[SOLVED] Testing various Gigabit Switches - here are the results...

[alexb75]

So, needed an 8-port switch, and went NUTS, bought FOUR different ones to try this upcoming weekend. I am sure ALL would do the job, BUT for FUN, I wanted to somehow test them and see which performs best, which could also be a good result for others in the market.

Question is, HOW do I do that? I'd like to basically know, the level of speed, packet loss, what's their limit transferring lots of data across all ports, and to possibly measure noise they might add on the line. Which I could do at home with a few PCs, other wired devices without special equipment?!

Thanks

 

[bill001g]

You would likely not find much interesting. Many years ago switches used cpu based devices and software and there used to be some difference.

Modern equipment uses ASIC so they have the ability to run every port at full speed both up and down simultaneously. You will see many 8 port switches advertise a backplane speed of 16gbit.

It technically is more complex because a 64byte packet takes just as much overhead as a 1500 byte packet. It is still almost impossible to detect the difference on a switch. You do see rates specified in PACKETS/sec for things like firewalls.

There might be some difference in switches that have 10gbit ports. Not all have enough backplane speed to run a bunch of 10g port at the same time. Still even very inexpensive 8 port 10g switches claim to be full wirespeed/non blocking switches with a backplane of 160gbit

 

[nigelivey]

Absolutely no point, would guarantee all will function exactly the same at full gigabit speed on all ports, the latency variance would be so small you wouldn't be able to measure it reliably anyway. Not much fun to be had!

 

[alexb75]

Thanks guys, I was hoping to do something like described here, but I guess in my case it wouldn't make much of a difference then! Damn...

Testing Suite And Methodology - How We Test Network Switches - Tom's Hardware | Tom's Hardware (tomshardware.com)

 

[ajohnson30]

nope. If you really want to, you can look up iperf (iperf2 or iperf3). You'll need two computers directly plugged into the switch you're testing. One computer runs iperf as a (listen) server, the other runs (client) testing commands.

Do that for each switch. Look at the results. Most likely if you test properly, each switch should come out about the same. If they don't, its probably something you did