EPICS Home

Experimental Physics and Industrial Control System


 
1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  <20052006  2007  2008  2009  2010  2011  2012  2013  2014  2015  2016  2017  2018  2019  2020  2021  2022  2023  2024  Index 1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  <20052006  2007  2008  2009  2010  2011  2012  2013  2014  2015  2016  2017  2018  2019  2020  2021  2022  2023  2024 
<== Date ==> <== Thread ==>

Subject: RE: device support for SNMP + Network statistics + vxStats
From: Al-Adwan Ahed <[email protected]>
To: "'Matthias Clausen'" <[email protected]>, Andrew Johnson <[email protected]>
Cc: Emmanuel Mayssat <[email protected]>, Janet Anderson <[email protected]>, Vladimir Sirotenko <[email protected]>, [email protected], Albert Kagarmanov <[email protected]>
Date: Fri, 8 Jul 2005 11:24:30 +0200
> Matthias Clausen wrote:

> We are currently also working on devSNMP scanning the MIBs from our SUN 
> and Linux machines. Scanning the (Cisco) network switches is the next 
> step. Albert Kagarmanov is working on this on our side. I wonder whether 
> we are talking the same devSNMP - hope so! - Let's check next week when 
> I am at ANL...

> SNMP-Server (not the right term but best to understand the 
> functionality) for VxWorks IOC's would be really useful. The current 
> implementations are commercial and really expensive. Is Emmanuel 
> volunteering to implement one? This way we could get symmetric 
> information from the CA-client and - server machines. Which includes 
> network load etc. besides the well known CPU/Mem/FileDesc... we all get 
> via channel access from the EPICS-IOC's.


At the SLS, we have added some information about the Network Interface
statistics to vxStats (we call it now IOCMON not to confuse it with many
different versions of vxStats), here is the MEDM snapshot:

http://people.web.psi.ch/adwan/img/iocmon.gif

We just have installed it in all ioc's and we are in the phase of running
tests, we still have to work with the networking people to define better
test scenarios (suggestions are welcome, like, how to drain the ioc's
network stack data buffers?).

For example, we started to see output errors reported by the ioc's NI (EPICS
3.13.9, MVME2300, vxWorks 5.3.1) once the Switch is configured to Full
Duplex. We assume if there is a miss configured device that is spamming
packets, ioc's in that subnet will start reporting collisions, the same
should be true when the Load goes very high.

This addition should serve the following purposes:
1- Detects problem in our network related to wrong equipment configuration
and the like.
2- Optimize our network stack configuration parameters (like buffer numbers
of different sizes) to fit our applications needs.
3- Anticipate future network operation needs.

We are really still in the test phase, and once we have a stable version and
more results, I will inform you.

Best regards,
Ahed

##########################################
# Ahed Aladwan                            
# SLS Controls / Paul Scherrer Institute  
# WSLA/208, 5232 Villigen PSI Switzerland 
# Tel:+41 56 310 4594                     
# Fax:+41 56 310 4413                     
# www.psi.ch                              
# http://people.web.psi.ch/adwan/


ì


Navigate by Date:
Prev: Re: device support for SNMP Matthias Clausen
Next: CAJ 1.0.3 release Matej Sekoranja
Index: 1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  <20052006  2007  2008  2009  2010  2011  2012  2013  2014  2015  2016  2017  2018  2019  2020  2021  2022  2023  2024 
Navigate by Thread:
Prev: EPICS <--> PS - Survey Martin Pieck
Next: CAJ 1.0.3 release Matej Sekoranja
Index: 1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  <20052006  2007  2008  2009  2010  2011  2012  2013  2014  2015  2016  2017  2018  2019  2020  2021  2022  2023  2024