. It would appear that I may have to physically suspend a project to get the other projects for the meantime until this is either changed or the devs state clearly that this is by design.
Gray
The switch intraval is not a 'change now' type thing. It really means at least that often the client will reevaluate what it should be working on, this may cause it to work on a different project or another round of the same project. In general fooling with the client to make it 'balance' properly will only increase the time it takes to find that 'balance', so don't suspend just let it run. There is one thing you may want to check to make sure the client is working on getting things straight, make sure in client_state.xml that all the 'debt' tags have numbers in them. Some of the older versions would put letters in there due to errors.
The best thing for you to do to get it to balance out sooner is reduce your queue. The effective size of the queue is number of projects * number of days, so if you want a 3 day queue and are attached to 3 projects you should set your 'connect every' setting to 1 day.
The 5.2.x versions also will get more accurate in their estimates of time to complete as you process work so if you are thinking double my queue since my computer takes half as long as estimated you need to rethink. This may also be behind the problem you are seeing. Since the estimates are (will be) better in the 5.2.x version they are given more weight in CPU scheduling. If the estimates are currently too high this may cause EDF mode prematurely.
Just a quick report back - as you know the "run based on preferences" is selected, in which I have said to change every hour - the idea being to rotate between the projects. Granted it is early days still and BOINC may still be settling in as it were, however Einstein is still grabbing total time. I admit that the deadlines for Einstein are a mere 11 days away, while Rosetta is only due mid November and Climate is due October 2006, so that might be the problem right there - BOINC is deciding what work units to do based purely on deadlines, rather than on preferences that I choose. I am not sure if this is by design or not. It would appear that I may have to physically suspend a project to get the other projects for the meantime until this is either changed or the devs state clearly that this is by design.
Gray
The switch intraval is not a 'change now' type thing. It really means at least that often the client will reevaluate what it should be working on, this may cause it to work on a different project or another round of the same project. In general fooling with the client to make it 'balance' properly will only increase the time it takes to find that 'balance', so don't suspend just let it run. There is one thing you may want to check to make sure the client is working on getting things straight, make sure in client_state.xml that all the 'debt' tags have numbers in them. Some of the older versions would put letters in there due to errors.
The best thing for you to do to get it to balance out sooner is reduce your queue. The effective size of the queue is number of projects * number of days, so if you want a 3 day queue and are attached to 3 projects you should set your 'connect every' setting to 1 day.
The 5.2.x versions also will get more accurate in their estimates of time to complete as you process work so if you are thinking double my queue since my computer takes half as long as estimated you need to rethink. This may also be behind the problem you are seeing. Since the estimates are (will be) better in the 5.2.x version they are given more weight in CPU scheduling. If the estimates are currently too high this may cause EDF mode prematurely.
Hello Michael
OK I will not suspend the BOINC projects, unless I'm writing a CD - sorry, I'll not take a chance of destroying one just to crunch another 2% of Einstein...
For the rest I can comment that BOINC 5.2.2 seems to get out of the way far more effectively than previous versions when I want to play a game ("Settlers - Heritage of Kings", if you want a name). I have yet to do any serious artwork with Paintshop Pro though. I HAVE told the projects to not get more work just yet. Do you think that would affect the ability to balance itself there ?
Just had a thought - BOINC is running on my main, and only (sob) PC, and my usage thereof varies - some nights I'll go read a book (not much of late - I seem to be spending quite some time doing my best to drive you mad on this message board...) - so I wonder whether BOINC will ever be able to work out guestimation times etc - if I fire up a game, BOINC gets just the odd cpu cycle, while if I am surfing as per now then BOINC kinda dominates the cpu.
Just had a thought - BOINC is running on my main, and only (sob) PC, and my usage thereof varies - some nights I'll go read a book (not much of late - I seem to be spending quite some time doing my best to drive you mad on this message board...) - so I wonder whether BOINC will ever be able to work out guestimation times etc - if I fire up a game, BOINC gets just the odd cpu cycle, while if I am surfing as per now then BOINC kinda dominates the cpu.
Gray
Gray,
Boinc is very good at taking into account the varying amounts of CPU time that it gets.
The "old" Boinc client was far off in it's estimation of To Completion values for enqueued WUs, but once a WU started being crunched, To Completion time rapidly converged with an accurate estimation. That was my experience of it, anyway.
