Qos-Confusion for Video and Best effort traffic

[anilkumar.cisco]

Hello Friends,

with regard to the Qos.. I am little bit confused with below things..

1. what is difference between video media and video signalling..

I know how to categarize the voice traffice and media in Qos as follows:

ip access-list extended voice-signaling

permit tcp X.X.X.0    0.0.0.255 any range 2000 2002

permit tcp X.X.X.0    0.0.0.255 any eq 5060

permit udp X.X.X.0    0.0.0.255 any eq 5060

and to categorize Voice media  we used the below access list.

ip access-list extended voice-bearer

permit udp X.X.X.0   0.0.0.255 any range 16384 32767

Now , My question,

How I will categarize the video media traffic..

is the below class map  is good to use for video media or video signaling 

class-map match-all video-traffic

match access-group name voice-bearer

match ip dscp af41

Also, my video traffice is of jabber and normal Cisco 9971 IP phone traffice, so it will be categorize as 

a)  Interactive Video (general guideline\standard from Cisco to mark voice with a dscp 46 or ef, Interactive video 34 or af41)

or

b) Streaming Video (Streaming video as dscp 32 or cs4)

so , I am confused with marking of the traffic..

and also,

what is difference between marking of data traffice to best effort versus marking it default unconditionally..

Kindly advise.

Best Regards

Anil Singh 

[Wayne DeNardi]

1. "Video Media" is the actual video packets of a TelePresence equipment.  The "Signalling" is all the other talk the endpoints need to do to work out how the call progresses.

2. The "Video" traffic from a TelePresence endpoint should be considered as "Interactive Video", not as "Streaming Video".

For all the in-depth information, I suggest you have a read through the Enterprise QoS Solution Reference Network Design Guide.

[Francesco Molino]

Hi

For Cisco devices, there are 3 DSCP values used to take into consideration:

- voice signaling --> CS3

- voice traffic --> EF

- video traffic --> AF41

Then your class-map for video is good as you're using match-all to match the acl and the dscp.

Everything will work correctly except Jabber on laptop. Jabber application don't have default dscp marking and won't match your video class-map.

Here a Cisco doc for jabber qos:

https://www.cisco.com/c/en/us/td/docs/voice_ip_comm/jabber/11_5/CJAB_BK_D00D8CBD_00_deployment-installation-guide-cisco-jabber115/CJAB_BK_D00D8CBD_00_deployment-installation-guide-cisco-jabber115_chapter_010001.html

For normal traffic, if you don't want to do anything with it, it will be catched by class default and no dscp marking. It will keep dscp 0. 

This is the way we handle normal/usual traffic.

Have i answered your questions?

[anilkumar.cisco]

thanks      Francesco Molino  for clearing my doubt..

So you want to say.. whether I will do classification and marking of video traffic in my service policy and applied to the voice port where both phone and PC (Jabber) connected..

PC will still send Jabber traffic without any marking thus my switch port would not be able to do any treatment and would be consider it as normal data traffic, correct??

And Still I am confused with the statement which state allow 10mbs for video signaling per voice port??

how I can allow 10mbps for video signaling , ideally it should be video media..

Also, where we are using AF31.. and how we can segregate video signaling via voice signaling and where we are using H.323 in between.

Kindly advise.

[Francesco Molino]

To answer your question about jabber.

The traffic coming from your computer with jabber application won't match your video class-map because it won't be tagged.

Now, if your voice class-map had only an acl match with rtp ports, the jabber will match that class-map. When you're making call from Jabber it'll user same udp ports as your IP Phone then all traffic from Jabber will be classified in your voice class-map and tagged as ef no matter if it's a voice or video call. This is by default if nothing is changed in CUCM and on Windows QoS policy. 

There's no video signaling but i think they're talking about video calls. To limit the bandwidth to 10Mbps, you've to use the command police under your output policy-map for video class.

Is that more clear?

[anilkumar.cisco]

Thanks for your reply and below is my understanding.

So by default all jabber traffic will be treated as voice traffic only..

and it will be treated as per the voice treatment in my policy map.

So if my question is asking to allow 10 MBPS for video signaling (means media) .. then under this only video call from 9971 IP Phone will categorize , not jabber..

and if I want to do it for jabber then I have to do some setting changes as suggested by you in both PC and CUCM.

What will happen if jabber in PC is replaced by IP Communicater, the behavior will be remain same or do I need to change some things.

I am also categorizing voice traffic from data vlan.. how it will help me here, or it don't require.

and at last what is the use of AF31

[Francesco Molino · Accepted solution]

No, by default jabber traffic is not tagged and will be like any usual traffic.

