From mboxrd@z Thu Jan  1 00:00:00 1970
From: bruce@it.usyd.edu.au (Bruce Janson)
To: 9fans@cse.psu.edu
Subject: [9fans] Re: network throughput measurements
Message-Id: <20020309035920.93AB219980@mail.cse.psu.edu>
Date: Sat,  9 Mar 2002 14:48:25 +1100
Topicbox-Message-UUID: 62a9c060-eaca-11e9-9e20-41e7f4b1d025

  ...
  From: andrey@lanl.gov (andrey mirtchovski)
  ...
  Message-ID: <20020308210559.2E0D819981@mail.cse.psu.edu>
  Date: Fri, 8 Mar 2002 21:13:40 GMT
  ...
  Here are a few numbers with regards to network throughput of p9 and
  other operating systems.  No substantial discussion is presented, but
  we would like to draw your attention towards the small data size
  transmission results.
  ...
  Note: we're currently looking at ways to improve the latency penalty
  for small data size packets and suggestions are most welcome.
  ...
  --- linux ---
  ...
  100BaseT on Linux (RH 7.2):
  ...
  10000 iterations over data size 64:   111.30 mbits/s...     0.05 seconds
  ...

andrey,
    Nice numbers but I would rather you ran each test for a fixed
number of seconds (8?) rather than a fixed number of iterations
(here, 10000).  When looking at run durations of 0.05 seconds you
are getting quite close to the resolution of some operating systems'
clocks (e.g. it used to be hard to get better than 100Hz resolution
out of linux).  But maybe your netpipe port will do this for you...

Cheers,
bruce.