The 5.x version of Boinc behaves almost exactly the opposite. I don't want to go into it further now, because everything that I had observed and had to report about is already included, in the first few days of posting on this thread - you can read it there, in detail.
I tested and reported, and received an explanation. Then I expressed pretty strongly my opinion and suggestions, which may have been going well beyond the function of a beta tester. Having said everything I had to say on the subject, I let it go, and returned to further testing.
Michael
microcraft
"The arc of history is long, but it bends toward justice" - MLK
A good piece of news (sort of);
I unattached myself from Rosetta and then went around all the websites of BOINC projects that I belong to and reset the days between contacting the servers to 3 days on each project. I updated Einstein and Climate on my PC and suddenly BOINC pre-empted Einstein (sorry :) ) and started Climate all by itself, so it looks like my problem was too great an interval period between connections to the servers. Whatever the reason, it seems as tho my issues are sorted. One last comment - when uploading a result I often see 101% in the transfer window !! Possibly this should be adjusted for the next release ?
A good piece of news (sort of);
I unattached myself from Rosetta and then went around all the websites of BOINC projects that I belong to and reset the days between contacting the servers to 3 days on each project. I updated Einstein and Climate on my PC and suddenly BOINC pre-empted Einstein (sorry :) ) and started Climate all by itself, so it looks like my problem was too great an interval period between connections to the servers. Whatever the reason, it seems as tho my issues are sorted. One last comment - when uploading a result I often see 101% in the transfer window !! Possibly this should be adjusted for the next release ?
Gray
Just one final note: I am now running Einstein, Rosetta, Seti, Climate and Predictor (LHC are not up-to-date with their website) and have specified connecting to webserver interval as one day. BOINC seems to happily swap between each one. There is a delay each time a project is added to the manager before BOINC will add the project to its routine and in my case this seems to be a day - the connection interval. Obviously the main thing is to avoid large connection intervals in the general preferences or alternatively merely run one project. I suppose if one were to wait the 10 days that I originally specified then the function would be the same - regualr swapping - but realistically speaking 1 day seems fine, and with 5 projects on the go, it is unlikely I will run out of work to crunch...
RE: (Now back to "normal
)
Stick,
I'll take the silver-tongue plea under consideration. I haven't made a move yet, but I was pretty much waiting, "hanging in there", until my testing was concluded. Now that 5.2.2 has gone Prime-Time, the devs probably won't need any more of my feedback, especially that of the vehement, or as Gary so graciously called it "vigorous", sort. lol
As to the "normal mode" topic - I posted my thoughts on the credit thing in my Automatic Benchmarking" thread in Cruncher's corner a couple of months ago, when I was just getting my feet wet in the message boards. My feelings on that haven't changed one iota in the interim, but reality has set in. The bench/quorum method is patently unfair, and will always be so. While Einstein WUs invariably require a set amount of work, on some other projects, the amount of processing needed can vary widely from one WU to the next. That was sometimes the case with SetiClassic, sets of "broken" WUs distributed occasionally. The reality is that now, everything is "locked-in" by dependence upon Boinc for the purposes of cross-project compatibility in credit count, which is what draws the crunchers and keeps 'em crunching. I, for one, greatly appreciated that aspect of SetiClassic, that you did a day's (WU's) work and you received a day's (ahem) pay.
Michael
microcraft
"The arc of history is long, but it bends toward justice" - MLK
Walt, My Seti result was
)
Walt,
My Seti result was reported just as you explained it would be. (It's nice to have software that does what it's supposed to!) ;-) Anyway, I will remember your explanation (and this post) as, inevitably, the question will come up again as users begin to change over.
Stick
RE: Walt, My Seti result
)
Hello All
Those last couple of posts came not a moment too soon: I was about to post a query about that behaviour (grin). Jusdging by the way some of the servers have struggled to keep up with their loads, I reckon this is actually quite a useful idea...
Gray
RE: So if I go with 4.19
)
Michael,
BTW: If you are really serious about cranking up your claimed credit you should look here for an optimized BOINC client (http://www.marisan.nl/seti/). I think I told you that I installed one a couple of weeks ago (when I also installed an optimized Seti app). The increased benchmarks I got increased my Einstein claimed credits by about 20%. (And, if you remember, at the time, I had no idea why my numbers were so much better than everybody else's in the quorum.)
