| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: FightAIDS@Home Phase 2
Thread: FightAIDS@Home, Phase 2 - Crunching Chart |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 64
|
|
| Author |
|
|
Speedy51
Veteran Cruncher New Zealand Joined: Nov 4, 2005 Post Count: 1288 Status: Offline Project Badges: 
|
I thought some people might be interested to see I have a 20th generation task from batch 76 FAH2_ 000076_ avx17684_ 000076_ 0029_ 020_ 0-- When the task completes is there a way of telling if there will be another generation of task created from the task I complete even if the 100,000 steps is reached in the log?
---------------------------------------- [Edit 2 times, last edit by Speedy51 at Mar 4, 2016 7:56:46 AM] |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
You must not have read the post right before yours. The generation series consist of 30 x 100,000 steps [3 Million]. There's also WU's with -ls- in them, these consist of 40 x 100,000 steps [4 million]. So to answer your question, no -ls- in yours, thus 10 more WU's at least will be generated, more if there are results that do not compute the full 100k.
----------------------------------------A tech wrote several times [uplinger I think] the steps count [in future] is subject to change. The scientists were discussing optimization from the test series, so what comes after... [Edit 1 times, last edit by SekeRob* at Mar 4, 2016 8:09:42 AM] |
||
|
|
Speedy51
Veteran Cruncher New Zealand Joined: Nov 4, 2005 Post Count: 1288 Status: Offline Project Badges:
|
Thank you for the explanation. I now understand how things work in terms of the generations. The time that the work unit won't go out again after been completed is if it is a 30 at the end or in the case of been a -is- it will keep going out unless there is a 40 at the end of it.
---------------------------------------- |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
When you see jobs with an extra suffix like
FAH2_000059_avx17257_000085_0014_026_wcgfahb00070000_0 it means that this one kicks off from step 70000 of generation 26. These jobs still have 100,000 in them, but the end job, generation #31, will have just the tail bit of whatever is left. 30000 if the steps are a round 3 million. How or why, but it was written that the scientists preferred for a single generation to compute at least 80000 steps in a run, so the above and this FAH2_000058_avx101140_000077_0018_026_wcgfahb00060000_0 indicates there are blocks around that are not ideal... but we're in a test/eroerterungs phase. Whatever the solution when the real real stuff commences... |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
Getting the half-ways in higher succession ATM. Now
FAH2_avx38781-ls_000084_0019_039_wcgfahb00040000_0 An -ls-, generation 39 starting at 3940000 [or 3940001?]. Batch 84, so long-step job series is close to being done. Would be interesting to know when _001_ was first put in the feeder, to know how long it took to complete the set. [Think it could be around 110-120 days]. No way of us donators to know unless we get tech info or start spreadsheeting from the history. |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7655 Status: Offline Project Badges: 
|
No way of us donators to know unless we get tech info or start spreadsheeting from the history. I do keep track of the WU I do in a spreadsheet. Upon looking at the completed units I am a little confused as to exactly how to interpret the naming and numbering system of the WU's. I know there are two different kinds of units and what is the batch number and what is the job number. It appears each WU name has a number of batches associated with it and each batch has a four digit extension. Then comes the three digit number which identifies which sequence of 100,000 steps is being worked on (up to 30 for some and up to 40 for the -is units. for instance: FAH2_000061_avx17285_000026_0043 _009 _wcgfahb00050000_0 1/16/16 14:55 1/18/16 9:08 FAH2_000061_avx17285_000026_0053 _017 _0 1/19/16 12:40 1/20/16 21:37 FAH2_000061_avx17285_000026_0038 _026 _0 2/27/16 14:21 2/29/16 9:58 Even if the batch numbers are the same, I don't know if the number in red means these are still the same WU series or if they all belong to a different series. The number in pink is the sequence number. The dates are sent time and returned time. If these do belong to the same basic WU sequence, then I could get an indication of how fast the entire 3 or 4 million step job is progressing through the system. If the batch number and the four digit extension need to be identical for this to work, then I don't think I have a big enough sample size for this to have occurred. Then I found this: FAH2_avx17684-ls_000040_0010 _001_0 10/3/15 5:58 10/5/15 11:37 FAH2_avx17684-ls_000040_0016 _001_0 10/3/15 5:58 10/5/15 11:37 I think this is the same unit, because both the sent time and received times are identical, yet the four digit extension to the batch number is different. So now I don't know if it is even possible from the data I have available to track the progress of the 3 or 4 million step jobs through the system. (If this is too off topic for this thread, I will move it elsewhere.) Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
Your last 2 are batch 40, sequence 10 and 16 respectively. There's -ls- and both 001_0 at end meaning they are the first 100K step kick-off jobs of a 4M WU sequence. Did track project launch at the beginning and noted an _030_0 showed up and completed after 78 days. If the number of 200,000 sequences in circulation stands with a current 7 day average of 61K results per day, it takes ~ 3.27 days for a single cycle to return, so 30 x 3.27 = ~98 days. For the -ls- it would be ~130. No wonder the 4 day deadline.
