1 percent status and Time to Completion

Message boards : Number crunching : 1 percent status and Time to Completion

To post messages, you must log in.

1 · 2 · 3 · Next

AuthorMessage
bzag00

Send message
Joined: 17 Aug 06
Posts: 4
Credit: 4,765
RAC: 0
Message 37986 - Posted: 18 Mar 2007, 22:57:59 UTC
Last modified: 18 Mar 2007, 23:26:11 UTC

I just had an update to the Rosetta cruncher and I've noticed two things:
1. The 1 percent status seems to remain abysmally long - like an hour and a half
2. The Time to Completion is always increasing.

On the face of it, this implies the program is no longer working, so rather waste CPU time I detached from the project in BOINC. I don't know if I can reattach and get my statistics back, but I'm not looking to reattach if the 5.51 version is not going to work for me. The 1 percent status has always been an annoyance, and such things should be deprogrammed from occurring, whatever it takes to make it happen, or else those of us with little patience will get disenchanted with the project's effort sooner than later.

If I've drawn any wrong conclusions on the techy side of the above arguments, I'd like to hear what went wrong, if anything went wrong. I'm running W2KSP4 up-to-date with security patches. My other BOINC projects are still going fine (I am running V5.2.7 Boinc Manager).
ID: 37986 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Ethan
Volunteer moderator

Send message
Joined: 22 Aug 05
Posts: 286
Credit: 9,304,700
RAC: 0
Message 37990 - Posted: 19 Mar 2007, 2:09:28 UTC

What you're seeing is normal. . check out Feet1st's FAQ:

https://boinc.bakerlab.org/forum_thread.php?id=1754&nowrap=true#17972
ID: 37990 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
sriganesh

Send message
Joined: 3 Mar 07
Posts: 2
Credit: 4,830
RAC: 0
Message 37994 - Posted: 19 Mar 2007, 4:05:32 UTC - in response to Message 37986.  

Hi,

I have been running a task, that had about 11.5hrs to completion. But in-spite of it running for nearly a day (many hrs for this alone), the Time to completion was still 10+hrs. Actually it reduced to 9+hrs, but later it increased again. I am don't know why it happened.

I aborted this task (had to give others a useful time).

Please let me know, the reasons for this and how to avoid it in future.

Thanks
:) Brijesh
ID: 37994 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
sriganesh

Send message
Joined: 3 Mar 07
Posts: 2
Credit: 4,830
RAC: 0
Message 37995 - Posted: 19 Mar 2007, 4:09:07 UTC - in response to Message 37994.  

I am sorry, I had not read the link.

Thanks,
:) Brijesh

Hi,

I have been running a task, that had about 11.5hrs to completion. But in-spite of it running for nearly a day (many hrs for this alone), the Time to completion was still 10+hrs. Actually it reduced to 9+hrs, but later it increased again. I am don't know why it happened.

I aborted this task (had to give others a useful time).

Please let me know, the reasons for this and how to avoid it in future.

Thanks
:) Brijesh


ID: 37995 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Angus

Send message
Joined: 17 Sep 05
Posts: 412
Credit: 321,053
RAC: 0
Message 37998 - Posted: 19 Mar 2007, 6:10:30 UTC - in response to Message 37990.  
Last modified: 19 Mar 2007, 6:14:53 UTC

What you're seeing is normal. . check out Feet1st's FAQ:

https://boinc.bakerlab.org/forum_thread.php?id=1754&nowrap=true#17972


None of this is "Normal"

The progress % done should rise evenly from the start of the WU to the finish, and the time to completion should decline in a linear fashion, from start to finish.

You're just making excuses for sloppy programming of the interface.

Neither of the only two progress indicators is worth diddly-poo here.

