testing QoS with conventional PCs

Have a question or want to start a discussion? Post it! No Registration Necessary.  Now with pictures!

Threaded View
Hello,

what are the methods to test QoS functions (specifcally I'm interested in
traffic shaping, scheduling) without packet generators Smartbits or Ixia,
but having only a few PCs and a target board running some code on L2 switch?
So basically, I'm trying to implement QoS support for a Ethernet chip, i.e.
QoS per se is done in hardware, my software is doing configuration and
monitoring, and thus I want to check if it really works or not.

Let's say, connect one PC to port 1 of the board and second PC to port 2,  I
can generate traffic via port 1 and expect to be shaped or scheduled
accordingly -- what is the sdcenario, or procedure, to do this? What tools,
utilities I could use for this, provided that both PCs and the board run
Linux?

Thanks.



Re: testing QoS with conventional PCs
["Followup-To:" header set to comp.protocols.tcp-ip.]
On Mon, 2011-08-29, Mark wrote:
Quoted text here. Click to load it

I'd expect you to need more than two hosts ... ? But I don't know your
requirements.

Quoted text here. Click to load it

Assuming you really only need a handful of endpoints:

- All the normal command-line tools and scripting languages
- Its configurable IP stack
- Its firewall
- ...

The "normal tools" I'd use would include ping, netcat, wget, ssh and
rsync. And the TCP and UDP flavors of the 'echo' and 'chargen' services.

Maybe you want one single tool which gives you one simple answer instead.
Someone else will have to mention those -- I don't know them.

/Jorgen

--
  // Jorgen Grahn <grahn@  Oo  o.   .     .
\X/     snipabacken.se>   O  o   .

Re: testing QoS with conventional PCs

Quoted text here. Click to load it


I would not presume to call it a drop-in replacement for smartbits or
ixia, but I do have a certain fondness for netperf:

http://www.netperf.org /

which can be used to generate bulk streams, request/response traffic,
etc.

happy benchmarking,

rick jones
--
It is not a question of half full or empty - the glass has a leak.
The real question is "Can it be patched?"
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...

Re: testing QoS with conventional PCs
On Mon, 29 Aug 2011 14:34:13 -0400, "Mark"

Quoted text here. Click to load it
1st bit is that it is difficult to do any QoS testing for a switch or
router unless you can overload a port.

So - traffic into 1 port faster than the output, or into 2 ports same
speed as the output.

2nd is that before you test your switch, test the PCs "back to back" -
because they may well not get to wire speed on a Gigabit Ethernet
port...

Quoted text here. Click to load it

The way we do this at work is to start simple and add complications.

a. 2 PCs back to back
b add switch without any qos enabled.
c add qos
etc

you are looking for how things change between tests.

1 place the Smartbits etc win is with scripting the tests so you can
re-run a varient of the same test, increasing load each time and show
the trends in the results, (or adding more IP subnets, or multicast
addresses or whatever you are stressing).

Quoted text here. Click to load it
Good luck
 - most tests seem to find problems with the test rig and software a
big chunk of the time :)
--
Regards

stephen_hope@xyzworld.com - replace xyz with ntl

Re: testing QoS with conventional PCs
Thanks everyone for input. I was rather looking for the principles of
testing of specific QoS features, for example, I can assign a 802.1q
priority value to a specific output queue (with low, mid or high priority).
So, for this particular test case the question is: how, having just a couple
of PCs and a L2 development design in between, to verify that the L2
hardware really does what I ask it to, i.e. placing a frame with COS
priority value 2 to TX queue 0 (as example), and a frame with COS value 7 to
TX queue 3.

(I'm trying to implement such functionality on the hardware and want to make
sure that it does what it claims in datasheet).

Will be very thankkful for hints and advices !


Mark



Re: testing QoS with conventional PCs
["Followup-To:" header set to comp.protocols.tcp-ip.]
On Wed, 2011-09-07, Mark wrote:
Quoted text here. Click to load it

That's not functionality; that's stuff that is internal to the chip,
and can only be verified by looking at its counters if it has any.

The /functionality/ would be how traffic with different COS values
affect the dropping of each other during heavy load.

I suppose you could set up something like this:

  host A: addr A ----- SUT ------ addr B0 : host B
                                  addr B1
                                  addr B2

And add iptables rules to rewrite COS on traffic
to B0, B1 and B2.

/Jorgen

--
  // Jorgen Grahn <grahn@  Oo  o.   .     .
\X/     snipabacken.se>   O  o   .

Site Timeline