Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Fri, 15 Oct 2010 09:33:45 -0500

Hi Jeff,

I wonder if this is more likely to be a channel connection issue.  Mark hasn't 
reported how busy his R3.14.9 IOC is or what architecture it's running on, but 
if the UDP processing task doesn't get to run very often that would slow down 
the caput operations without affecting throughput as such.

- Andrew

On Thursday 14 October 2010 17:20:19 Jeff Hill wrote:
>
> I ran some quick tests of R3.14.11 on my pc.
> I see a maximum throughput closer to a million puts
> per second when doing routine tests of the ca client library
> against the ca example server (excas), but this is not connecting
> and disconnecting the channel with each put. I can also
> connect around 50000 channels per second. This is probably
> with a debug (no optimization) build.
>
> It takes about 8 seconds to run a script that does 100 caput
> command line requests against the example server on my workstation.
>
> Jeff
> ______________________________________________________
> Jeffrey O. Hill           Email        [email protected]
> LANL MS H820              Voice        505 665 1831
> Los Alamos NM 87545 USA   FAX          505 665 5107
>
> Message content: TSPA
>
> > -----Original Message-----
> > From: [email protected] [mailto:tech-talk-
> > [email protected]] On Behalf Of Mark Boots
> > Sent: Thursday, October 14, 2010 2:20 PM
> > To: [email protected]
> > Subject: slow caput from R3.14.11 client to R3.14.9 IOC
> >
> > Hi all,
> >
> > We've noticed at the CLS that channel access puts (caput) from a
> > default R3.14.11 build (linux-x86) to a 14.9 IOC are very slow.  As a
> > test, we ran a shell script with ~100 caput (command line) calls; it
> > took ~ 24 seconds to complete using the 14.11 version of caput, and
> > only 6s for the 14.9 version.
> >
> > Asynchronous caputs (using the channel access library, instead of the
> > command line) seem to be even slower -- sometimes taking over a minute
> > for the IOC to respond to all 100 puts.
> >
> > We're still debugging to see if this is a peculiarity of the IOC.  I'm
> > just wondering if anyone has seen similar issues involving channel
> > access between different EPICS versions?  Any suggestions for tools or
> > methods to help diagnose the problem?

-- 
If a man is offered a fact which goes against his instincts, he will
scrutinize it closely, and unless the evidence is overwhelming, he will
refuse to believe it.  If, on the other hand, he is offered something
which affords a reason for acting in accordance to his instincts, he
will accept it even on the slightest evidence.  -- Bertrand Russell

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System