[nos-bbs] trace files

Michael Fox - N6MEF n6mef at mefox.org
Wed Sep 22 19:34:28 EDT 2010

Interesting Maiko.

I would have assumed that JNOS decodes the packet and then hands it off  -
either to the display routine or to the OS file handler.  Either way, after
decoding the packet and handing it off, I would have assumed the impact on
JNOS would be the same and that it's then up to the display routine or the
OS to write the output.

We know the OS can log things very quickly.  The voluminous messages at boot
time are generated and logged in real time. (And that's on an ancient laptop
with 512MB of memory!)

With today's processors and disk speeds, in my day job, I know that large
volumes of debugging info can be written in real time with very little
degradation in performance.  TCPdump and other such applications can capture
lots of data at Mbps speed - much more stressful than 1200 bps.  So I wonder
if what you saw was simply due to a very underpowered machine.  Or I wonder
if it's a limitation of DOS.  And I wonder why writing to a file is slower
than writing to the screen (again, from the JNOS perspective - since the OS
will handle the I/O buffering).

I realize I could close the trace, but then it's stopped.  What I'm trying
to do is watch the traffic as it comes in.  The F9 session is useless for
that since after three packets or so, it scrolls of the screen and is lost.
And, right now, the log file is useless for this purpose, because there's
nothing in it until much, much later.  So, right now, I'm stuck.  I'm hoping
you can have a look at it in the near future.


-----Original Message-----
From: nos-bbs-bounces at tapr.org [mailto:nos-bbs-bounces at tapr.org] On Behalf
Of Maiko Langelaar
Sent: Wednesday, September 22, 2010 2:19 PM
To: TAPR xNOS Mailing List
Subject: Re: [nos-bbs] trace files

> how long/much does it buffer [snip] what possible purpose ?

If one were to FLUSH every single call to the trace file or socket,
you might see a fairly large degradation in JNOS performance on a
whole. I may have tried (just for the fun of it) a few years ago
to do just that, let me tell me, JNOS was not able to handle any
incoming packets at all. It just stalled out.

Now, mind you, I didn't give it alot of thought as to how I did
it back then. I can certainly look into it. You will always pay
a price for the so called 'value' in something.

I have a similar complaint, sometimes I need to see things on
the fly, and the best approach right now is to close the trace
when you have passed the event that you wanted to trace.


nos-bbs mailing list
nos-bbs at tapr.org

More information about the nos-bbs mailing list