+1 for showing the total memory usage per node/machine. When I looked at the one in 3.14 I was thinking that it would be nice to know that information for all processes sharing the same memory.
On Fri, Feb 10, 2012 at 2:11 PM, Burlen Loring <[email protected]> wrote: > Great! let me know if there is anything I can do! More recently I had > revised the panel to make launching remote gdb sessions easier, and making > it easier to navigate through larger runs (256+ procs). > > I just noticed that the memory inspector panel in 3.14 is not reporting > per-process values which is what I was expecting. So I was wrong, there is > no bug in the units. Sorry for the false alarm. > > Burlen > > > On 02/10/2012 11:14 AM, Utkarsh Ayachit wrote: > >> Burlen, >> >> That looks pretty slick! If I had known earlier I would have started >> with your code :). Any case, what you have definitely sounds >> interesting. We should work on integrating the two in the future. >> >> Utkarsh >> >> On Fri, Feb 10, 2012 at 1:30 PM, Burlen Loring<[email protected]> wrote: >> >>> Hi Utkarsh, >>> >>> Those sound like wonderful additions especially the new transfer function >>> editor, I can't wait to start using 3.14! >>> >>> I want to share some experience re: the memory inspector as I had written >>> something similar as a source plugin called the Process Monitor a while >>> back >>> (screen shot attached). In the memory inspector panel it would be >>> useful to >>> organize the server ranks by host name and display the total memory used >>> on >>> each host by the pvservers. This sum is necessary to know how close you >>> are >>> to actually running out. >>> >>> It may also be useful to display memory usage as both absolute and % >>> used. >>> The latter would easily enable adding visual ques to the panel (eg. row >>> color is changed to red when more than 75% of memory is used on an >>> individual host) to warn the user that they are in danger of a crash. >>> >>> You may want to consider adding an "override" for reporting the amount of >>> free ram (and % used) when the server is run on an SMP such as NICS >>> Nautilus. Although Nautilus has 4TB of ram, users need to request a >>> specific >>> amount of ram per core when submitting a job. The kernel will kill the >>> process that uses more than the requested amount, so what the system >>> reports >>> as free memory is fairly useless. >>> >>> An old version of the Process Monitor is in the SciVisToolKit plugin, >>> note >>> that mac server side code is not finished so only windows and linux >>> servers >>> are currently are handled. The process monitor has a number of additional >>> debugging features such as the ability to attach gdb to remote servers, >>> and >>> enable backtrace on sig SEGV etc. >>> >>> You should also double check that the memory used value you report is >>> accurate. When I tested the 3.14 binary I noticed that it reports 4G used >>> when top reports only using 425m. There may be a bug in your unit >>> conversion. >>> >>> Hope this feedback is useful, I had intended to share this work but just >>> haven't had the time to finish the mac implementation. It's great that >>> you >>> have added it! >>> Burlen >>> >>> >>> > ______________________________**_________________ > Powered by www.kitware.com > > Visit other Kitware open-source projects at http://www.kitware.com/** > opensource/opensource.html<http://www.kitware.com/opensource/opensource.html> > > Please keep messages on-topic and check the ParaView Wiki at: > http://paraview.org/Wiki/**ParaView <http://paraview.org/Wiki/ParaView> > > Follow this link to subscribe/unsubscribe: > http://www.paraview.org/**mailman/listinfo/paraview<http://www.paraview.org/mailman/listinfo/paraview> >
_______________________________________________ Powered by www.kitware.com Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html Please keep messages on-topic and check the ParaView Wiki at: http://paraview.org/Wiki/ParaView Follow this link to subscribe/unsubscribe: http://www.paraview.org/mailman/listinfo/paraview
