Hi,

Here is the situation :

Subnet A <-> Cisco A <-> Public Inet cloud <-> Cisco B <-> Subnet B
^ ^
| |
v v
Phone Phone
station A station B

Cisco A has (symmetric) upstream link of 2 Mbps. Cisco B - symmetric 256 Kbps.
They route VoIP traffic between two office phone stations, plus IP traffic
between two subnets, as well as other IP traffic with other Inet destinations.

The problem is that Cisco B with its narrow upstream link employs various QoS
features to raise VoIP quality (like prioritization/queueing/fragmentation/etc).
But Cisco A feels that its upstream link is wide enough and simply does not
employ these QoS features when sending outgoing IP+VoIP traffic to Cisco B. As
a result, Cisco B experiences congestion and delays on its incoming traffic,
that degrades its VoIP reception quality.

Now what I would like to hear is your practical help on what to do on Cisco A
to explain it that for Cisco B and Subnet B it should keep outgoing traffic in
bandwidth of no more than 256 Kbps ? I.e. to turn on VoIP QoS features not for
certain interface, but for certain routing direction outgoing through this
interface ?

Thank you in advance for your replies. Hope I am not asking for too much.
I have read qcfbook.pdf, but so far I could not deduce how to combine features
described there to reach my goal, and there were no obvious solution.

Please cc: your reply to my email.

Not sure what model routers you are using, but, if supported, you
should check into RSVP. This will ensure that the voip bandwidth is
reserved end to end before a call is placed.

I assume that you use a VPN tunnel to connect your two sites, because
all the routers in the path must be configured for RSVP. Also, I assume
you are taking advantage of the lower bandwidth needs of G.729 for
your voice codec.

http://www.cisco.com/en/US/partner/p...00b75 bb.html


:The problem is that Cisco B with its narrow upstream link employs
various QoS
features to raise VoIP quality (like
prioritization/queueing/fragmentation/etc).
But Cisco A feels that its upstream link is wide enough and simply does
not
employ these QoS features when sending outgoing IP+VoIP traffic to
Cisco B. As
a result, Cisco B experiences congestion and delays on its incoming
traffic,
that degrades its VoIP reception quality.

Kevin Widner wrote:
On both sides, they are 36xx models.

Of course yes. g.729 is the default, why should we change it ?

We use tunnel only for IP routing between subnets. VoIP goes
straight between routable IP addresses of Cisco interfaces,
through 10+ hops of public Inet.

I am learning further, and what if :

1. create access-list including CiscoB and SubnetB
2. create class-map matching this access-list (access-group)
(and may be something like 'match ip rtp' ?)
3. create policy-map, specifying bandwidth limit and fair-queuing for
this class
4. assign service-policy output for this policy on external interface of
CiscoA

Please correct me if I am wrong.

Wladimir Mutel wrote:
Since voip is symmetric you need to limit your voice bandwidth on rtr A
to the same as on rtr B. Don't use autoqos!