Allied Telesyn QoS testing

Hi guys,

I need to qualify an Allied Telesyn router from a delay, jitter, packet loss and, most importantly, QoS. The router is GbE, so we are arranging a raft of linux PC's to drive the router to the congestion point and verify that the QoS parameters are being adhered to.

Additionally, we then need to feed a number of QoS-managed video feeds through with some noise traffic to ensure that the QoS policies protect the video feed.

The intention for the bulk of the testing is to use IPerf as the traffic generator, enforcing the TOS bits appropriately to validate the links.

As I am a bit new to this, and we are on an incredibly tight project timescale, I wonder if anyone who has had similar experiences would care to share the common gotchas and advice they gathered whilst performing similar testing?

Many thanks for your kind assistance.

All best, Chris B.

read the RFC.

formatting link

we have an internal lab / reference network where we test various things before they go into deployment.

1st suggestion is to avoid the issue and ask the manufacturer to do it for you (this only works if you are likely to spend serious money on the kit, and are willing to buy something else if they dont play ball). you also need to check what they do to make sure they dont cheat to avoid an issue they already know about.....

testing depends on whether you are willing to spend some money - but if you havent done this before, and want quick results try using a test specialist (be prepared for a serious bill) - i have had good experience with

formatting link

(but that probably doesnt help if you arent in Europe)

if that is too steep, then a "real" test device will help, since they provide traffic generation, emulation and measurement all in 1 box, and come with applications and scripts for common scenarios. if you only need it the one time, you can rent it rather than buy.

if all else fails, look at the white papers on the testers sites...

formatting link

and probably lots of others

the rules: testing is mainly black magic. it takes 10+ times as long to set up a test than to run it. you always need to run the tests several times. you cant do the testing until you understand how the equipment under test works, set it up properly and what the test actually does. you need to know what you want from each test before you start to design it. most tests dont tell you what you want to know manufacturer tests give best possible numbers - so derate what you expect from the kit (by maybe a factor of 2) to allow for working in a realistic environment.

given all this, starting from scratch on a tight deadline is a high risk place to be.

if all you want is to try a couple of boxes then the testing costs will swamp the cost of the hardware - so use high end manufacturer kit unless you have a good reason not to.

and if you are using expensive kit, the manufacturer may have already paid someone else to do a similar test already and you can just use their results.