QOS Help 2960x/X2948G/2821

[e_dunnington]

Hello,

I am hoping that someone can help me. I am pretty good with Cisco but not so great with QOS, let me explain my situation.

Our current location has 2 different l2 switches that are not doing any routing at all. Those are the 2948 and 2960. I have phones connected to the 2960, on the ports i have put 

 switchport access vlan 55
 switchport mode access
 switchport voice vlan 20
 srr-queue bandwidth share 1 30 35 5
 priority-queue out 
 mls qos trust dscp
 auto qos trust 
 spanning-tree portfast
 spanning-tree bpduguard enable
 service-policy input Voice_POLICY

and on the trunk port for the 2960 i have:

 description UPLINK to CORE
 switchport mode trunk
 srr-queue bandwidth share 1 30 35 5
 priority-queue out 
 mls qos trust cos
 auto qos trust 
 spanning-tree portfast disable

I just started working here and we have roughly 20 - 30 phone agents on the phones all the time and we are getting static and voice loss when talking. I looked and there does not appear to be any QOS enabled anywhere except what i have done above. I am wanting to enable QOS on the 2948 (That is how the 2960 connects back to our main router and also enable QOS on the 2821. I am at a loss when it comes to setting up QOS on the 2948 and for the 28281 i have:

class-map match-any VOIP
 match ip dscp ef 
 match protocol rtp
 match ip precedence 6 
 match ip precedence 5 
class-map match-any VOICE
 match access-group 130
class-map match-any MATCH-EF
 match ip dscp ef 
!
!
policy-map VOICE-MATCH
 class MATCH-EF
 class VOICE
  set ip precedence 5
policy-map VOIP-QOS
 class VOIP
  priority 960
 class class-default
  fair-queue
  random-detect
!
!
access-list 130 permit ip any 10.10x.x.xxx 0.0.0.255

interface GigabitEthernet0/0
service-policy output VOIP-QOS

interface  GigabitEthernet0/1.4
service-policy output VOICE-MATCH

My questions are:

1) Is there anything else i need to do on the 2960?

2) For the 2948 how do i enable QOS so my voice packets still take priority?

3) The 2821 config i believe is correct but want to make sure i am heading the correct direction.

*Also we have 2 differnet types of phones, altigen 720's and Polycoms.

Please help 

[Joseph W. Doherty · Accepted solution]

"1) Is there anything else i need to do on the 2960?"

On your 2960, this "service-policy input Voice_POLICY" I believe overrides "mls qos trust dscp". Further if you're using DSCP tagging replace "mls qos trust cos" with "mls qos trust dscp".

Unfortunately, I forget what the "auto qos trust" does and whether it overrides your other trust statements.

"3) The 2821 config i believe is correct but want to make sure i am heading the correct direction."

What's the available bandwidth on your g0/0?

Your policy-map VOICE-MATCH's class MATCH-EF precludes application of class VOICE setting IP Prec 5 to its traffic - this is intended? Also DSCP generally uses DSCP EF for VoIP (bearer), your class VOICE is tagging all traffic that matches the ACL with IP Prec 5.(?)

IP Prec 6 isn't normally used for VoIP traffic. You're sure this is correct?

I would recommend you not use RED, especially as you're also using FQ.

[e_dunnington]

The 2960 config ill look into what you said.

For #3.

The port is connected at a gig, the network it self is not setup properly so i am working through ALOT of hurdles.

This config i found offline, researched the commands and it semi made sense as to why i reached out.

Thanks, Hope this helps some.

What about the CATos?

[Joseph W. Doherty]

Ok, if you have a full gig's worth of bandwidth, you might consider allocating more than 900k to LLQ.

As for the 2948, I recall the 29xx series QoS support varied whether it had an "enhanced" image or not. They also didn't have as many QoS features as the later switches. It might support some global wrr-queue statements. Check the IOS documentation for your device and IOS version.