I'm saying it will be treated and tagged ef like voice if you have a simple class-map voice with a simple match on acl based on udp rtp ports. If your voice class-map is like your video with match acl statement and match dscp ef statement with a class-map match-all then jabber traffic will go the default class like normal user traffic.

There're lot of Cisco documentation, how to configure QoS for jabber on windows machine and/or over wifi on laptop or smartphone. The last link is talking about cucm changes and windows changes. 

For ip communicator, you can't do video. For voice traffic it'll be treated like jabber on the switch side.

You're categorizing voice traffic from data vlan, this is exactly the idea of jabber app running on laptop data vlan and needing voice priority

By default, there's no Cisco default AF34 tagging. Voice is EF, Video is AF41, Signaling is CS3, Telepresence is CS4

[anilkumar.cisco]

thanks , Superb.. +5 star

You made my day today..

Hope my below  understanding is correct

If on the 3750  switch, I am doing mls qos map cos-dscp 0 8 16 24 34 46 48 (globally)

and when video traffic related to the 9971 would be classified on would match my class map(mention below ) and take the treatment as per Policy map configured (shared earlier on the post) Correct..

Because I have configured match-all class map and matching both RTP (port) and considering Video marking AF41 as well

class-map match-all video-traffic

match access-group name voice-bearer

match ip dscp af41

Also, Do you know where we will be H.323 signaling (port 1720) and what is the use of  

Also it is good practice to consider voice or video traffic generating from Data port and do the treatment as have done by me, apart from voice vlan.

Kindly advise.

[R0g22]

Why not configure MS QoS Policy group locally on the machine to mark the jabber traffic or use medianet metadata ?
For H.323, H225 can be classified easily but H245 cannot since it is ephemeral so cannot use "match protocol h323" or an ACL to classify/mark-remark that traffic.

[anilkumar.cisco]

Thanks Nipun for your reply.. My requirement is very clear.. All Voice and Video traffic management polices to be enable on voice port with IP Phone attached.

there are two type of switch available on the customer site..

the phone connected to Cisco 3750 switch (one site ) and the phone connected to Ether Module of 2900/2800 series router..(rest two sites).. all these site are interconnected with frame relay network. 

phone models are 7965 and 9971.

In voice port the PC also connected with Jabber client

So I have advised the customer that end to end Qos is not possible because Ether Module of 2900 and 2800 switch would not support Service Policy configuration in inside interface of EtherModule..

and Qos will only means when we are doing end to end.

while segregating traffic on the site where 3750 switch are their, I have consider both voice and video traffic from voice vlan and data vlan.. and because jabber is not marking its traffic as AF41, thus video traffic of jabber from data vlan will be treated as voice traffic and I am able to meet the customer requirement as he want to allow only 16kbps of voice signaling per voice port.

Hope my understanding is correct and I would be great if you can make me double sure that understanding of my previous post is correct.

I am not sure whether we need to think about H323 protocol or not.. but these sites are interconnected via Cube only.

[R0g22]

1. Ethermodule ports are L2 only so yes you cannot apply service-policy but you can apply service-policy to the VLAN interface. Create a class-map(s) to match the source based on the phone MAC, add policy-map as desired and then apply inbound service-policy on the VLAN interface.

2. Like everyone said, jabber will not mark traffic. If you want to, then you can create MS QoS Policy groups locally on the machines running Jabber or you can look into medianet. Jabber for Windows supports it.

[anilkumar.cisco]

if this is the case then we need to do  end to end qos configuration..

means on Cisco 3750 switch then on all three router  out and inside interface which interconnecting these site, including vlan of Ether switch and Cube..

Hope My understanding is correct.

Best Regards

Anil Singh

[R0g22]

Absolutely.

[anilkumar.cisco]

Can you pls share the sample template to achive the below requirement considering IP Phone and PC connected to the ether module ports..

you said "1. Ethermodule ports are L2 only so yes you cannot apply service-policy but you can apply service-policy to the VLAN interface. Create a class-map(s) to match the source based on the phone MAC, add policy-map as desired and then apply inbound service-policy on the VLAN interface. "
thanks for your help

[R0g22]

Sure. It would be something like -

1. Create a class-map to match on based on mac address -

class-map match-all PHONE1
match source-address mac AAAA.BBBB.CCCC

2. Create a policy-map to enforce a policy/treatment -

policy-map POLICING
class PHONE1
police cir 512000 bc 8000
conform-action <action>
exceed-action <action>
violate-action <action>

3. Assign the policy to let's say vlan 100 -

interface Vlan100
service-policy input POLICING

You can tweak the above to let's say classify/mark/re-mark the traffic. I just used a policing example.