As for watching the "steps" progress, that's not displayed in the "Manager" at all, and is only visible if you decide to risk total system crashes by trying to run the buggy BOINC graphics.
Proudly Banned from Predictator@Home and now Cosmology@home as well. Added SETI to the list today. Temporary ban only - so need to work harder :)



"You can't fix stupid" (Ron White)
ID: 37998 · Rating: -1 · rate: Rate + / Rate - Report as offensive    Reply Quote
John

Send message
Joined: 18 Mar 07
Posts: 24
Credit: 0
RAC: 0
Message 38003 - Posted: 19 Mar 2007, 12:53:23 UTC

How come WCG can do this nicely but Rosetta can't?
ID: 38003 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Christoph

Send message
Joined: 10 Dec 05
Posts: 57
Credit: 1,512,386
RAC: 0
Message 38004 - Posted: 19 Mar 2007, 13:09:54 UTC

I have already posted a proposal for the progress computation here.
ID: 38004 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Feet1st
Avatar

Send message
Joined: 30 Dec 05
Posts: 1755
Credit: 4,690,520
RAC: 0
Message 38006 - Posted: 19 Mar 2007, 13:46:37 UTC

Angus, I know you've been away for a while, so you probably missed the news, but the graphics problems that were causing some tasks ("tasks", not "systems") to fail have been corrected for some time.

To tell people they are risking total system crashes really just sounds like you are trying to be an alarmist. Especially since you've posted no reports of such a "total system crash".
Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/
ID: 38006 · Rating: -1 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Angus

Send message
Joined: 17 Sep 05
Posts: 412
Credit: 321,053
RAC: 0
Message 38008 - Posted: 19 Mar 2007, 14:18:21 UTC - in response to Message 38006.  
Last modified: 19 Mar 2007, 14:44:28 UTC

Angus, I know you've been away for a while, so you probably missed the news, but the graphics problems that were causing some tasks ("tasks", not "systems") to fail have been corrected for some time.

To tell people they are risking total system crashes really just sounds like you are trying to be an alarmist. Especially since you've posted no reports of such a "total system crash".


Ignoring the fact that the topic is the "1% status" and "Time to Completion" problems, you certainly must realize that BOINC, not Rosetta, has had problems displaying graphics or screensaveers since it was first released.

It's not uncommon to have a system freeze or blue-screen when attempting to run the BOINC graphics, if you even have a video card and OS and driver combination that will run the openGL graphics at all.

http://boinc.berkeley.edu/dev/forum_thread.php?id=1605
http://boinc.berkeley.edu/dev/forum_thread.php?id=1261
https://boinc.bakerlab.org/rosetta/forum_thread.php?id=1388

The point is, the user should NOT have to depend on buggy or non-existent graphics to see if the WU is actually working properly.

It should be accurately displayed in BOINC Manager, and it's not. You cannot argue that fact.



Proudly Banned from Predictator@Home and now Cosmology@home as well. Added SETI to the list today. Temporary ban only - so need to work harder :)



"You can't fix stupid" (Ron White)
ID: 38008 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Feet1st
Avatar

Send message
Joined: 30 Dec 05
Posts: 1755
Credit: 4,690,520
RAC: 0
Message 38011 - Posted: 19 Mar 2007, 14:47:19 UTC

What I'm questioning Angus, is whether it is constructive to speak about monitoring the graphic, and a risk of a "system crash", and "sloppy programming" in a thread created by a participant with 1,000 credits and a RAC of 1.

The project has the watchdog to watch after things for you. Monitoring tasks at all is not a requirement to successfully run the project.

