[OpenAFS-devel] RX Performance

Lawrence Greenfield leg+@andrew.cmu.edu
Thu, 19 Jun 2003 11:26:10 -0400


   From: chas williams <chas@cmf.nrl.navy.mil>
   Date: Thu, 19 Jun 2003 11:00:20 -0400
[...]
   while most is good, afs does run on quite a few platforms not listed.
   i have no idea how aix, hpux (despite its bsd lineage), dux, and nt handle 
   this.  tcp connections also consume more resources than a single udp socket.

The fact that people run web servers on all of the platforms on earth
and care about performance pretty much kills the idea that TCP is
expensive or slow. The only question is the amount of effort to recode
AFS for RX-with-better-performance or blah-via-TCP. 

[Aside: NT has killer mechanisms to deal with large numbers of TCP
connections: i/o completion ports, inherited from VMS.]

   i might also add that tuning tcp is not really a simple matter either :)

It's a lot more well understood than RX.

Larry