Stick
RE: Hello Michael Just a
)
The switch intraval is not a 'change now' type thing. It really means at least that often the client will reevaluate what it should be working on, this may cause it to work on a different project or another round of the same project. In general fooling with the client to make it 'balance' properly will only increase the time it takes to find that 'balance', so don't suspend just let it run. There is one thing you may want to check to make sure the client is working on getting things straight, make sure in client_state.xml that all the 'debt' tags have numbers in them. Some of the older versions would put letters in there due to errors.
The best thing for you to do to get it to balance out sooner is reduce your queue. The effective size of the queue is number of projects * number of days, so if you want a 3 day queue and are attached to 3 projects you should set your 'connect every' setting to 1 day.
The 5.2.x versions also will get more accurate in their estimates of time to complete as you process work so if you are thinking double my queue since my computer takes half as long as estimated you need to rethink. This may also be behind the problem you are seeing. Since the estimates are (will be) better in the 5.2.x version they are given more weight in CPU scheduling. If the estimates are currently too high this may cause EDF mode prematurely.
BOINC WIKI
BOINCing since 2002/12/8
RE: RE: Hello
)
Hello Michael
OK I will not suspend the BOINC projects, unless I'm writing a CD - sorry, I'll not take a chance of destroying one just to crunch another 2% of Einstein...
For the rest I can comment that BOINC 5.2.2 seems to get out of the way far more effectively than previous versions when I want to play a game ("Settlers - Heritage of Kings", if you want a name). I have yet to do any serious artwork with Paintshop Pro though. I HAVE told the projects to not get more work just yet. Do you think that would affect the ability to balance itself there ?
Gray
Hello Michael Just had a
)
Hello Michael
Just had a thought - BOINC is running on my main, and only (sob) PC, and my usage thereof varies - some nights I'll go read a book (not much of late - I seem to be spending quite some time doing my best to drive you mad on this message board...) - so I wonder whether BOINC will ever be able to work out guestimation times etc - if I fire up a game, BOINC gets just the odd cpu cycle, while if I am surfing as per now then BOINC kinda dominates the cpu.
Gray
RE: Hello Michael Just had
)
Gray,
Boinc is very good at taking into account the varying amounts of CPU time that it gets.
The "old" Boinc client was far off in it's estimation of To Completion values for enqueued WUs, but once a WU started being crunched, To Completion time rapidly converged with an accurate estimation. That was my experience of it, anyway.
The 5.x version of Boinc behaves almost exactly the opposite. I don't want to go into it further now, because everything that I had observed and had to report about is already included, in the first few days of posting on this thread - you can read it there, in detail.
I tested and reported, and received an explanation. Then I expressed pretty strongly my opinion and suggestions, which may have been going well beyond the function of a beta tester. Having said everything I had to say on the subject, I let it go, and returned to further testing.
Michael
microcraft
"The arc of history is long, but it bends toward justice" - MLK
Hello Michael A good piece
)
Hello Michael
A good piece of news (sort of);
I unattached myself from Rosetta and then went around all the websites of BOINC projects that I belong to and reset the days between contacting the servers to 3 days on each project. I updated Einstein and Climate on my PC and suddenly BOINC pre-empted Einstein (sorry :) ) and started Climate all by itself, so it looks like my problem was too great an interval period between connections to the servers. Whatever the reason, it seems as tho my issues are sorted. One last comment - when uploading a result I often see 101% in the transfer window !! Possibly this should be adjusted for the next release ?
Gray
RE: Hello Michael A good
)
Just one final note: I am now running Einstein, Rosetta, Seti, Climate and Predictor (LHC are not up-to-date with their website) and have specified connecting to webserver interval as one day. BOINC seems to happily swap between each one. There is a delay each time a project is added to the manager before BOINC will add the project to its routine and in my case this seems to be a day - the connection interval. Obviously the main thing is to avoid large connection intervals in the general preferences or alternatively merely run one project. I suppose if one were to wait the 10 days that I originally specified then the function would be the same - regualr swapping - but realistically speaking 1 day seems fine, and with 5 projects on the go, it is unlikely I will run out of work to crunch...
Gray