----------------------------------------Remember, anything goes in any thread at WCG. On Topic is an impossibility in this place, but yes, if batch / sequence tracking became a specific topic, it would be better to place in a separate thread. BTW, on the OP images, I've removed the source for a little, hoping PB will eventually delete those ghost copies too. All active charts deleted, 11 old left to see, yet it says there are 486 images in the daily folder. If they don't in 48 hours, I'll sent them a mail again. Placed a browsable slideshow link in my sig and there's just now only those 11 to be seen, there should be ATM. [Edit 1 times, last edit by SekeRob* at Mar 6, 2016 8:13:35 AM] |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7655 Status: Offline Project Badges:
|
Your last 2 are batch 40, sequence 10 and 16 respectively. There's -ls- and both 001_0 at end meaning they are the first 100K step kick-off jobs of a 4M WU sequence. Did track project launch at the beginning and noted an _030_0 showed up and completed after 78 days. If the number of 200,000 sequences in circulation stands with a current 7 day average of 61K results per day, it takes ~ 3.27 days for a single cycle to return, so 30 x 3.27 = ~98 days. For the -ls- it would be ~130. No wonder the 4 day deadline. Thanks, that calculation is what I wanted to achieve. I still find the numbering system somewhat confusing, but I suppose it makes sense for the scientists. The highest number I have seen so far is 100 for the sequence number for any batch.( Might be the limit for this "test" run.) Thus I think it works this way - Name, then batch number (6 digit field), sequence (4 digit field), step number index ( 30 or 40 of the 100,000 step part) and finally the job number (single digit, mostly zeroes as this is a single redundancy project, but a few ones etc. for repair units). Given the size of the fields, they have provisioned themselves for the number of WU's to be in excess of 10^10 jobs, surely way more than they anticipate. But better to have excess numbering capacity than not enough and have to change the system part way through the project. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
Just when you think to see in the data that things are getting towards the end zone, the gradual reduction in production and highest batch 102 being crunched of total 106 batches, there's this jolt again, almost as if new sequences were loaded onto the grid [manual labour, no automation to keep 200,000 sequences on the grid]. Today's projection even hints at a new day record of 86K results, but no telling with this project and the morning half rarely does extrapolate to a full day number. E.g. my 8 core only has FAHB on it and will complete 2 tasks in the afternoon session (runs them just over 24 hours each), foreseeably 4 tomorrow morning and then 2 tomorrow afternoon i.e. my contribution stats are all over the place.
|
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
New record of 86K in a day results it is ... 85,920 to be precise [happening in the weekend on top]. T'was for 119.64 runtime years which is an 'almost', the green vertical bar in the left bottom mini graph indicating the last record for this category.
In picture speaking a thousand words:  |
||
|
|
|