Experimental Physics and
Industrial Control System

"Zelazny, Michael S." <[email protected]> · Fri, 15 Nov 2002 14:19:24 -0800

Let's see if I can remember exactly why we wanted to do this.

1.) We've frequently seen failure modes when writing files from the IOC causing corrupt files.  
2.) People have an annoying habit of powering off IOCs just as the files were being written.
3.) We wanted to off load some of our older IOCs.  File writes are very expensive operations.
4.) We have a need to save and restore waveforms.
5.) Much easier to add/remove channels.  Operations doesn't like us to touch IOCs during operations.  UNIX process restart is OK.
6.) Network traffic is negligible because we use CW for channels that don't change very often.
7.) We wanted the ability to log channel changes.
8.) When an IOC has a problem loading databases the IOC s/r process kills the files by writing "Search Issued" to channels that don't connect.  CW reads in the previous saved files for default values.

Maybe Stephanie Allison and Ron Chestnut can think of other reasons.

Here's a link to all of our previous EPICS collaboration meeting presentations, including the one I'm giving next week at Jefferson Labs: http://www.slac.stanford.edu/comp/unix/package/epics/extensions/ChannelWatcher

Here's a link to the documentation that describes all of the features (implemented and planned): http://www.slac.stanford.edu/comp/unix/package/epics/extensions/ChannelWatcher/ChannelWatcher.html

....and yes, CW does support better precision for floats and doubles.

Mike

-----Original Message-----
From: Pete R. Jemian [mailto:[email protected]]
Sent: Friday, November 15, 2002 1:46 PM
To: Zelazny, Michael S.; EPICS Tech-Talk
Subject: Re: Channel Watcher V2.0

>
>
>The Channel Watcher moves the save part of save/restore from the IOC to UNIX.  
>
Why is this an improvement over save/restore?
Doesn't it add to network traffic?
Don't get me wrong, I have problems with the save/restore process 
failing frequently.
I'm just curious about the motivations here in case they may
improve my systems by adding more reliability.

>CW V2.0 supports the caGet/caPut file formats,  
>
Hopefully, CW supports floating point numbers better than does caGet
which, say for a number 1.2e-9, reports that number as 0.000.
No way to tell caGet to use scientific notation instead.  Bummer.

Pete Jemian
UNICAT

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System