flow generating tool

Iperf(linux) and jperf(windows) does generate traffic flows. You simply set up sender/receiver and it'll generate traffic load. It can generate 10gig line rate unicast flows but multicast replication is limited because it is software based.

Iperf/jperf however, is not as flexible or anywhere as feature rich and powerful as some of the other hardware solutions as mentioned earlier - ex: ixia, spirent etc.


Does anyone follow a network performance testing methodology, using hardware
from companies like ixia/spirent?

I know that basic testing is typically done for validation of configs, but i
assume other issues would make themselves apparent when pushed to these
higher loads.



Probably more/more formal than you want, but:


And I think you may find that some of the vendor folks are active in the WG.

---George Jones

It really depends on the product you are testing. If forwarding performance is what you want to measure, you would do it with various routing table sizes (starting small and ending with a global table). Packet size is also something you should look at.

We could provide better suggestions if you tell us what your product is.

Vlad Galu

We have quite a number of companies using hardware from companies like Ixia
and Spirent - the key is to use tools that makes it easy to setup a testing
methodology without hiring a support staff.

You can contact me directly for more information and how others like Cisco,
At&T, Vodafone and many more are doing this sort of testing.



The test plan you use depends upon what you want to test - raw pps throughput, route convergence time, qos performance, etc.
We use Exfo (http://www.exfo.com) testers working to a mac-swap loopback for commissioning testing of Ethernet access circuits, looking at the usual loss/throughput/latency/jitter metrics and burst size.
When checking out new equipment in the lab we also use scapy scripts (http://www.secdev.org/projects/scapy/) to look at things like Ethertype and L2CP transparency.