Message boards : Number crunching : Flag to opt-in to run big (200+MB memory) WUs?
Author | Message |
---|---|
Dimitris Hatzopoulos Send message Joined: 5 Jan 06 Posts: 336 Credit: 80,939 RAC: 0 |
I just noticed a pretty big (many amino-acids) protein, named 7521_1_fullatom_relax_... which has a "working set" (i.e. memory requirement) of 200MB and a virtual size of 646MB. I also understand that more big WUs will come in the future, as part of the vaccine and cancer studies within Rosetta@home I wonder if it would be possible for BOINC to either: 1/ have BOINC feeder only send bigger WUs to hosts with e.g. >768MB or 2/ let users opt-in to run bigger WUs using a project-specific BOINC setting, like the "run-WU-for-X-hours" Best UFO Resources Wikipedia R@h How-To: Join Distributed Computing projects that benefit humanity |
Keck_Komputers Send message Joined: 17 Sep 05 Posts: 211 Credit: 4,246,150 RAC: 0 |
Option one is possible by customising the memory requirement as the tasks are created. No one has gotten option 2 working yet even though there are a couple of projects that could use it. BOINC WIKI BOINCing since 2002/12/8 |
Dimitris Hatzopoulos Send message Joined: 5 Jan 06 Posts: 336 Credit: 80,939 RAC: 0 |
I'd like to re-assert the issue of "opting-in" to run bigger (>100MB memory) WUs for Rosetta. I don't think the impact on donor userbase should be understimated, i.e. some people will stop crunching (see misc unit err thread). Rom wrote Most BOINC projects (except CPDN) require 1/10th or even less memory than Rosetta currently does: most Rosetta WU's need 70-120MB and bigger ones 250MB (and 650MB v.size) compared to e.g. Einstein 12-15MB, SIMAP 5-7MB etc This BigWU setting would ideally be tunable PER HOST, otherwise per separate home/work/school preferences. IMHO the best method is how Folding@home does it:
source: F@h settings Best UFO Resources Wikipedia R@h How-To: Join Distributed Computing projects that benefit humanity |
River~~ Send message Joined: 15 Dec 05 Posts: 761 Credit: 285,578 RAC: 0 |
This is actually a very different request to the recent one about the length of WU. At least, it looks similar to the user, but is very different in terms of the implementation. The recent mod was to the application code, and affects how the application decides whether it is finished. As such it was totally within the Rosetta team's domain to make this change, and I am glad they did. In contrast, other opt in flags would need to affect which work was assigned to the computer before it even was issued. This would need mods to the BOINC code in the server as well as mods to the preferences database. As well as the flag suggested here, I can think of at least four other flags that might be useful, and if any of them is implemented it seems to me to make sense for the BOINC folks to provide the hooks for all of them at the same time - all of them have the same basic idea - that the user sets a flag, and the project team give the scheduler enough info to know which WU are allowed and which excluded by that flag. I have started a separate thread to discuss whether we feel there is enough demand for such changes to be worth asking the BOINC people to do this, see here |
Message boards :
Number crunching :
Flag to opt-in to run big (200+MB memory) WUs?
©2024 University of Washington
https://www.bakerlab.org