| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Beta Testing
Forum: Beta Test Support Forum
Thread: How can I force more beta units? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 83
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I understand that the scheduler may ignore the size of my buffer. Even so, I have now seen over a dozen Beta WU aborted on my machines because the other WU have been validated by my wingmen.
Therefore the fact that I had a larger buffer than my wingmen means that I did not get the WU started early enough and the system then issued an abort on my unstarted WU. Therefore a smaller buffer would have increased my chances of crunching those aborted WU and finally the size of my buffer DIRECTLY affects my ability to crunch Beta WU. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I have got a steady stream.
1 Set connect & buffer to be .1 day. 2 set only project to be Clean energy - There aint any of these for windows. 3 If I look like running out of work I switch to CMD2 and get a few of these before going back to Clean Energy......  |
||
|
|
Crystal Pellet
Veteran Cruncher Joined: May 21, 2008 Post Count: 1320 Status: Offline Project Badges: 
|
2. When you decrease your buffer back to .1 days, you prevent your computer from asking for more work. This prevents your system from acknowledging the completed WU and may cause your WU to be marked as Too Late. All finished jobs will automatically acknowledged without asking for new work when they are within 24 hours from the deadline. |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: 
|
You have made a couple of deadly mistakes in your suggested method. 1. If you change your buffer to 8 days, you are not guaranteed to get 8 days of Beta. You might get only 1 Beta job and 7.9 days of other tasks. When you abort those tasks, you negatively affect the reliability rating of that machine for many days. Plus you place extra load on the scheduler who then has to resend the jobs you aborted to other crunchers. Not to forget, with a 1.5 days deadline on beta-wu's like now, you won't get 8 days of beta-wu's, and even "best-case" will be stuck with 6.5 days of other work, if you'll not so "lucky" that the sub-project(s) selected has little or no work to give-out. 2. When you decrease your buffer back to .1 days, you prevent your computer from asking for more work. This prevents your system from acknowledging the completed WU and may cause your WU to be marked as Too Late. Totally wrong, you don't need to ask for more work to report finished work. Finished tasks will be reported if: 1; Asks for more work. 2; If been "ready to report" for atleast 24 hours. 3; If less than 24 hours until deadline. 4; If time until deadline is less than "Computer is connected to the Internet about every N days". 5; By manually hitting "update". 6; If sends a trickle-up (currently CPDN only). 7; If configured <report_results_immediately> (not recommended, since adds unnessessary extra server-load). 8; Based on <next_rpc_delay>, WCG is using 4 days, some projects as little as 1 hour, while many doesn't use this at all. 9; Not so common, but if uses account manager, there are instances connection to all BOINC-projects are triggered. As long as you manages to finish a task before it's deadline, BOINC will report it before it's deadline, if connection isn't blocked for any reason. Like everything involved in WCG you need to have patience and you need to go gently whenever you change settings. Massive changes as suggested above are not the solution. I do not know everything about managing this project but I do have 62 days of Beta WU, so I have had some success. The biggest problem with setting a large cache is, especially now with a long string of short beta-wu's, the <duration_correction_factor> can be much too low, and this can lead to much too much non-beta-work downloaded with a large cache-setting. On one of my computers the DCF was roughly 1/10 of the correct, meaning with a 10-day cache-setting you'll risk getting close to 100 days downloaded...  "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
rilian
Veteran Cruncher Ukraine - we rule! Joined: Jun 17, 2007 Post Count: 1453 Status: Offline Project Badges: 
|
Now: how to force Beta Wu's: The moment you notice that there are Beta's available, change your buffer from 0.01 days to any higher number like 6, 8, 10 days. Your computer will automatically scoop up work for that longer period and so scoop up Beta's too. Right after the download, put your buffer back to 0.01 days and wait again... When the work you have is finished, do the same thing again. I thought you get only one Beta wu per cpu core  |
||
|
|
TBirdTheYuri
Advanced Cruncher France Joined: Mar 5, 2006 Post Count: 115 Status: Offline Project Badges:
|
Non, on peut recevoir autant d'unités Beta que les réglages le permettent, tout comme pour les autres projets. J'ai déjà reçu des lots de 15 ou 20 unités beta, surtout récemment avec les beta CMD qui durent parfois que quelques secondes.
Simplement, les unités Beta c'est comme les unités AC@H, c'est donné au compte-gouttes. --- No, you can receive as many Beta units as settings permit, as well as for other projects. I have already received packages of 15 or 20 beta units, especially recently with the beta CMD that sometimes last only a few seconds. Simply put,Beta units it's like Units AC@H, it is given to a trickle. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
2. When you decrease your buffer back to .1 days, you prevent your computer from asking for more work. This prevents your system from acknowledging the completed WU and may cause your WU to be marked as Too Late. Highly unlikely. More importantly recently is that if your system contacts the scheduler and a beta in your queue has already been completed by a quorum and validated, that beta will be aborted for you. If your system doesn't contact the scheduler at all until betas are completed, there's no risk of them being told to abort. Setting buffer back to prevent communicating with the scheduler is good. For true paranoia, block WCG in your firewall so that you can't possibly receive abort messages.  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
....or do what I do and set just Clean Air as your project.....you get lotsof Beta and if you start them running, using suspend and resume then they do not get aborted, they only abort if you have not started them.....
|
||
|
|
Steve WCG
Senior Cruncher Joined: May 4, 2009 Post Count: 216 Status: Offline |
These last two methods force WCG to process unnecessary results. Special irony goes to including the use of CEP in the suspend/ resume approach so you can waste power to get an electonic badge.
---------------------------------------- [Edit 1 times, last edit by Steve WCG at May 24, 2009 9:13:19 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
There are no CEP WUs to suspend/resume, that is why I gat all Betas so your comment about wasting power is not appreciated!!!!!!!
|
||
|
|
|