[RC5] DNETC is not backing down with CST needs processing power

Philippe Faure philippe at faure.ca
Thu Feb 24 10:25:01 EST 2011


I don't know if CST is a .net software or not.

As a solution, I have only allowed DNETC to attach itself to 1/2 the  
total number of processors in the system.  This seems to allow for CST  
to proceed without any issue.

I haven't been able to get the "pause when running" option to run for  
the time being since I don't have access to the processes from other  
users.  I haven't tried to start dnetc as a service.  I will give that  
a try.

Thanks.

Philippe




Quoting "Kevin G. McCoy" <kgmccoy at idk-inc.com>:

> On 2/23/2011 5:59 PM, Philippe Faure wrote:
>> Hello,
>>
>> I have a large server (24 logical CPU's) that can run DNETC, but for
>> some reason when an application is started by someone else, DNETC
>> doesn't recognize this, and doesn't relinquish the CPU processing time.
>>
>> The software is called CST (http://www.cst.com/).  I was wondering if
>> someone else has come across this issue?
>>
>> Since the user running CST and the DNETC user are not the same, I
>> can't check from processes manually.  This is happening on an Windows
>> Server 2008 R2.
>>
>> For now, I have backed off the user of all CPU, but this is a
>> sub-optimal solution.  If someone else has come across this issue, I
>> would like to know how they solved it.
>
> I used to work for HP/EEsof, so I am familiar with microwave RF design
> and circuitry simulation software.
>
> Is CST a .NET product?
>
> I have noticed that .NET apps hang-crash-act weird-run slow when run on
> the same machine as DNETC.  I suspect that .NET memory garbage
> collection runs at idle priority and since DNETC is always on, garbage
> collection never occurs. Eventually the .NET app dies.
>
> You can sometimes play games with processor affinity to force the .NET
> app to use certain core(s) or CPU(s) and force DNETC to use different
> core(s)/CPU(s).  That way everybody plays nice on the same hardware.
> This is not an optimal solution for DNET, but sometimes it is the only
> way to get it to work.
>
> You may have troubles with CST if it uses all the cores or CPUs during
> heavy simulation calculations.
>
> Best regards,
>
> Kevin G. McCoy
> _______________________________________________
> rc5 mailing list
> rc5 at lists.distributed.net
> http://lists.distributed.net/mailman/listinfo/rc5
>





More information about the rc5 mailing list