Ethernet performance tests

Stefano Gridelli sgridelli at gmail.com
Fri Oct 29 00:25:23 CDT 2010


I tried the "ultra high throughput" script just for fun to see how much I
could push ... I got a solid 920 mbps stream for the entire time I run the
test (circa 30-60 seconds) with not spikes. The hardware in that case were
two IBM hs-20 blades with broadcom chipsets.

I said for fun because if we use ixchariot for throughput tests usually is
just for small T1 sites (max 3xT1) so I have never seen the issue you
mentioned. Usually on the same T1, we fill the data VLAN with traffic and
then we run x voice pairs on the voice vlan to validate QoS (MOS score).

On Thu, Oct 28, 2010 at 5:16 PM, Jonathon Exley <Jonathon.Exley at kordia.co.nz
> wrote:

> How smooth is the Ixchariot data stream? When Chariot was a NetIQ product
> it seemed to generate regular spikes as the algorithm tried to correct the
> total throughput over a time interval.
> It's not a problem for slow data rates but when testing near the limit of a
> circuit's capacity the spikes could sometimes overflow the buffers of
> Ethernet media converters and give false results.
>
> Jonathon
>
> -----Original Message-----
> From: Stefano Gridelli [mailto:sgridelli at gmail.com]
> Sent: Friday, 29 October 2010 1:08 a.m.
> To: Diogo Montagner; Tim Jackson; nanog at nanog.org
> Subject: Re: Ethernet performance tests
>
> Hi Diogo
>
> We use ixchariot endpoints installed on linux laptops to test sites for
> voice readiness. Ixchariot calculates for you the MOS score and, depending
> from the NIC, can also push close to 1 Gig of traffic. For larger bandwidth
> tests (I believe 6-7 Gig) and fast re-route testing (ms failover) we use
> ixia hardware.
>
> Ciao
>
>
> This email and attachments: are confidential; may be protected by
> privilege and copyright; if received in error may not be used,copied,
> or kept; are not guaranteed to be virus-free; may not express the
> views of Kordia(R); do not designate an information system; and do not
> give rise to any liability for Kordia(R).
>
>
>



More information about the NANOG mailing list