Wireshark-users: Re: [Wireshark-users] Huge VoIP Problem :(
From: "Jim Young" <SYSJHY@xxxxxxxxxxxxxxx>
Date: Thu, 18 Jun 2009 14:19:55 -0400
Hello Mark,

>>> Mark Jeffers <mantramark@xxxxxxxxx> 06/17/09 12:31 PM >>>
><snip>
> The results are horrible...<snip> several of them had "Max Delta" in 
> the thousands (some in the 9000s), resulting in 90+% packet loss! 
>
> I'm at a loss here.   Obviously severe network issues, or the Phone Switch
> is bad.   I've tried everything I can think of to no avail.  Anybody have
> any ideas of what might be wrong, or what further information I should
> gather to help pinpoint the issue?   I'm going nuts here and any help would
> be greatly, greatly appreciated.  :)
<snip>

Just a WAG, but anyone verified that all the equipment port's duplex modes 
have been setup correctly?

I've seen vendors come in and unilaterally setup their equipment at full
duplex but neglect to let the network guys know!

Reports of "late collisions" on any network interface is usually a smoking 
gun.  The fact that any equipment is counting collisions events of any 
kind implies that the port is operating in "half-duplex" mode.

Others may argue on this point, but with the _current_ network equipment 
it's often best to let everything auto-negotiate their ethernet speed and 
duplex.   

The problem is this:  

Is one device port is setup to auto-negotiate its ethernet duplex but the 
peer device port is commanded to "full-duplex" then you will _very likely_ 
end up with a duplex mis-match.  

In this case because the "full-duplex" side will typically _NOT_ advertise 
how it is configured (that it can ONLY support "full-duplex" mode), the 
auto-configure side will NOT see a link advertisement regarding the 
duplex states supported by the peer and will be forced to assume the 
lowest common denominator (the peer can only do "half-duplex").  
The auto-negotiate side is virtually guaranteed to configure itself to 
"half-duplex" mode which will result in a classic duplex mis-match.

When a duplex mis-match is present I often see Packet losses of 90%.
And these packet losses are usually much worse with packet flows in one 
direction versus the other.

Perhaps this info may help.

Good luck,

Jim Y.