[Tmrg] Queue size - Towards a Common TCP Evaluation Suite

[garmitage at swin.edu.au (grenville armitage)]

Tom Quetchenbach wrote:
	[..]
> Has anybody else done any investigation of this? Especially interesting
> would be if you could use Linux's tcp_probe module to plot the TCP RTT.
> Unfortunately my modem only works in Windows, but I may be able to rig
> something up using another computer as a gateway and get some
> measurements this way.

<possible_tangent>

You might find http://caia.swin.edu.au/tools/spp/ of interest.
A small tool we put together to passively estimate RTT between
two points on the network based on arbitrary flows of packets
seen heading in each direction. (It also works with flows that
are asymmetric, i.e. more packets/sec in one direction that the
other.)  Basically you capture packets at either end of your link,
pass both files to SPP and out pops a sequence of RTT estimates
versus time. By filtering out different application flows from your
captured traffic you can isolate the RTT vs time experienced by
different flows passing the same points of the network.

Right now the tool's only been compiled under FreeBSD. But it reads
standard pcap files so you can do the capture using whatever hosts
happen to be at the points of interest in your network.

</possible_tangent>

cheers,
gja