Testing 10Gb fiber optic link between switches

LPT Administrator · ‎10-18-2013

I have a Cisco 3750X 48 port switch connected via a OM-3 fiber optice cable a Cisco 2960S 24 port switch via the 10G-SR SFP+ transceivers. I am trying to determine the maximum throughput of the link.

In theory, the link should be able to do 1310720 kB/s. The best I have been able to get with TTCP is an order of magnitude lower at around 1316 kB/s

The results are 67108864 bytes in 49770 ms.

I am using the default settings except I set the TCP Recieve Window size to 65536 (or higher, doesn't matter).

Am I reading this utility wrong or is it just not that useful of a test of throughput between switches?

I have tried tweaking every setting I could find and nothing seems to make any difference.

Thanks

rfalconer.sffcu · ‎10-18-2013

My guess is the devices can't generate any more traffic than that. When running tests using pcattcp on end points, I need to run multiple streams (4 or 5) to get even close to filling a 1Gb link.

LPT Administrator · ‎10-18-2013

Yeah I think you may be right. If I connect a computer to each switch and run iPerf I get a very consistant 938Mbps. I tested the two computers with a crossover cable to establish a baseline and that is what I got also.

I tried saturating the link with various number of multiple streams and the link speed was unaffected. I maxed out the ability of the computers to keep up.

What did you look for when you did your tests? Dropped packets? throughput speed reduction?

I wish Cisco IOS had an easy way to determine total actual throughput between devices like an LED bargraph on the front panel!

Thanks

rfalconer.sffcu · ‎10-21-2013

I'm typically not trying to judge throughput when I do these tests. I use it more for link saturation to test QoS settings.

Do you have Netflow set up? This could probably give you some interesting data on throughput between devices.

Leo Laohoo · ‎10-21-2013

I maxed out the ability of the computers to keep up.

How many computers?

Say you put 24 computers into a 24-port switch and do a simultaneous test to push data traffic at 1 Gbps. You'll max out the buffer of the entire switch very quickly.

Joseph W. Doherty · ‎10-22-2013

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

Both those switches should be able to drive a single 10g port to full capacity.

Unless you have special performance testing equipment, it would be difficult for you to verify this. If you need verification, search for performance testing review reports (Cisco sometimes lists those it "charters".)

BTW, there's much more involved whether a switch will deliver the performance you expect. You mention using TTCP, but TCP has its own quirks depending on a particular host's implementation of TCP. Even the best and optimally configured TCP setup may have issues with "high" bandwidth (which is why there are so many proposed TCP improvements).