| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
| World Community Grid Forums
|
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 16
|
|
| Author |
|
|
rendition54
Master Cruncher USA Joined: Aug 16, 2005 Post Count: 2609 Status: Offline Project Badges:
|
As long as I eventually get credit for the arp WUs I completed, I have no complaints with the current process.
----------------------------------------![]() |
||
|
|
Acibant
Advanced Cruncher USA Joined: Apr 15, 2020 Post Count: 126 Status: Offline Project Badges:
|
Keep in mind that my goal was to provide context. To make it clear, my personal thought is that the platform (BOINC) already has a system that the tech, knreed, is adjusting to keep work units from languishing too long. That is, the definition of a "reliable" machine which knreed already adjusted down from one that returns within 2.5 days to one that returns within 2 days.
----------------------------------------So in the worst-case scenario where people are hogging a load of work units on their machines and not returning in a timely fashion, once those all time out they will get resent to reliable machines and be likely to get returned within that 36 hours proposed in the OP, especially if the definition was cut to 1.5 days. If that were to continue without further action that would mean you'd be adding a week for that initial timeout to occur for each generation which would be bad for getting ARP done sooner. However, knreed is also prioritizing older generations even for newly-generated work units if they're falling too far behind the head of the pack. It is stated that they try to have 20% of work units sent out going to reliable machines. With aggressive tweaking of the aforementioned settings they could get that to, say, 80%. But then you would see a backlog appear again as slower machines would not qualify for any eligible work units. So I don't think they will have much interest in modifying points based on return time (if they even could) because they have tools already and have gone as aggressive as they desire at this point. ![]() |
||
|
|
yoerik
Senior Cruncher Canada Joined: Mar 24, 2020 Post Count: 413 Status: Offline Project Badges:
|
should've clarified that the only part of my post directed at you specifically was just the thanks for the additional context - thanks for this additional context as well.
----------------------------------------![]() |
||
|
|
Martin Schnellinger
Advanced Cruncher Joined: Apr 29, 2007 Post Count: 128 Status: Offline Project Badges:
|
Hello Fellow Crunchers,
I got two points: 1. I very much doubt, whether the changes proposed would really speed up research, as the overal amount of operations needed to complete a workunit and a project does not change. There maybe less resends, but the number of calculations stays needed to complete a projects stays the same. 2. I fear, that owners of slow machines and cruchers just starting to contribute ("newbees") will be frustrated and bewildered, if they are "punished" for not contributing enough work power in a certain amount of time. Personally, I would not request this features, but others, more powerful ones.....GPU...old topic....I know. Sorry about it, but I had to say it. I will keep crunching 24/7, even if it is hot in my room. Best whishes Martin |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Lots of scary thought here above. Credit does nothing on these heavies, can assure you even if given 10 fold I'd not tolerate the systeem bog down at checkpoint save and uploading to allow more than 1 at the time computi... . I'd like to be able to use my computer without impairment. Via profile I allowed 2 in buffer and via app_config and Total buffer size controle 1 is proccssed, 1 is waiting and always the oldest and first to start when the previous finishes. Upload and download of ARP same time is killing so that wont start until the reports handshake is made with my setup. Just an upload take upwards of 20 minutes, which is the second reason to keep 1 in wait.
Anyway by this setup no single result goes back later than 48 hours. Begs the questione with this project report of achieving a cycle completion in 4.4 days how I had 8 in PV jail on thursday. That's 8 consecutive days of result return given my computer crunches 1 on average in 24 hours. .If 48 hours tops return puts my machine in reliable than strictly from soonest credit. msybe i should increase the buffer to 4 or more as it anyway takes 8+ days to get credited. Then the wigman gets to wait on me instead if doing quckies. Selfish me. |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7844 Status: Offline Project Badges:
|
I complete about 1 per day on my 3770 running Win 7. I do not recall having to wait for any significant amount of time to get another one. For me the supply has been quite dependable. I am running 7 MCM with 1 ARP on a 8 thread machine. I don't see any impact on normal use of the machine. I only have two in pending validation.
----------------------------------------Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
|