| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Help Conquer Cancer
Thread: Clock is counting up, not down |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 16
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I can confirm this. Indeed the remaining time is counting small steps upwards, and then it suddenly decreases a big step.
|
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges: 
|
Now that you write it Lawrence I remember better. Indeed you are right, HCC does not behave exactly like DDDT (which is very strange) but it is very not-linear too.
---------------------------------------- |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I'm still unsure whether there is a problem needing investigation here, or whether it is all explicable by estimation errors and work unit variations.
So, if anyone can provide an example of a repeatable, significant clock error, please say so. Repeatable means that the same error happens under the same circumstances, even on different work units. Significant would be an error of, say, about 5% of the work unit length. Thank you! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I gazed at it again and it seems the time decreases only when the percentage jumps up a step.
This kind of makes sense (my theory): because the percentage does not grow, the time estimation algorithm thinks the calculation is not advancing, and thus grows the estimated time left. Possible solutions are to update the percentage more often (preferably as often as the time estimate), or to enchange the time estimation algorithm so that does not stare just the percentage, but a longer-term average of calculation speed. This from a software engineer who knows absolutely nothing about what's under the hood of this client :) |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges:
|
I'm still unsure whether there is a problem needing investigation here Regarding the DDDT strange-percentage-counting problem I do not know if it is the result of a mistake or if it is caused by some tricky technical constraint. If it is a mistake it would certainly be more elegant if it were fixed sometime when there is nothing more urgent to fix. But in any case this is not an important problem since it does not forbid the WUs to go their way to their end and the variations of times are not exceeding 2 or 3 minutes vs what they would be if the percentage were counted correctly. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Yes, this is only cosmetic, and it doesn't stop us from conquering cancer, but it's pretty darn confusing at first.
----------------------------------------BTW, it's way more prominent on slower machines, I installed BOINC to another host with more muscle, and it's not as disturbing, as the percentage is updated more frequently. So if you want to reproduce the problem, try it with a more ancient unit. edited for language...cih [Edit 1 times, last edit by Former Member at Dec 10, 2007 1:46:47 PM] |
||
|
|
|