[Asterisk-Users] ADSI phone vs. IP phone (and proper implementation thereof)

[Rich Adamson]

> > Probably because it's well known that these setups are prone to failure
> > of either the PC's connection, the phone's connection, or degredation of
> > one/both.  It also breaks switch envirenments where spanning-tree
> > portfast is enabled (not as big of a deal if the deployment is in
> > concert with the infrastructure group, as it should be).
> > 
> > Vendors should NEVER have implemented this functionality into phones
> > unless it was working under all conditions.  Personal experience shows
> > that it is most definitely not on Cisco and 3Com products.  Others have
> > told me their stories with other manufacturer's equipment.  None of it
> > was good.
> > 
> > It's not a production-stable way to deploy phones.  Period.
> 
> I'm wondering if what you say is actually true.  According to recent media 
> releases, Cisco has shipped over 2 million of their IP phones.  They must be 
> doing something right.  Their phones are _designed_ to function and cooperate 
> with the switch.  Obviously, the installer has to be totally familiar with all 
> phone, switch, router and network settings in order to have a successful 
> installation.
> 
> The switch needs to be configured with specific port, vlan, and class of 
> service settings.  Accepted practice is to provide a voice vlan and a data vlan.
> 
> On the phone side, the phone knows to send voice on the specific vlan told to 
> it by the switch , and to pass through data from the pc through the vlan told 
> to it by the switch.  The phone knows to prioritize voice traffic over data 
> traffic.  So does the switch.  And so on through the connection of switches and 
> routers.  This ensures voice quality and precedence through out the network.
> 
> Voip quality is not necessarily about bandwidth (because it works on T1 data 
> lines as well as GB ports), but about instantaneous bottlenecks in the 
> network.  These instantaneous and random bottlenecks can occur in the cad 
> environment mentioned.  But with appropriate COS (layer 2) and TOS (layer 3) 
> settings in the phones, switches, and routers, these bottlenecks become non-
> issues.
> 
> In addition, what many people forget, or learn by experience, is that you 
> absolutely _must_ have everything running full-duplex, and to physically check 
> errors and statistics on each port of the switch in order to verify that you 
> have error free links.  You won't believe how many networks out there are 
> broken because noone checks and fixes these issues.  A voip network _must_ have 
> managed switches so you can verify these things.
> 
> There was mention of a heavy cad environment.  Say your computer is connected 
> to the 100mbps port of the phone.  A g.711 call comes through.  The call takes 
> around 80 kbps.  If I've done the math properly, the voice call takes only 
> 0.08% of the bandwidth, hardly something that will interfere with 'heavy cad 
> users'.  More likely the opposite, the heavy cad users will interfere with the 
> call, _but_ _only_ if the switch and phone are not configured properly for 
> vlan, cos, tos, speed, and duplex settings.
> 
> So having said this, you mentioned that you have had personal experience where 
> this functionality is built into, or does not work in Cisco's case.

Couldn't agree with you more (and I'm not the original poster). We've spent a
number of years conducting independent (no vendor alliances) network performance
assessments for corporations in more then 40 states, and have found a large 
percentage of network managers and technicians just don't pay attention to 
these things (for lots of reasons).

As far as the switch function built into many of the sip phones, there has been
a fair number of folks on this list that have had problems with it. If I
recall correctly, John Todd (very experienced) recently queried the list
relative to unusual C7960 switch problems. Unknown as to whether the root-cause
was hardware failures, STP, or what, but maybe John will post his findings.

Given our extensive experience with performance analysis, I would not use the
switch function "if" it was limited to 10 meg half duplex except in very low
usage office environments.

It would be very interesting to hear from those that have real life experience
with the switch function in network environments that are much larger then
the typical SOHO shops, and that have invested the time/effort to properly
diagnose the real root-cause of such issues.

Rich