Bandwidth & QoS opinion

[snootalope]

Hello

I've been struggling with what I think is a QoS problem, but i just can't seem to make it better.

I've got a remote office, conected via a fractional T1 with a cir of 768k. A 1751 router sits on each end of the T. I've only got maybe 10 people in the office and they all use dumb terminals that emulate a session at the home office via Citrix. Each user has a VoIP phone at their desk that talks to the call managers also in the home office.

Only 10 people shouldn't be able to utilize all that bandwidth so much that they're getting disconnected from the sessions, well that's what I'm guessing..

here's my QoS/config on each router:

!
class-map match-all VoIP-RTP
match access-group 105
!
!
policy-map QoS-Policy
class VoIP-RTP
priority 600
class class-default
fair-queue
!
!
interface Serial0/0.16 point-to-point
description Austin
bandwidth 768
ip address 172.17.17.5 255.255.255.252
no ip mroute-cache
frame-relay interface-dlci 16 CISCO
class voip_768
frame-relay ip tcp header-compression
frame-relay ip rtp header-compression
!
!
map-class frame-relay voip_768
frame-relay cir 768000
frame-relay bc 7680
frame-relay be 0
frame-relay mincir 768000
frame-relay fragment 960
service-policy output QoS-Policy
!
access-list 105 permit ip any any precedence critical
access-list 105 permit udp any any range 16384 16400
access-list 105 permit ip any any dscp ef
!

Can some expert out there tell me if I've got this configured wrong maybe? I'm giving priority to the voice traffic, which a Cisco TAC guy helped me with, but I don't know if that's choking down the link or what?? Anyone got any advice??

 

[KiscoKid]

Firstly, I can't see anything fundamentally wrong with your configuration.

Secondly, I think you are flying a bit near the wind giving voice nearly 80% of the available line rate. The reason for this is that I'm led to believe that Citrix typically utilises between 20-40k per session per user. So with 10 users you could be maxing out your line (assuming the voice is fully in use).

You can confirm what your QoS policy is doing realtime using the 'show policy interface' command. The next time your users complain about the performance, use this very useful command to determine which class (voice or default) is eating all the available capacity and perhaps resize accordingly. This command shows very useful output inc. the QoS policy drops. You only see drops during congestion and only for traffic that is outside the defined limits.

 

[BuckWeet]

Question..

do you really have frame relay between sites, or is it a point to point t1?

also, if it is frame, do you really have a 768K CIR, or is the pipe just 768k and you have a 384k cir?


BuckWeet

 

[snootalope]

Well, our previous provider was for sure a frame circut, but reading our new providers "description of service" document all it says is it's a "Meshed WAN" with PVC's and specific DLCI's.. so, it doesn't exactly say Frame-Relay.

Yes, the CIR's are 768k. Positive on that.

 

[snootalope]

Ah! did find one portion mentioning it.. "The Mesh will be provisioned with ports and PVC's at the CIR selected by us via either ATM or Frame-Relay encapsulation.

 

[BuckWeet]

I've had issues with the fragment command.. you might try to remove it.