| Info | Message |
|---|---|
| 1) Message boards : BOINC Manager : My Wish List - part 3.
Message 36605 Posted 26 Jan 2011 by Trev |
On the stats charts for both total & average credits for the host / user it would be really good to have a second Y axis showing the ‘run time’ used each day to generate the recorded credits for each project on the chart. (Perhaps this could be displayed using a dashed line of the same colour as the project’s credit results) I have a number of PC running Boinc for different amounts of time each day (5 hours, 16 hours, 24 hours) and although I can compare credits generated, I can’t easily work out the effort required to generate the credits. I want to be able to easily say that leaving on my new PC ‘A’ (2.5GHz 8 Core CPU + 1 GPU) for another 45 minutes each day is equivalent 3 hours of my older PC ‘B’ (3GHz 2 Core CPU). This may seem a little controversial because it encourages using Boinc as a race for high credit scores, which I do enjoy although 100K per day is still out of my league. This will assist the core Boinc supporters to contribute to their chosen projects, but doing so in the most efficient way possible whilst keeping their electricity bills down to a minimum. In my example above PC ’A’ probably only draws 25%-40% more power than PC ‘B’, but if I use PC ‘A’ more or instead of PC ‘B’ to generate the same amount of credits it costs me less in electricity. Using the same data it would be good to be able to compare Hosts on the website stats by their generated credits divided by the ‘run time’ used to generate it. |
| 2) Message boards : BOINC Manager : My Wish List - part 3.
Message 36508 Posted 21 Jan 2011 by Trev |
Hi Just remembered the other thing I was going to subject. A number of times I have noticed that when I have got a mix of tasks I get the following situation. Tasks with a long (+80 hours) time to complete and with a deadline 40-60 days away are running as high priority. And at the same time tasks with a short (0-3 hours) time to complete remaining and with a deadline 0-32 hours away but these are running normally or not running at all. I know that Boinc is running the long tasks at high priority otherwise they will never get completed before the deadline, but with my PC just running on average 5 hours a day, without my intervention it is unlikely that the short task would get finished before its dead line. Can I suggest that in addition to the current scheduling, that when a task gets within 24 hours of its dead line the scheduler’s priority switches to completing as many tasks as possible, starting with the smallest time to complete first. Also no new tasks are started while there are still tasks within this 24 hour window, overriding the switch between applications function. This may result in tasks exceeding their dead line without having been started, but that is better than having a number of tasks that exceeding their dead line but only had 0-3 hours left to run. I have suggested a 24 hour window because I am well aware that Boinc doesn’t know when the PC will be processing again, so that the scheduler needs to have time to switch priorities and get the task completed. A similar thing could be done for tasks that have (4-12 hours) time to complete remaining, but a window of 48 hours. Something that would be difficult to implement at the moment but is worth considering in the future as GPU processing becomes more main stream, is the ability to allocate project resources for GPU separately from CPU. The other minor request was to have a different row background colour for tasks that are currently running. With a 8 core CPU + GPU task switching and a buffering of tasks waiting to start you end up with quite a page full, even with the ‘Show active tasks’ filter on its not immediately noticeable which ones are running. Trevor |
| 3) Message boards : BOINC Manager : My Wish List - part 3.
Message 36507 Posted 21 Jan 2011 by Trev |
Hi, Another feature requests. On the task list of the Boinc manager can we have multi column sorting. I normally have the tasks sorted by ‘To completion’ time, but I would also like to be able to sort by report dead line as well. If 3 column sorting is possible that would be even better. |
| 4) Message boards : BOINC Manager : My Wish List - part 3.
Message 36460 Posted 16 Jan 2011 by Trev |
Hi All, here are a couple of features that I would definitely find useful. I have only been using Boinc for 4 months so sorry if these items have been suggested in the past. I only have my home PC on in the evenings which averages out at 5.25 hours a day, but I often getting work units down loaded that I am not likely to complete within the deadline. I know you can abort work units but I don’t have the manager open all the time, so I often find some that have been processed for 3 – 8 hours before I spot that I am not going to complete the other 54 hours before the dead line. I wish: There was a setting that you could set your expected average compute hours per day, and that Boinc would not download / or auto abort work items that are not going to be completed before the deadline. (Auto abort may be easier to implement) If Boinc already dose this or similar automaticity based on processing time in the last week etc, it doesn’t seem to be making a very good job of it and could we have a manual override for the hours per day value. I have access to a Server 2008 R2 system that is running Hyper-V during the day, but there is not much activity during the evening. At the moment Bonic uses the system CPU usage to detect when it should suspend processing, but on a Server 2008 R2 Hyper-V system this value does not include CPU usage of any virtual PCs that are running just the parent OS ‘Server 2008 R2’. The true system wide CPU usage is shown on one of the Perfmon performance counters for Hyper-v, I forget which one. I wish: Can Boinc auto detect or a manual check box be provided to allow it to monitor the Perfmon performance counter instead of the normal system CPU usage. Something ells that would be good but less important for me at the moment is to have two sets of the ‘Day-of-week override’ values each set having its own ‘use at most CPU time’ value. I would like to have been able to set some of my systems to 100% CPU for 17:30 – 07:30, and then a trickle processing of 10-30% from 07:30 – 17:30 to make sour that it doesn’t interfere with normal operations. This may also get around the problem with Server 2008 R2 system that is running Hyper-V I mentioned above. Thanks for listening. |
Copyright © 2026 University of California.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License,
Version 1.2 or any later version published by the Free Software Foundation.