Message boards :
Number crunching :
Cache on fast or optimsed clients
Message board moderation
Author | Message |
---|---|
Dave Harry Send message Joined: 28 Feb 03 Posts: 42 Credit: 8,021,290 RAC: 22 |
I would like to be able to keep a cache of around 7 days. With the cache setting on 10 days, my BOINC (now using 4.43) downloads WUs based on the presumption that they will take around 23,000 seconds on this computer. However, since they only typically take around 7500 seconds (on 4.11) to complete on an optimised client (or about 10,000 on the old 4.09 client), the cache is emptied signifiacntly faster than BOINC caclulates. So a '10 day' cache lasts around 2-3 days. I'm not au fait with the way BOINC (or seti?) calculates the work time, but I would assume it relates in some way to the general benchmark test. So why is it so far out, even for the 4.09 client? With a 14 day deadline, I could set a cache to a month and still not miss a deadline. |
mikey Send message Joined: 17 Dec 99 Posts: 4215 Credit: 3,474,603 RAC: 0 |
I would like to be able to keep a cache of around 7 days. UMMM..1st thing, the max you can set the cache for is 10 days. 2nd...the schedular is sort of intuitive, it will see that your settings are for more days than it is providing and over time will give you more units. There IS a hard limit of 100 units per machine per day, so you will never get more than that. Although that would be a VERY fast machine!!! |
Angus Send message Joined: 26 May 99 Posts: 459 Credit: 91,013 RAC: 0 |
I'll disagree that the scheduler is intuitive. It is making no attempt to keep my queue filled to the REAL requested level. I'm running SETI only, with a 10 day cache requested. On this PC, an average WU runs between 5000 and 7000 seconds, or 1.6 hours. My cache is hovering at 44 WUs, each supposedly taking 6:07 to run, to make up the 240 hours of work requested. Unfortunately, the effective queue is only 2.9 days (44 x 1.6 hours). There is no way to keep the queue filled to the requested level. Going into a long weekend, this is a scary situation. One screw-up by Berkeley and this PC will run dry. And NO! I will NOT attach to another project. I have no interest in any of the other BIONIC projects. |
jimmyhua Send message Joined: 16 Apr 05 Posts: 97 Credit: 369,588 RAC: 0 |
Which version of BOINC is adaptive? I am running version 4.25 and I have my time to connect at 3 days. It has consistently cached up 2 days worth of work since day 1. Jimmy |
Angus Send message Joined: 26 May 99 Posts: 459 Credit: 91,013 RAC: 0 |
How does BIONIC Manager determine the 'Time to Completion" for new WUs? It can't be a constant, because it varies slightly. My 6:07 has changed in the last couple of hours to 6:02. Does it depend on the (broken) benchmark scores? |
Angus Send message Joined: 26 May 99 Posts: 459 Credit: 91,013 RAC: 0 |
BUMP No takers? Can anyone explain how to keep a full cache? |
[ue] Toni_V Send message Joined: 6 Apr 03 Posts: 52 Credit: 141,788 RAC: 0 |
As far as I know the Boinc estimates calculation time from the benchmark results. And the benchmark changes a little, that's why the estimated time changes too. For me the estimated time is 5:07-5:30 usually, and calculation time with optimised client 1:17. So basically my options say I have 5 days of cache, but it's more like one day in reality. I guess there's nothing to be done exept a optimised compilation of the benchmarkin routine. I think someone has been making this - not sure about the current situtation. I'm guessing what you can do is now just raise the cache... |
Dave Harry Send message Joined: 28 Feb 03 Posts: 42 Credit: 8,021,290 RAC: 22 |
I'm guessing what you can do is now just raise the cache... I've raised as far as she'll go! I canna raise no more! |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
Can anyone explain how to keep a full cache? Right now, this is a BOINC Version dependent question. But, simply enough, set the connect number for the number of days you desire. Then once a day, or every other day, do a manual update to flush the completed work. |
Angus Send message Joined: 26 May 99 Posts: 459 Credit: 91,013 RAC: 0 |
Can anyone explain how to keep a full cache? The point is , you ask for 10 days, and get 3, because the scheduler (or Manager, don't know which determines this) can't figure out how long it takes your PC to run jobs. That's not a full cache. If you ask for 10 days, you should get 10 days of work, and be able to shut off the network connection and keep crunching for 10 days, or keep crunching if the project runs out of work or goes offline for an extended period. |
shady Send message Joined: 2 Feb 03 Posts: 40 Credit: 2,640,527 RAC: 0 |
Can anyone explain how to keep a full cache? I agree with you and it is a big problem , its not very helpfull when you are running a large farm and only have 3 venue setting options and the accuracy of estimated crunch times varies from machine to machine and project to project. Shady Shady <img src='http://www.boincsynergy.com/images/stats/comb-1527.jpg'> |
©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.