2

for testing purposes I need to stress an 8-port managed ethernet switch. Essentially, I want to simulate a scenario in which two hosts on the switch are exchanging data and the other hosts connected to the switch are trying to use all the available bandwidth. The idea is to see how a network application behaves when the other hosts on the same switch are using the network too much. How would you do that? Thanks a lot in advance!

5
  • 1
    But WHY??? it's a SWITCH, so long as it's non-blocking that's all you need to know - what do you THINK is going to happen?
    – Chopper3
    Jul 31, 2013 at 12:30
  • I expect that the performance of the data transfer will drop. Is there something I'm missing?
    – Vindem
    Jul 31, 2013 at 13:13
  • On a non-blocking switch if say server A and B are using 100% of their own bandwidth talking to each other then that doesn't matter if server C and D want to do the same, A+B won't impact any other ports but their own, obviously if A+B are using 100% of their links then server C will struggle to transmit with with A or B as their links are clearly busy.
    – Chopper3
    Jul 31, 2013 at 13:29
  • If you really want to make sure that your switch is non-blocking (and I believe that even the cheapest ones these days are), you can try Iperf Jul 31, 2013 at 13:44
  • I thought there was a shared backplane that (depending on the manufacturer) has a throughput limit lower than the sum of the individual ports. For example (using figures out of my behind), an 8-port gigabit switch with a theoretical 16-gigabit FD throughput might have a backplane that's only capable of say 12 gigabits. Is my understanding wrong there?
    – fukawi2
    Aug 1, 2013 at 4:56

1 Answer 1

5

Purpose-built Ethernet test gear is highly specialized and can be very expensive. You might be able to find some used gear that does what you want for 100Base-TX applications, but I doubt you're going to find gigabit gear for anything you want to spend.

As the comments have already stated, most commodity gigabit Ethernet silicon today is non-blocking and you can take the manufacturer at their word that the gear is going to do what it says. (Pretty much everybody, except for the big "name brand" switch manufacturers, are just using ASICs from the same OEMs anyway).

Having said that, there's no reason that you can't do as @dusan.bajic says in the comments and use a traffic generator tool (iperf, ttcp, etc) to attempt to flood the device with frames (likely just to beat a dead horse). More than likely you're going to bottleneck on your host operating system's traffic generation capabilities before you hit maximum theoretical throughput on the switch. Purpose-built Ethernet test gear won't have that problem because it's built to run at 100% theoretical maximums.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .