SIP Trunk, need recommendation on QoS setup

Unanswered Question

I have a UC520 with a SIP Trunk and a Cable internet service (16mbps down, 2-3mbps up). I have noticed that at times users are downloading files from fast webservers, it will spike our bandwidth. This leads to poor quality voice calls, or voice calls not working alltogether.

What is the recommended method for configuring a UC520 to provide QoS or traffic shaping/policing/throttling and support a SIP trunk? Ideally I would want dedicated bandwidth for the voice traffic (SIP, RTP, RTCP, etc.) then the leftover would be the remaining internet traffic."traffic shaping" seems to be pretty simple, but I have also seen that it is not recommended for voice traffic. Should I be using LLQ?

Are there any sample configurations out there for this type of setup?

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
Maulik Shah Tue, 05/05/2009 - 13:57

You can enable shaping / queueing on the UC520 WAN interface (CLI is below) - this will prioritize the SIP / RTP packets out the WAN interface i.e on the upstream side. However it still does not guarantee the QOS for packets coming in which is where the congestion is from your description - there is not much you can do in terms of downstream as the QOS needs to be applied on the cable provider's side. That is why doing SIP trunks "over the top" is always a challenge - unless you convince your cable provider to ensure SIP Trunk traffic gets priority over other traffic on the downstream.

CLI config for doing QoS:

class-map match-any sip
match ip dscp cs3
match protocol sip

!
class-map match-any rtp
match ip dscp ef
match protocol rtp audio
!
policy-map queue
class rtp
    priority percent 30
class sip
    bandwidth percent 5
class class-default
    fair-queue

!

policy-map shape
class class-default
    shape average 16000000
  service-policy queue

!

interface FastEthernet0/0
service-policy output queue

Thanks, good information. I was also concerned about the upstream vs. downstream config.

This may be a dumb question, but what if i were to restrict traffic going OUT of my INTERNAL interface, this restricting the speed at which internal clients could download from the internet. Would this helpful or even possible?

Maulik Shah Tue, 05/05/2009 - 22:52

From your last post:

PC --- LAN -- UC520 -- WAN -- Internet

You want to limit the traffic going from UC520 to PC on the LAN - yes? If so that would not help a whole lot as the download traffic still comes in on the WAN pipe and causes congestion.Options to consider:

- limit specific download traffic from PC to UC520 at UC520 VLAN interface using Class based policing - may not be that useful but if used in conjunction with WRED may help with congestion avoidance if your download traffic is TCP based

Class based Policing http://www.cisco.com/en/US/docs/ios/qos/configuration/guide/cnfig_traffic_polcng_ps6350_TSD_Products_Configuration_Guide_Chapter.html

WRED - http://www.cisco.com/en/US/docs/ios/qos/configuration/guide/congestion_avoidance_ps6350_TSD_Products_Configuration_Guide_Chapter.html

- limit specific traffic from Internet to UC520 using Class Based Policing on the WAN interface - its not that useful as the WAN pipe is already congested and you are dropping at the destination rather than the source.

- Look at time of day QoS and force download type traffic to be after hours if possible

http://www.cisco.com/en/US/tech/tk543/tk759/technologies_tech_note09186a00801aa69d.shtml

The best option is really to get your provider to prioritize SIP Trunk traffic over data

Marcos Hernandez Wed, 05/06/2009 - 10:41

To verify DSCP settings, check the dial peer stats using "sh dial-peer voice XXX | inc ef", where XXX is your outbound SIP dial peer (typically in the 1000's)

You don't see matches because there has to be congestion for traffic shaping to kick in. As a quick test, you could force some policing under the "shape" statement, something like:

policy-map all-traffic
  class class-default
     shape average 60000000
       police 40000000 64000 128000
      service-policy voice

Let us know,

Marcos

I don't see anything about the DSCP settings. i just see qos 'best effort', is that it?

MCC-UC520#sh dial-peer voice 1010 | inc ef
        peer type = voice, system default peer = FALSE, information type = voice,
        answer-address = `', preference=0,
        technology prefix:
        req-qos = best-effort, acc-qos = best-effort,
        req-qos video = best-effort, acc-qos video = best-effort,
        req-qos audio def bandwidth = 64, req-qos audio max bandwidth = 0,
        req-qos video def bandwidth = 384, req-qos video max bandwidth = 0,
        fax NSF = 0xAD0051 (default)
        Playout-delay Minimum mode is set to default, value 40 ms
        Accepted Calls = 0, Refused Calls = 0,
MCC-UC520#sh dial-peer voice 1011 | inc ef
        peer type = voice, system default peer = FALSE, information type = voice,
        answer-address = `', preference=0,
        technology prefix:
        req-qos = best-effort, acc-qos = best-effort,
        req-qos video = best-effort, acc-qos video = best-effort,
        req-qos audio def bandwidth = 64, req-qos audio max bandwidth = 0,
        req-qos video def bandwidth = 384, req-qos video max bandwidth = 0,
        fax NSF = 0xAD0051 (default)
        Playout-delay Minimum mode is set to default, value 40 ms
        Accepted Calls = 0, Refused Calls = 0,

Maulik Shah Wed, 05/06/2009 - 13:43

You should see the settings as below:

uc520#sh dial-peer voice 2000
...

       ip media DSCP = ef, ip media rsvp-pass DSCP = ef <<<<<<<<<<<<<<<<<<<<
        ip media rsvp-fail DSCP = ef, ip signaling DSCP = af31,
        ip video rsvp-none DSCP = af41,ip video rsvp-pass DSCP = af41
        ip video rsvp-fail DSCP = af41,
        ip defending Priority = 0, ip preemption priority = 0

...


The default setting is EF. If you want to confirm the DSCP match is occuring - remove the "match protocol rtp" on the class-map, only keep "match ip dscp ef".

Eivind Jonassen Fri, 11/20/2009 - 03:22

Hi guys,

I'm also looking for a solution for QoS. For upstream I've used the CCA setup for QoS on the WAN interface, but what I'm struggling with is QoS for downstream. I've been in contact with the ISP and they only have solution for QoS on RTP ports 30000 to 300030. This is a bit difficult to do on the uc500 since the port range for RTP streams differs from 168xx to 32xxx

Any recommendation on how to apply a QoS feature that the ISP can use in their network??? Is there any proven/good methods?

Thanks,

Eivind

Maulik Shah Fri, 11/20/2009 - 10:23

Is the ISP the same as your SIP trunk provider - I take it its not? If they were the same the SIP Trunk provider could limit the source RTP ports to the range that they mark.

If not then your options are:

- check if ISP can mark based on DSCP markings - RTP packets are generally EF

- check if ISP can mark based on specific IP addresses (UC500 WAN IP if its static or SIP Trunk provider IP addresses)

FYI downstream QoS on the UC500 is not applicable as the congestion, if any will be at the edge router on the ISP side.

Actions

This Discussion