[netperf-talk] Bidirectional tests

[fonesurj at verizon.net]

Rick:

Hello.  I have been fiddling with some different tools (IPerf, pktgen, D-ITG, etc) testing the performance of a PC running the Vyatta OFR.  See http://www.vyatta.com/twiki/bin/view/Community/DellDimensionB110Test

I have found I am not a fan of IPerf due to the fact it doesn't really support 64 byte size L2 frames.  I'd like to run a series of tests, across all the packet sizes identified in RFC 2544.  

Since each of the PCs has only four fast ethernet interfaces, I figure I could push two bidirectional streams across the router, for an ideal total of ~382mbps @ L3 (measuring the IP/UDP headers plus payload).  In practice, of course, the PC I'm using in the test won't handle that.  The PC will be limited by the PCI backplane (roughly bandwidth limitation) or its ability to handle interrupts (roughly packets-per-second -- I know its not the same exactly).

So Netperf looks interesting, but I need some clarification on a post to this e-mail list you made concerning bidirectional tests.

===============
However, if one 
configures netperf with --enable-burst, and are careful about socket buffer 
sizes relative to the request/response sizes and the burst size, one can achieve 
the same thing.
================

1.  Can you elaborate on the relationship between the socket size, the R/R sizes, and the burst size?  Like "socket_size >= burst_size * RR_size" or something to that effect.

2.  The R/R sizes are the size of the TCP message (which is then subject to being broken up at L3)?

3.  The burst size is the number of packets (L3) or messages (L4)?



Thanks.  I find that being clear on these things helps a lot.  From one tool to the next definitions and references seem to change.

Derick Winkworth
CCIE #15672