without more information on exactly what you want to test
and the resources avaiable to you it is difficult to
give precise recommendations however I would want:-
Some way of verifying the actual path in use
network equipment commands
Some way of interrupting the path
Some way of verifying he failover time
Ssay use iperf to send stream of udp traffic
of some known rate and on the receiver use
wireshark to capture the packets. Then look at
the timestamps and the iperf interface to look at the
numeber of "lost" packets.
There are commercial test tools available
Spirent smartbits - maybe
There is a Japanese company that does high end test tools
I forget the name sorry.
netperf or tcpdump may be more to your liking than iperf.
One thing that you need to consider is that a real
in-service failure may not be as tidy as the "failure"
you get when pulling a wire. STP for example
(posssibly not RSTP - I am not really
that familiar with it) was and is susceptible to
unidirectional link failures where STP does not correctly
detect a link failure when one of the directions
of an ethernet link becomes disconnected but the other
direction does not.
Cisco has a fix - UDLD. UniDirectional Link Detection.
This uses multicast frames so the transport needs to
What i am using some tools are SMB and IXIA either check Rx/Tx
or use wireshark to see timestamps. Since the results are
not consistent, so I am looking for a dedicate tool or equivalent for
Ring/STP/RSTP,then I can repeat test it very quick. Does any
tools has Ring/STP/RSTP recovery time measurement feature.
just like a running watch.
at work we have used Smartbits and Agilent testers of various
however - what may help more is to describe how we do it.
you want a tester that generates a packets stream at "reasonable" load
- at least 1000 pps if you want 1mSec resolution, and you want the
recieve section to show a graph of no of packets/ unit time making it
thru the link.
then you break the working link and look at the graph during the
if you want some more paranoia, repeat at various load levels, from
very low load to more than 100% applied, since the stuff may react
differently under different traffic conditions.
make sure you simulate various faults - physical fibre disconnect gets
signalled very quickly.
however - i dont like the idea of relying on spanning tree of any
flavour to handle network repair.
spanning tree doesnt react well to some error types, since the
"suppress loop" packets go in the opposite way across links to
potential looping packets, some types of 1 way link fault can convert
a working "resilient" ring into a broadcast packet generator.
something like OSPF and routing copes much better with the hassles
real life throws at networks.
we used the graphing stuff that comes with the agilent console - but
all you need is something that shows traffic load as a graph and the
timescales you want.
if your traffic generator doesnt come with a stats display. then
wireshark may be able to do it.
note - this is just looking at arrival stats - ideally you want to tag
every frame you send and make sure how many you lose during the
topology change, and that the ones that survive, stay in order etc.
but - just seeing how big a gap between cut link and recovery gets you
most of the way there....
This is a fine plan from Stephen.
Send with say iperf (UDP) and run Wireshark on the iperf
receiver (or otherwise) and use the ioGraph as described.
(filter on the iperf traffic).
If iperf does not sequence number the packets in the data
portion then you could use the IP Identification field
verify how many packets had been lost and if any
had been re-ordered, if that was required.
Wireshark will display these graphs in real time
as the packets are captured.
I get the idea that you are looking for some sort of
automated tester that just magically displays the
failover time. I doubt that such a thing will be available.