| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Community
Forum: Chat Room
Thread: What is the purpose of the wingman ? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 14
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Things DO come up all the time, they are pc's after all, so don't get the idea that it is THAT easy to run a project. But there are computer nerds running small time projects everwhere just because Boinc is set up to be a 'run on its own' system. And don't get the idea that it is THAT difficult, either: Do not split the sum of the points equally between faster-done WUs and slower-done WUs. That is, instead of 50%-50% split, do a, say, 60%-40% split in favor of the faster-done WU. For finer gradations, calculate the relative weights from a comparison of the completionTimes of the WUs involved. Or, give the 50% to the slower-done WU, and 50% + X to the faster-done WU where "X" is a factor where other considerations are taken into account. If that is not easy, I have no idea what you are trying to talk about, including your reference to a 'run-on its own' system which has nothing to do with evaluating a method. ; Edit1:2011.10.25Tu.2033: spell-check [Edit 1 times, last edit by Former Member at Oct 25, 2011 8:32:26 PM] |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: 
|
And don't get the idea that it is THAT difficult, either: Do not split the sum of the points equally between faster-done WUs and slower-done WUs. That is, instead of 50%-50% split, do a, say, 60%-40% split in favor of the faster-done WU. Hmm, it's not very clear to me that you're meaning here, if by "faster-done WU" you mean something like: Wu-#1 on computer A takes 5 hours to crunch. Wu-#2 on computer A takes 2.5 hours to crunch. If this is your meaning, the fair thing to do is to give example 100 points for Wu-#1 and 50 points to Wu-#2. Your suggestion of giving example 60 points to Wu-#2 "because it's done faster" doesn't make much sence to me... Another meaning of "faster-done WU" can be something like this: Task-#1 for Wu-#1 on computer A takes 5 hours to crunch. Task-#2 for Wu-#1 on computer B takes 2.5 hours to crunch. If this is your meaning, the fair thing to do is in my opinion to give example 100 points for Wu-#1, regardless of this being done on computer A or computer B, assuming both Task-#1 and Task-#2 reached 100% done that is. Assuming both computers runs 24/7, and all wu's is equal, computer A will get 480 points/day while computer B will get 960 points/day. Meaning, since computer B finish each task in half the time computer A uses, computer B will get 2x the points/day. Your suggestion of giving 120 points to computer B and 80 points to computer A for Wu-#1 will mean 384 points/day for computer A and 1152 points/day for computer B, meaning computer B gets 3x more points/day than computer A even B is only 2x faster than computer A. Now, such a system can be an advantage in some instances, example if Wu-#2 can't be generated before Wu-#1 is validated, giving an extra incentive for fast turnaround-time can lead to more fast computers crunching this project, and therefore more wu's crunched per day for this project. A variant to this is if a project has limited server-resources so can only have N active wu's at a given time, the more wu's that's tied-up waiting for 2nd. task needed for validation, the less wu/day can be done. Another possibility is if computers wu-demands exceeds how many wu/day a project can manage to generate. But, if on the other hand a project has example 1000 million independent wu's that needs to be crunched, and the current demand is 100 k wu/day, giving an extra incentive for fast turnaround-time doesn't make much sense. Yes, you will likely get some users with very fast computers to run this project, something that let's say increases demand with 20 k wu/day. But, on the flip-side you're likely to alienate users with older computers that doesn't have any reason (or money) to upgrade, and let's say your "bonus"-system leads to demand decreasing with 30 k wu/day, meaning total effect is a decrease to only 90 k wu/day done... A user first alienated is unlikely to come back, even if you've removed the bonus-system. Also, if you removes the bonus-system again, chances are some of the new computer-power you've attracted will leave again, so you'll maybe drop to 75 k wu/day or something...  "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
mikey
Veteran Cruncher Joined: May 10, 2009 Post Count: 826 Status: Offline Project Badges: 
|
But, if on the other hand a project has example 1000 million independent wu's that needs to be crunched, and the current demand is 100 k wu/day, giving an extra incentive for fast turnaround-time doesn't make much sense. Yes, you will likely get some users with very fast computers to run this project, something that let's say increases demand with 20 k wu/day. But, on the flip-side you're likely to alienate users with older computers that doesn't have any reason (or money) to upgrade, and let's say your "bonus"-system leads to demand decreasing with 30 k wu/day, meaning total effect is a decrease to only 90 k wu/day done... A user first alienated is unlikely to come back, even if you've removed the bonus-system. Also, if you removes the bonus-system again, chances are some of the new computer-power you've attracted will leave again, so you'll maybe drop to 75 k wu/day or something... But increasing the wu's per day also means an increase on the Servers to compile the data coming in at a now faster rate, which they may not be able to do. And even if they can someone has to look at the data manually to see if it is really what is wanted and a project may not have the extra people to do that either. Some projects are maxed out in their resources and are just happy to be running. Seti, the original Boinc Project, was/is shutting down several days per week just to keep up with the data flow, among other things.   |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
WCG has a policy of sizing work units [if it can] to a mean average run time, so almost all can participate [with the appropriate hardware], whilst not jeopardizing the stability of the work scheduler. Many will remember how a second validator had to be put in place once the 4x faster HCC jobs started to come through in large number and the PV jail was building out of it's seams.
------------------------------------------//-- [Edit 1 times, last edit by Former Member at Oct 26, 2011 2:33:24 PM] |
||
|
|
|