If the project has done something to upset you, I would certainly prefer to see you take it up with them rather then making alarmist statements in the threads of new participants.
Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/
ID: 38011 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Angus

Send message
Joined: 17 Sep 05
Posts: 412
Credit: 321,053
RAC: 0
Message 38013 - Posted: 19 Mar 2007, 15:01:47 UTC

What upsets me is you preaching the "Everything is OK - ignore the broken monitoring tools", when it's obvious that they are not working correctly, and are of no use to those who are inquisitive enough to want to monitor the progess of the work they are donating.
Proudly Banned from Predictator@Home and now Cosmology@home as well. Added SETI to the list today. Temporary ban only - so need to work harder :)



"You can't fix stupid" (Ron White)
ID: 38013 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Feet1st
Avatar

Send message
Joined: 30 Dec 05
Posts: 1755
Credit: 4,690,520
RAC: 0
Message 38015 - Posted: 19 Mar 2007, 15:22:13 UTC

I certainly agree that it is an issue that can be, and should be resolved with programming changes. However, in programming terms, it is not "broken". It is "working as designed", and you (and I) happen to feel that the design could be improved to be more intuitive to new users. But things ARE working "correctly", the way they were designed to work.

So, I agree, and I'll take every opportunity to point it out to the project team. Which I did prior to you ever posting to this thread. But, I for one am glad to see them rolling out new code that begins the study of docking, and RNA, and new tasks that are studying potential vaccines for HIV. So, there's something to be said for perfecting what you've got before you add more to it, but the Project Team obviously chose to organize their priorities differently. ...and it's their decision to make.
Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/
ID: 38015 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
bzag00

Send message
Joined: 17 Aug 06
Posts: 4
Credit: 4,765
RAC: 0
Message 38020 - Posted: 19 Mar 2007, 17:08:57 UTC - in response to Message 38015.  

I find it sad that projects that are so important cannot be improved to a level to give a measure of confidence to the wider user community. Simply stating that it works is simply no longer good enough.

Suppose you walked into a car dealer and he told you he was selling vehicles that now could get you from NY to Boston South Station with automatic steering, brake control, and collision avoidance, but that the problem they haven't been able to overcome is that the windshield is opaque and you can't look outside while the car is moving. Would you buy the car, even though you had statistics that showed you the car was 99.9999999% perfect?

This is the problem with getting the wider community to accept the project. We must be able to see for ourselves that the project is working.


ID: 38020 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Feet1st
Avatar

Send message
Joined: 30 Dec 05
Posts: 1755
Credit: 4,690,520
RAC: 0
Message 38031 - Posted: 20 Mar 2007, 0:29:55 UTC

...and if you were given a car, and it had a fancy GPS navigation system with voice activation, and you jumped in and said "plan route home", and then asked it how long the trip will take, and what % complete we are right now, it would have a difficult time answering your questions. This is because it doesn't know where "home" is.

Now there is ANOTHER car over here, that SAYS it KNOWS the answers to those questions, but it turns out it is often wrong. ...which car is better? Is one car broken?

I would assert that BOTH cars could be "improved". So let's talk about ideas on how the improvements might be done.

What if you start with an estimated runtime that is equal to your target runtime, and then compute a new % complete every 10 seconds as follows:
total runtime so far / (last estimated runtime - 10 seconds)
That should result in a % complete that would multiply through to a estimated runtime that would be declining every 10 seconds. Then, at the end of each model, recompute the % complete the way it already does. This should give the desired illustration of "progress", and instantly adjust to new runtime preference (well... unless you make it smaller I suppose). And it would just have a shift up or down in the estimated time to complete at the end of each model, but that shift would be like really minimal compared to the jump it takes now.

If you interrupt a task with a runtime preference that is smaller then the amount already crunched on a task, then the estimate would have to be based on assuming one more model will be crunched, as compared to the number of models that were crunched previously on that task. This may not be accurate, because we might be 3/4 done with that last model, and have checkpointed that work, but for as often as it is going to happen, it would still be closer then the current system.
Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/
ID: 38031 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Feet1st
Avatar

Send message
Joined: 30 Dec 05
Posts: 1755
Credit: 4,690,520
RAC: 0
Message 38035 - Posted: 20 Mar 2007, 3:32:27 UTC

Of course the problem with that approach is a long running task on a slow system with a very low preferred runtime. Your preference might be for 1hr, but it may take your machine 3 hrs to crunch that required first model. And the method I just proposed wouldn't handle that very well. And there is no great way to know ahead of time how long it is going to take your machine to crunch that first model.

