| [HELP] VOIP and data traffic with qos and vlan So...we're having some issues with our Cisco phones, which are on their own "vlan"..and the PC is plugged into the IP phone. Below is an example of a switchport config for a port the phone/PC is plugged into...
switchport access vlan 112
switchport mode access
switchport voice vlan 111
switchport priority extend cos 7
srr-queue bandwidth share 10 10 60 20
srr-queue bandwidth shape 10 0 0 0
mls qos trust device cisco-phone
mls qos trust cos
macro description cisco-phone
auto qos voip cisco-phone
spanning-tree portfast
Below is a switchport config of an uplink port to another one of our remote offices:
power inline never
switchport trunk encapsulation dot1q
switchport mode trunk
srr-queue bandwidth share 10 10 60 20
srr-queue bandwidth shape 10 0 0 0
mls qos trust cos
auto qos voip trust
The issue is sometimes the remote offices have echo issues and scratchy voice quality. Plenty of bandwidth between offices...so i'm thinking its a switch config issue? Help!! |
|
| You dont say how the offices are connected. I would assume there is a router and/or firewall in there somewhere. If you have a point to point, or frame relay or MPLS, you will need to properly mark your egress packets to match up with the queues that your service provider honors (likely ef for voice). You will need to configure QoS on every device in the path that you control (router/firewall/other switches, etc...) If you are using an internet connection, there really isnt much you can do. If you are using a VPN connection, you can do qos pre-classify (your ipsec packets are properly marked based on what they contain). You are shaping at 1/10 the interface bandwidth, which should be at least 1/10 of 100Mb, which is plenty for voice. You can (should) use the bandwidth statement on whatever interface is your bottleneck (your T1 serial, your ethernet uplink to your service provider, etc...). Your queues wont even kick in unless you max out the bandwidth (for example, if you have a 100Mb interface plugged into a 10Mb cable modem, you can saturate your uplink, but never queue because the ethernet interface thinks it has 90% available).
HTH |
|
|
|
| Thanks for the response....
Yes, there is a point2point from the main office to each of the two remote offices, both p2p's are 8mbps up and down. The internet connection isnt being used in this scenario since all inter-office calling is on the LAN...either over the P2P's or thru the main building LAN. So, not sure if I should add or remove any commands to the uplink ports on any of the offices switches for the P2P or any additional setting change for the port config for the phone/PC port itself... |
|
| reply to gknopp
Port configs for the phones are fine. You would need to adjust the uplinks to the routers that terminate the point to points, and the routers themselves. Basically, you will be marking the packets coming in and queueing them going out. You need to contact your provider and find out what your queues are and what dscp values align with their queues. Also on your router uplink interfaces (that connect to the provider), type "bandwidth 8192", otherwise, you'll never hit your queues.
HTH |
|
| great...thanks....ill check on that. But, there isnt a router on either side...its just a fiber connection between our offices so everything should be open. Its been hard to pinpoint these issues. Anyone else out there experience voice issues with Cisco phones and have cisco equipment? |
|
| reply to gknopp
ok then on your fiber uplink ports, set the bandwidth to 8192 (or less). That might be enough. It doesn't do anything to actually limit the bandwidth, it is only to help higher layer protocols make more intelligent decisions (dynamic routing protocols, QoS, etc...). You'll need to do this on you remote sites as well as your main office site. If your point to point is layer 2, don't bother calling your provider for queue information, because it wont matter. |
|
| great. thanks! I'll give it a try... also... can you help me understand what these two commands do or what they mean?
srr-queue bandwidth share 10 10 60 20
srr-queue bandwidth shape 10 0 0 0 |
|
| reply to gknopp
What is the round trip time between the offices?
Does the call go thru an analog leg along the path?
Do you have voice gateways that handle the calls? |
|
| There is no time between offices...its a 8mbps up/down fiber point to point...also, no its all IP traffic..no analog. We've got a server handling phones and the configs. |
|
| reply to gknopp
So are phones in the remote office are registered with the phone systems in HQ?
Are you using SIP or SCCP for the phones?
Which end hears the echo? remote or HQ?
can I guess....... this is a non-Cisco phone system? |
|
| reply to gknopp
Hello, here are your explanations:
srr-queue bandwidth share 10 10 60 20 - You have 4 egress queues on this interface. Voice uses queue 1. For each cycle, 10 packets are processed from queue 1, 10 from queue 2, 60 from queue 3 and 20 from queue 4.
srr-queue bandwidth shape 10 0 0 0 - You are shaping queue 1, which means that any packets in queue 1 are processed first, at 10% of the interface bandwidth. The other queues are not shaped, so they are processed with sharing.
There is a better explanation in this document (the technical writers at Cisco are a lot better explaining this stuff than I am)
»www.cisco.com/en/US/docs/switche···p1162303
Something that I missed last night (long day) was that you should add "priority-queue out"
HTH |
|
| your recommending adding priority queue out on phone/pc ports or on the phone server/uplink/trunk ports?
also..to answer the other post...yes, the phones at the remote offices connect to the HQ and they are cisco phones going to a asterisk/free pbx server in VM. Echo is mainly heard at the one remote office. And...is that shaping command, 10% bandwidth right? Should that be changed? |
|
| Use priority-queue out on all of your interfaces with egress queueing. So to answer your question, both. Without it, your priority traffic is not de-queued first. With asterisk, you must be using the sip load. I dont have any experience with asterisk or the sip loads, but you may find something on an asterisk board. The shaping command is 10%. Whether you should change it is up to you. During off hours, play around with it until it works. Run some debugs, see what you get. |
|
| reply to gknopp
Imo lack of\or misconfigured QoS should not result in echo per sa, the overall call quality should suffer.... choppy,call drops etc. Echo results from leakage of TX traffic leaking into RX at delays greater than 25 milliseconds.
- Calls made after hours or during times of NO congestion should NOT result in this echo, assuming QoS is the issue. Unless, you are doing Policing along the path
- No analog to digital conversion is occurring along the path since this is IP based. Confirm sites phone's point to the same SIP server/proxy
- Another thing you may possibly look into, is the firmware version the problem phones |
|
| Right..we're looking into upgrading the cisco phone's firmware, but are having trouble figuring it out and getting it to work from the phone server's tftp server |
|
| reply to gknopp
Just to confirm, was the firmware of the problem site different from the other working sites? |
|
| nope...same firmware on all cisco phones... |
|
