[c-nsp] IPerf alternative

[James Bensley]

On 9 August 2017 at 10:51, CiscoNSP List <CiscoNSP_list at hotmail.com> wrote:
>
> Hi James - Yes I tested this also....it still will not go above 50%...at first I though it may be the carriers link, but I also tested the bidirectional/simultaneous test, and it produced similar results (Single direction test were fine.)  Single direction tests are fine.
>
>
> I think its a problem with IPerf, potentiallly the server nics...Or maybe IPerf is "limited" to the bandwith of the link (1G, but can only send/receive a "total" (50% in both directions)
>
>
> There are a few reports of issues with the bidirectional tests, but I cant locate any "fixes".....so Id prefer to spend time looking at a more robust (proven) testing tool.
>
> Cheer

Hi,

We have iPerf servers and one trick we use to pretend we have multiple
clients (and thus multiple flows) sending traffic through a device is
to assign multiple IPs to the interfaces of the sender/receiver.
Perhaps try something like that, multiple iPerf instances sending
traffic between two boxes but using multiple IPs on each side;

box1  <>  box2
10.0.0.1 tx > 10.0.0.3
10.0.0.2 rx < 10.0.0.4

We had an issue a while back with some VMs only trasfering files at
only few Mbps between each other and in one direction only, we fired
up iPerf3 and sent 6Gbps of traffic (the hosts had 10G NICs) strait
away without issues or any host tuning etc. So iPerf will definatly do
>1Gbps as Ryan also showed.

Cheers,
James.