"Use the BOINC benchmarks" people will say, "this will tell you the machine's capacity". But it's not so. The benchmarks tell you how fast a program that requires nil for memory can perform floating point operations, which doesn't always give an accurate estimate of the machine's efficiency running Rosetta. You can see this in the credit claimed as compared to credit granted figures. For some machines they are consistently within 2%, for others, they are consistently more like 50% off... and off in some cases under estimating and in other cases over estimating.

Would there be any way to use the experience with the task on Ralph?? Ralph's credit per model is really the best estimate available about how extensive the task's models are.
Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/
ID: 38035 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
bzag00

Send message
Joined: 17 Aug 06
Posts: 4
Credit: 4,765
RAC: 0
Message 38043 - Posted: 20 Mar 2007, 12:30:40 UTC - in response to Message 38031.  

...and if you were given a car, and it had a fancy GPS navigation system with voice activation ...


Two points:

1. I accepted the original argument about the steppings at the beginning of this post and have reattached, and remained reattached after testing it and watching it for awhile to verify it is working.

2. It is absolutely irrelevant to the argument itself that I am "given" the car, independent of the fact that, on the face of it, I am not the one getting the direct benefit, but actually "giving" CPU time to the project. My point, regardless of my indirect benefit as a member of society that suffers from these diseases, has more to do with getting on with it than with paying homage to gratuitous stroking.



ID: 38043 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
MattDavis
Avatar

Send message
Joined: 22 Sep 05
Posts: 206
Credit: 1,377,748
RAC: 0
Message 38058 - Posted: 20 Mar 2007, 18:43:29 UTC - in response to Message 38020.  



This is the problem with getting the wider community to accept the project. We must be able to see for ourselves that the project is working.



Most of the tens of thousands of users HAVEN'T complained - it sounds like the community accepts the project and only a few people have issues.

ID: 38058 · Rating: 9.9920072216264E-15 · rate: Rate + / Rate - Report as offensive    Reply Quote
BennyRop

Send message
Joined: 17 Dec 05
Posts: 555
Credit: 140,800
RAC: 0
Message 38059 - Posted: 20 Mar 2007, 19:03:33 UTC

The 1% issue was confusing when I first started here, and is currently being discussed in this thread here: 1% status and in one of the FAQs at System seems stuck at 1%

ID: 38059 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Angus

Send message
Joined: 17 Sep 05
Posts: 412
Credit: 321,053
RAC: 0
Message 38069 - Posted: 21 Mar 2007, 3:01:14 UTC
Last modified: 21 Mar 2007, 3:02:17 UTC

1% bug still there in 5.54

Been running a new WU for 23 minutes on a 1 hour CPU crunch time preference. Should be showing about 33% by now.

Estimated run time was 3:46 when the WU started - now up to 4:05, again on a 1 hour WU run time preference. Two problems here - the estimated time should not be almost 4 hours for a 1 hour crunch time WU, and the time to completion is rising, instead of counting down to zero.

No apparent fixes here.

Proudly Banned from Predictator@Home and now Cosmology@home as well. Added SETI to the list today. Temporary ban only - so need to work harder :)



"You can't fix stupid" (Ron White)
ID: 38069 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
P . P . L .

Send message
Joined: 20 Aug 06
Posts: 581
Credit: 4,865,274
RAC: 0
Message 38071 - Posted: 21 Mar 2007, 3:40:28 UTC

Hi Angus.

As your computer is hidden it's hard to help you much

but if your computer is slow, or you are doing other

things with it. That could be what is causing the

problems for you, because some W.U's are more complexed

it takes longer than one hour to do one model.

I know my computer run time is ten hours and sometimes

it sits on one percent for around half an hour then it

jumps up in precentage.

ID: 38071 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
1 · 2 · 3 · Next

Message boards : Number crunching : 1 percent status and Time to Completion



©2024 University of Washington
https://www.bakerlab.org