| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: Mapping Cancer Markers Forum
Thread: MCM Progress Reports |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 957
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Reads like 1, you know the batch count, 2, the 'research page percent' is 3 weeks ahead of the actual completion. Since there has not been a word of hitting onto the second target, yet you show 25%, one of 4 targets, one could derive we're under 3 weeks away of the 2nd, if continuing seamlessly. Makes for a guess of the error rate being higher than anticipated. More assumption, 350 / 4 (target) = 87.5 million per target, if equal sized. With 115 million validated at this time makes the rate 31% higher than the original estimated distribution of 2 plus original redos projection. Maybe borrow harry potter's wand and do the 'reveal' spell on you.
|
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7844 Status: Offline Project Badges: 
|
I do not have a real good way to track how many resends there might be be, but I have looked at the last 9484 MCM1 jobs I have processed (from about April 1). Of this number I extracted the last digit. If the last digit is a "0" or a "1" it means that WU is not a resend (granted, there may be resends for the unit which went to a wingman, but I have no way of knowing this.) If the end digit is a "2" or greater, then that WU is a resend for sure. Of those I have 8396 are either a "0" or a "1" which is 88.528%. Of the remaining 11.472%, 972 are "2", 107 are "3", and 9 are "4".
----------------------------------------Because I maintain short queues (24 hours or less) I presume I get resends that are about average for the project. Therefore the original 350,000,000 WU would need to be incremented by about 12% or 42,000,000. The next time I do an estimate, maybe I will include this figure in addition to my original figures. I know I am making a lot of assumptions here, but I am open to any other ideas people may have to try to be more accurate. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges: 
|
If you want an actual batch number we have loaded batch 5789 out of the 17,500 batches they are estimated to give us. As for if the targets are at the 25% mark for each one, I do not know the answer to that. It could be the first of the 4 targets is the most difficult and is actually going to be 50% of the work, I don't know. But remember, The estimate given is based off of time.
Thanks, -Uplinger |
||
|
|
cjslman
Master Cruncher Mexico Joined: Nov 23, 2004 Post Count: 2082 Status: Offline Project Badges: 
|
There are several possibilities. First, you may be correct, they have not updated their research page. Second, they may have changed the total number of workunits they anticipate having in the project, if they have added more workunits, it will affect the percentage done. Third, they may be using an anticipated number of hours for the project in their calculations rather than number of workunits. Since the workunit length is highly variable in this project, this would affect the percentage done on a time basis. Fourth, there may be some other reason of which I have not thought. If I learn any new information, I will incorporate it into the estimates. Cheers I checked the Research page and MCM is now showing 38% progress. I also asked in the Website Support forum how often is the progress updated and if it is automatic or manual. CJSL Crunching for a better world... |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Tried to model the uplinger info and got garble, gigo effects i suppose. The closest I get from "This is based off how fast batches are loading" is, that round about batch 6650 is now at the backend of the conveyor to the feeder spout, the conveyor 3 weeks of work supply long?
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The percentages progress jumped ahead for all, fa@h from 77 to 88 and cep2 from 95 to 97.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Uplinger,
Can you check these WU's: MCM1_0005556_0916 MCM1_0005572_5566 MCM1_0005591_5584 Both of these the wingman errored out around 7-days after the WU was sent to them. I would have expected no reply or would an error show up if they started the WU but the time ran out? I have seen quite a few resends where the wingman just errors out 6 to 7 days after the WU was sent to them. |
||
|
|
cjslman
Master Cruncher Mexico Joined: Nov 23, 2004 Post Count: 2082 Status: Offline Project Badges:
|
The percentages progress jumped ahead for all, fa@h from 77 to 88 and cep2 from 95 to 97. ha... CEP2 went down to 62% ... go figure. CJSL Crunching for a brighter future... |
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges:
|
Uplinger, Can you check these WU's: MCM1_0005556_0916 MCM1_0005572_5566 MCM1_0005591_5584 Both of these the wingman errored out around 7-days after the WU was sent to them. I would have expected no reply or would an error show up if they started the WU but the time ran out? I have seen quite a few resends where the wingman just errors out 6 to 7 days after the WU was sent to them. Basically in this case the client realizes it won't complete the work unit by the deadline. It sends a -233 exit code which means not started in time. This is similar to the No reply, but the agent recognizes the issue before the server does and sends a signal back. No reply is we never hear back from the client about that particular result. Thanks, -Uplinger |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Basically in this case the client realizes it won't complete the work unit by the deadline. It sends a -233 exit code which means not started in time. This is similar to the No reply, but the agent recognizes the issue before the server does and sends a signal back. No reply is we never hear back from the client about that particular result. Thanks, -Uplinger Thanks, so most likely they have too deep of a buffer for how much work their computer can actually complete then. I had another WU where it was resent after the 7-day and the wingman showed as no reply. Two days later, the wingman reported the result so it showed as valid. So, if the resend was already started, then three results get returned and a waste of CPU time; just another example of too deep of a buffer. The CPU time was only four hours for the wingman; so if it took them a little over 9-days to return it, either the machine is not on very much at all; or they run a deep buffer which the variable WU run times just make it worse. Thanks for checking on them. |
||
|
|
|