[PLUG] Verizon FIOS - Traffic shaping against VOIP?

[Brent Jones]

Keith Lofstrom wrote:
> When looking at path delay between my Verizon FIOS home machine and
> a machine in Portland on Alternet, I noticed some "interesting"
> ping results.  1000 pings sent at 0.2sec intervals sent to the 
> Verizon POP in Seattle revealed that the vast bulk of the pings
> returned between 12 and 15 msec - about what one would expect from a
> lightly loaded network.  However, about 50 (5%) of the pings returned
> between 20msec and 80msec, with a fairly uniform distribution.  
> 
> This is not what I would expect;  I would instead expect that the
> delays would follow a Poisson distribution, smoothly tapering towards
> zero with longer delays.  However, I don't know a lot about network
> traffic delays.  Is this normal, or are some of the packets being
> intentionally delayed to screw up services like VOIP?
> 
> This weekend, I will run some longer ping trials to get a better
> idea of the shape of the outliers, perhaps by running two pings
> in parallel with slightly different intervals to see if (1) the
> delays correlate and (2) they interfere because of some throttling
> mechanism.  I will also run a looping ifconfig on the firewall to
> do a total packet count on the interface.  Anything else I should
> try,  such as setting particular QOS bits on the ping?
> 
> Also, it would be nice to try the ping tests against another 
> Washington County Verizon FIOS customer.  Volunteers with an
> IP address?
> 
> Keith
> 
> 

Any large ISP is probably going to set the QoS priority on pings to be 
pretty low; which makes performance trending using only ping pretty 
difficult.

I would try hping using your protocol of choice (IP/UDP/ICMP) against 
the service ports you are trying to use.
Vary your packet size, etc to try and simulate normal usage.

Brent Jones