[e_dunnington]

Yes I looked on the 2948 and there isn't much and what there is, it is very limited.

Is there a way for me to take 15% of the bandwidth for VOIP? rather than bytes?

Also I researched the 2960x config and couldn't find anything. Any suggestions on changes to it?

[Joseph W. Doherty]

"Is there a way for me to take 15% of the bandwidth for VOIP? rather than bytes?"

On the 2821? If so, latter IOS versions support "priority percent #".

"Also I researched the 2960x config and couldn't find anything. Any suggestions on changes to it?"

Well you have the most important thing, i.e. PQ enabled.

2960s have little buffer space, and their default buffer management locks buffers to interface egress queues, which often results in a high drop rate. You may find you'll need to adjust buffer parameters to reduce such drops.

[e_dunnington]

Ok thank you, my last question would be and what I just discovered by doing research.

My company is running 2 PRI's for our phones to go across. Based on what I read is that with PRI's I will not really need QOS because PRI's do not operate within the OSI model and that the PRI is possibly the cause.

Thoughts on this? The reason for all the questions is because both incoming and outgoing calls we are having major jitter and sound loss while on the phone with people outside the building. I immediately thought of QOS VOIP but based on what I just read that might not be the solution and that we might be maxing out our throughput on the PRI's.

[Joseph W. Doherty]

If your PRIs are only carrying VoIP, then QoS is unlikely to help. However, if your PRIs are carrying VoIP and anything else, then QoS might make a huge difference.

[e_dunnington]

No all my traffic (data) is flowing out of my SonicWALL. I just checked my firewall to see if any voice traffic was flowing that way and there isn't. So I would assume that my voice is going out of PRI's only and no data.

I do not know how to check the PRI's for IPs or data traffic but I do not believe it is having any extra data going through it at this time.

[e_dunnington]

Sorry for the delay i have really been digging into this.

So after further discussion i need some more input. Our voice vlan is 10.104.103.xx

our Adtrans which have our PRI's and our PBX are on a different vlan 10.104.1.xx

We are still having the issues and packet loss showing up inside the adtrans stats page but i am not seeing anything on the port itself that match the phone that is having problems at the time. My managers are telling that PRI's only come into play when leaving the building.

Can you possibly shed some light on this? I am under the assumption that our possible problem might be when the phone traffic is jumping from phone vlan to another and back again. Any help is appreciated.

Thanks,

[Joseph W. Doherty]

Didn't you already mention you believe there's congestion on your PRIs carrying VoIP? If so, that might be the root of your problem.

Or, are you saying there VoIP issues within the same building, just running across your LAN topology?

[e_dunnington]

Yesterday I went around to our call center members and asked them to detail out their issues.

Some said, they only have problems calling external which tells me to look at the PRI. Then some of them said they also have problems calling internally to extensions inside the same building which makes me think our virtual PBX might be the cause or possible network issue.

Basically I was just wondering if since the Adtran's and the PBX virtual server are on our main data vlan and not setup in our voice vlan would there possibly be an issue when the voice traffic is moved from one vlan to another.

I am trying to rule out network problems, a vendor manages our virtual server for the PBX and for the Adtrans, I am trying to cover my bases from a networking stand point.

Sorry for any confusion.

[Joseph W. Doherty]

So, you haven't found any switch ports showing drops?

Excessive delay or jitter can cause VoIP issue too, but switches generally don't queue deep enough, as LAN bandwidths, to usually cause an issue.

[e_dunnington]

Ok,

So those being on 2 different vlan's would not in most cases cause problems.

I am just making sure I understand properly.

[Joseph W. Doherty]

Generally not on the same switch.

If your VLANs are on different switches, than inter-switch connection might be a point of congestion.

Although in your case, you route between VLANs using the 2821? If so, the 2821 is a WAN router, suitable for under 100 Mbps total traffic. Inter-VLAN traffic might be bottlenecked by it.