[Info-vax] Networking Benchmarks

[Kerry Main]

> -----Original Message-----
> From: Info-vax <info-vax-bounces at rbnsn.com> On Behalf Of johnwallace4--
> - via Info-vax
> Sent: September 19, 2019 3:37 AM
> To: info-vax at rbnsn.com
> Cc: johnwallace4 at yahoo.co.uk
> Subject: Re: [Info-vax] Networking Benchmarks
> 
> On Tuesday, 17 September 2019 17:40:47 UTC+1, Bob Gezelter  wrote:
> > On Saturday, September 14, 2019 at 1:47:29 PM UTC-4, Jeffrey H. Coffield
> wrote:
> > > On 09/13/2019 09:35 AM, Bill Gunshannon wrote:
> > > >
> > > > Has anyone ever done any benchmarking of TCPIP vs. DECNET under
> > > > VMS?  It seems that there should be less overhead and therefore
> > > > greater throughput using DECNET.
> > > >
> > >
> > > $ write sys$output f$cvtime()
> > > 2019-09-14 10:43:35.86
> > > $ copy/ftp t.tmp vms3"jeffrey blender"::t.tmp $ write sys$output
> > > f$cvtime()
> > > 2019-09-14 10:43:43.71
> > > $ copy t.tmp vms3::
> > > $ write sys$output f$cvtime()
> > > 2019-09-14 10:43:51.24
> > >
> > >
> > > FTP> put t.tmp
> > > 200 TYPE set to IMAGE.
> > > 200 PORT command successful.
> > > 150 Opening data connection for SYS$SYSDEVICE:[JEFFREY]t.tmp;
> > > (10.10.12.10,49163
> > > )
> > > 226 Transfer complete.
> > > local: SYS$SYSDEVICE:[JEFFREY]t.tmp;2  remote: t.tmp
> > > 116359168 bytes sent in 00:00:03.26 seconds (34.04 Mbytes/s)
> > >
> > > copy       7.85 seconds
> > > copy/ftp   7.53 seconds
> > > ftp	   3.26 seconds
> > >
> > > ymmv
> > > Jeff
> >
> > Jeffrey,
> >
> > With all due respect, may benchmarking be so simple.
> >
> > Since both ftp and fal remote elements can do connection-to-connection
> optimizations (e.g., process reuse), such simplistic tests tell little
about actual
> performance.
> >
> > There are also questions of file extent size and buffering. Both lead to
> dramatic performance differences for DECnet FAL, and depending upon
> implementation, the FTP server of a particular TCP package.
> >
> > Superficial benchmarks do not answer any useful question.
> >
> > - Bob Gezelter, http://www.rlgsc.com
> 
> Well said that man.
> 
> There's a reason people in the know used to call it "benchmarketing".
> 
> If people know what they want to measure and compare, by all means
> compare them, and understand the relevance of the results of the
> comparison. If there isn't a proper set of properly understood performace
> indicators, then the results are likely to be just hot air.

To that point, comparing the current native OpenVMS TCPIP stack which has
not been updated performance wise in the last 15+ years, was basically a
ported copy of the Tru64 UNIX stack which itself was also a ported copy of
an older UNIX stack and which has now been more or less retired by VSI to a
modern TCPIP stack is likely not very useful.

As someone mentioned, comparing a modern Windows/Linux TCPIP stack to the
latest versions of Multinet (which the new VSI stack is based on) would be a
tad more apples to apples comparison. 

Even then, I suspect most people will be more interested in how the OpenVMS
X86-64 V9.* TCPIP stack (when it is prod ready) will compare to an
equivalent Windows/Linux TCPIP stack  on the exact same X86-64 hardware.


Regards,

Kerry Main
Kerry dot main at starkgaming dot com