| 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 AsyncRE work units started |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 47
|
|
| Author |
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges: 
|
Greetings everyone,
We are starting FightAIDS@Home - Phase 2 AsyncRE work units. These work units behave a little bit different than the previous FAH2 work units. 1. They are shorter in runtime ( 1.5 - 5 hours ) 2. Trickle messaging and intermediate uploads will not happen. 3. Work units have a short deadline of 24 hours. 4. We are limiting results to 1 per core, similar to how beta runs. The work units for FAH2 will have an additional backend component that we have been testing that'll help improve the results for the researchers before the subsequent work units are released. This is still the same children/grand children concept from before, so work units will appear to have the same numbering sequence to the members. One benefit of this is the overall CPU time used to compute the results for a given batch will be less. To start off with the first early batches we will be doing some experimenting with both the size of the work units, as well as total number of steps required to complete a batch. The initial release will have 451 batches with 10k steps and 451 batches with 50k steps. With the results from these batches we hope to find an optimal setting going beyond these batches. Thanks, -Uplinger |
||
|
|
branjo
Master Cruncher Slovakia Joined: Jun 29, 2012 Post Count: 1892 Status: Offline Project Badges: 
|
Thanks uplinger for info.
----------------------------------------NI!  Crunching@Home since January 13 2000. Shrubbing@Home since January 5 2006  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Since we only get 1 per core, can they be given priority in the feeder scheme?
|
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
The dirty trick is to set a MinB buffer greater than 12 hours and BOINC will go nuts on them in EDF. If you don't care what else is crunched, set FAH2 as main and the "If there is no work for..." Theoretically AsyncRE will always jump queue, up to a point when the other work deadlines become too tight.
----------------------------------------Far as i know there's nothing really but the deadline value to compel a client to prioritize. This presumes there won't be a supply chain shortage. [Edit 1 times, last edit by SekeRob* at Nov 1, 2017 9:04:27 PM] |
||
|
|
wolfman1360
Senior Cruncher Canada Joined: Jan 17, 2016 Post Count: 176 Status: Offline Project Badges: 
|
I just got 4 invalids in a row on these new WUs.
----------------------------------------Is this something to be concerned about?
Crunching for the betterment of human kind and the canines who will always be our best friends.
AWOU! |
||
|
|
PMH_UK
Veteran Cruncher UK Joined: Apr 26, 2007 Post Count: 764 Status: Offline Project Badges:
|
So far I have 5 invalids and 3 in progress.
----------------------------------------None valid for wingmen yet. I had no invalids during beta AFAIK. Looks like there may be an issue. Paul. Edit: typo.
Paul.
----------------------------------------[Edit 1 times, last edit by PMH_UK at Nov 1, 2017 10:18:12 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Oh dear, me too. Two invalid on Linux, 2 more in progress and 2 in progress on windoze.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The dirty trick is to set a MinB buffer greater than 12 hours and BOINC will go nuts on them in EDF. If you don't care what else is crunched, set FAH2 as main and the "If there is no work for..." Theoretically AsyncRE will always jump queue, up to a point when the other work deadlines become too tight. Far as i know there's nothing really but the deadline value to compel a client to prioritize. This presumes there won't be a supply chain shortage. That's a good idea once they reach the client. My concern is I won't get very many as they will be overwhelmed by the other work. It would be nice if they sent FAH2 until <num_cores> reached and then fill in with other work. Once one finishes, the feeder should first try and send a FAH2 before looking for other eligible work. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
All four on Linux went invalid. I still have two in progress on windoze. I've switched away from the project until we hear from the techs.
|
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges:
|
Tech here...The invalids are due to the wrong version being loaded when the results started. I had pushed version 7.18 but forgot to restart the feeder. So the feeder was still giving out work units for 7.14/15. This should be fixed going forward as 7.18 is the official release for this application.
The fix in 7.18 is that it will upload the proper rst and out file required for workunits that do not have trickle messages and intermediate uploads initiated. I am sorry about this and at the moment only had 2 batches that were released, FAH2_001319 and FAH2_001320, both were 10k step workunits. Thanks, -Uplinger |
||
|
|
|