Message boards :
Number crunching :
Petition to control how the scheduler works.
Message board moderation
Author | Message |
---|---|
ksnash Send message Joined: 28 Nov 99 Posts: 402 Credit: 528,725 RAC: 0 |
Please add a reply if: 1. You would like to be able to set scheduler based on your setup.( connection rate if you use a modem/ or real deadline if you have dedicated line) 2. You want to control if you want to use debt scheduling. If forced to run, make it so that project can't run anymore. |
Astro Send message Joined: 16 Apr 02 Posts: 8026 Credit: 600,015 RAC: 0 |
I agree the scheduler has problems, but what you describe is the older versions of Boinc (older than 4.35). just run one of those and you get what you want. tony |
AnRM Send message Joined: 29 Feb 04 Posts: 3 Credit: 300,282 RAC: 0 |
I am very relieved that the developers have the wisdom to allow a range of BOINC versions to coexsist ie. 4.19/4.25 for those who like to be 'hands on' and believe in the KISS principle and the more automated newer versions for those who are more 'plug and play'. MTCW. Thanks! |
ksnash Send message Joined: 28 Nov 99 Posts: 402 Credit: 528,725 RAC: 0 |
I am very relieved that the developers have the wisdom to allow a range of BOINC versions to coexsist ie. 4.19/4.25 for those who like to be 'hands on' and believe in the KISS principle and the more automated newer versions for those who are more 'plug and play'. Thanks! The older boincs didn't have intelligent scheduler, well nether did this one. They just followed the resource share blindly. I would be happier if the scheduler would show some intelligence. Maybe something like hey this work unit isn't in danger of missing deadline. So how are we supposed to decide which version to run. The earlier versions, listening to the threads, have some big bugs. Who ever decided to have a single variable to control cache size and connection rate started the biggest problem. It would be nice if the people developing the new software could go back and read some archived classic message board threads. They might see how real people voluteered thier computer resources to the project. a year ago They determined the infrastructure couldn't take everyone contacting Seti servers after each work unit was fininished and down load a new one. The people who ran Setiathome said everyone needs to look into using Setidriver or connect to Setiqueue servers so main servers wouldn't get overloaded. Boinc was supposed to have its own cache so the third party programs weren't needed anymore. So what does the new software do now, make it so you can't use the only cache setting or you get ran out of a project for a long time. I keep hearing wait wait it will even out. It hasn't yet. My so called debt has increased a magnitude of 10 in a week. I am asking for simple changes and the responses on the board from real users echoes the same thing. count how many times someone posts about one project ran dry and won't download. |
Nightbird Send message Joined: 2 Feb 03 Posts: 73 Credit: 53,523 RAC: 0 |
I would like to know how disabling this "new" scheduler. I tried the 4.37, the 4.43 and the 4.44. Before the machines ran without problem with the 4.19. With this new 4.4x, i needeed to do something like "babysitting" in order to understand how it worked and to control it. Finally, i removed it and i installed the 4.27. I don't need an "intelligent" scheduler, I prefer something that i can control. Is it difficult to understand ? Do you want to get banned for 31 years and your account & credits deleted at a Boinc project ? Predictor@home is your best choice. |
MikeSW17 Send message Joined: 3 Apr 99 Posts: 1603 Credit: 2,700,523 RAC: 0 |
Frankly, I have given up worrying about the scheduler. No matter how cleverly they tweak just the code, it will never be right. If you feed any algorithm, no matter how good, duff and inaccurate data, it will produce odd results. While the scheduler is fed a queue of processing requests, whose run-time is inaccurately known, it can never properly do its job. Until someone comes up with the simple concept of a host storing the real run-times of work units actually done on that host, neither the 'days between connection' nor the scheduling will work as expected. Once a host 'knows' how long an average work unit will take, the the first thing that will work correctly is the amount of work downloaded. The knock-on effect of that will be that panic mode will hardly need to trigger at all. And then, hey-presto no more complaints about project shares. If you guess you have 20 seats on a bus and you sell 20 tickets, then it turns-out it is only a 12 seater, no matter how clever you order the people getting on the bus, eight people get left behind. |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
If you guess you have 20 seats on a bus and you sell 20 tickets, then it turns-out it is only a 12 seater, no matter how clever you order the people getting on the bus, eight people get left behind. That is why there are racks on the top of the bus ... just tie them in place and off you go ... |
Nightbird Send message Joined: 2 Feb 03 Posts: 73 Credit: 53,523 RAC: 0 |
Frankly, I have given up worrying about the scheduler. This is not the question to give up or to worry up. Perhaps some or many people like it because it's easier (?)(but how ?) but i had no problems with a 4.19 on machines running 3 or 4 projects. Now how this "new" scheduler works is not satisfactory for me. Do you want to get banned for 31 years and your account & credits deleted at a Boinc project ? Predictor@home is your best choice. |
W-K 666 Send message Joined: 18 May 99 Posts: 19063 Credit: 40,757,560 RAC: 67 |
I have to agree with MikeSW17 the scheduler is not going to work correctly unless the processing time can be reasonably accurately predicted. You cannot have a connection time of X days, because X days work is completed in 0.n * X days on most computers. To overcome communication problems most of us, I believe, would like to keep a few days of work in our cache, but connect more frequently. Predicting the completion time might also reduce the queries about other gripes. Andy P.S. Does anybody know how many people are using V4.4x? Andy |
Speedy67 & Friends Send message Joined: 14 Jul 99 Posts: 335 Credit: 1,178,138 RAC: 0 |
|
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.