Mailing List Archive: Cisco: NSP

Basic QoS on ATM subinterfaces

Index | Next | Previous | View Threaded

djweis at internetsolver

Nov 24, 2009, 6:05 AM

Post #1 of 6 (802 views)
Permalink

Basic QoS on ATM subinterfaces

Hello All

I've got a PA-A3-OC3 that is terminating a large number of PPPoA
connections. I need to do basic QoS/prioritization for voice traffic. I
am using a subinterface per VPI with a vc-class to reference the
virtual-template.

I have set up a parent/child policy-map as the documentation suggested
but trying to apply it doesn't work:

router(config)#int atm4/0
router(config-if)#service-policy output VOICE-PARENT
GTS : Not supported on this interface

Most of the examples I saw that did work involved one or two interfaces
of a known speed. The customers that are connecting to this vary from
256k to 30 megabits. I don't want to individually configure any portion
of the solution per customer.

I do have control of the radius configuration and can insert attributes
that contain the speed that each customer is provisioned for if that
would help.

Any ideas? Below is what I ended up with.

class-map match-all EVERYTHING
match access-group name EVERYTHING
class-map match-all IS-VOICE
match access-group name IS-VOICE
!
policy-map IS-VOICE
class IS-VOICE
priority percent 75
set dscp ef
class EVERYTHING
set dscp default
policy-map VOICE-PARENT
class class-default
shape average percent 100
service-policy IS-VOICE
!
ip access-list standard EVERYTHING
permit any
!
ip access-list extended IS-VOICE
permit ip 192.168.221.0 0.0.0.63 any

Thanks!
dave

--
Dave Weis
Internet Solver
Your Technology Partner
515-224-9229
www.internetsolver.com
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

tim at pelican

Nov 24, 2009, 7:14 AM

Post #2 of 6 (756 views)
Permalink

Re: Basic QoS on ATM subinterfaces [In reply to]

Hi Dave,

> I've got a PA-A3-OC3 that is terminating a large number of PPPoA
> connections. I need to do basic QoS/prioritization for voice traffic.
> I
> am using a subinterface per VPI with a vc-class to reference the
> virtual-template.
>
> I have set up a parent/child policy-map as the documentation suggested
>
> but trying to apply it doesn't work:
>
> router(config)#int atm4/0
> router(config-if)#service-policy output VOICE-PARENT
> GTS : Not supported on this interface

No, this won't work. You've got several places you can apply the template:

-On the sub-interface
-On the PVC, with the outer shaper removed
-On the virtual-access (via the virtual-template)

If you're bulk-terminating a bunch of PPPoA sessions, I'd suggest that you want it applied to the virtual-access interface. You can do this by either applying it to the virtual-template (if you're sure you always want the same policy for all the users), or push it back from RADIUS as a Cisco-avpair as each virtual-access interface is cloned.

Regards,
Tim.
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

djweis at internetsolver

Nov 24, 2009, 7:45 AM

Post #3 of 6 (757 views)
Permalink

Re: Basic QoS on ATM subinterfaces [In reply to]

Tim Franklin wrote:
>> I've got a PA-A3-OC3 that is terminating a large number of PPPoA
>> connections. I need to do basic QoS/prioritization for voice traffic.
>> I
>> am using a subinterface per VPI with a vc-class to reference the
>> virtual-template.
>>
>> I have set up a parent/child policy-map as the documentation suggested
>> but trying to apply it doesn't work:
>>
>> router(config)#int atm4/0
>> router(config-if)#service-policy output VOICE-PARENT
>> GTS : Not supported on this interface
>
> No, this won't work. You've got several places you can apply the template:
>
> -On the sub-interface
> -On the PVC, with the outer shaper removed
> -On the virtual-access (via the virtual-template)
>
> If you're bulk-terminating a bunch of PPPoA sessions, I'd suggest that you want it applied to the virtual-access interface. You can do this by either applying it to the virtual-template (if you're sure you always want the same policy for all the users), or push it back from RADIUS as a Cisco-avpair as each virtual-access interface is cloned.

OK, something like this:

class-map match-all EVERYTHING
match access-group name EVERYTHING
class-map match-all IS-VOICE
match access-group name IS-VOICE
!
!
policy-map IS-VOICE
class IS-VOICE
priority percent 75
set dscp ef
class EVERYTHING
set dscp default

vc-class atm pppoa-1
encapsulation aal5mux ppp Virtual-Template1

interface Virtual-Template1
ip unnumbered Loopback0
ip accounting output-packets
no logging event link-status
peer default ip address pool adsl1
ppp authentication pap chap radius-ppp
ppp authorization radius-ppp
ppp link reorders
ppp multilink
ppp multilink fragment disable
service-policy output IS-VOICE

ip access-list standard EVERYTHING
permit any
!
ip access-list extended IS-VOICE
permit ip 192.168.221.0 0.0.0.63 any

I have applied this configuration but the only interfaces that show up
in show queueing are MLP bundles. The PVC's that show up after that
section all list the queueing as FIFO still:

router#show queueing
Current fair queue configuration:

Interface Discard Dynamic Reserved Link Priority
threshold queues queues queues queues
Virtual-Access180 64 256 256 8 1
Virtual-Access207 64 256 256 8 1
Virtual-Access450 64 256 256 8 1
Virtual-Access541 64 256 256 8 1
Virtual-Access573 64 256 256 8 1
Virtual-Access574 64 256 256 8 1
Virtual-Access575 64 256 256 8 1
Virtual-Access595 64 256 256 8 1
Virtual-Access597 64 256 256 8 1
Virtual-Access599 64 256 256 8 1
Virtual-Access640 64 256 256 8 1
Virtual-Access651 64 256 256 8 1
Virtual-Access654 64 256 256 8 1

Current DLCI priority queue configuration:
Current priority queue configuration:

List Queue Args
Current custom queue configuration:

VC 15/155 -
VC 15/155: Per VC queueing is FIFO.
VC 14/99 -
VC 14/99: Per VC queueing is FIFO.
VC 13/43 -
VC 13/43: Per VC queueing is FIFO.
VC 11/187 -
VC 11/187: Per VC queueing is FIFO.
VC 10/531 -
VC 10/531: Per VC queueing is FIFO.
VC 10/275 -
VC 10/275: Per VC queueing is FIFO.
VC 15/156 -
VC 15/156: Per VC queueing is FIFO.

Have I missed something else?

Thanks
dave

--
Dave Weis
515-224-9229
djweis [at] internetsolver
http://www.internetsolver.com/
Please check out our Complete Support Service
http://www.internetsolver.com/completesupport/
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

b.turnbow at twt

Nov 24, 2009, 8:30 AM

Post #4 of 6 (755 views)
Permalink

Re: Basic QoS on ATM subinterfaces [In reply to]

You can't do it with ubr/ubr+ interfaces ,you need to set a different class of service.
Here is an example technote
http://www.cisco.com/en/US/tech/tk39/tk824/technologies_configuration_example09186a0080094cf6.shtml

Brian

-----Original Message-----
From: cisco-nsp-bounces [at] puck [mailto:cisco-nsp-bounces [at] puck] On Behalf Of Dave Weis
Sent: marted� 24 novembre 2009 16.45
To: Tim Franklin
Cc: cisco-nsp [at] puck
Subject: Re: [c-nsp] Basic QoS on ATM subinterfaces

Tim Franklin wrote:
>> I've got a PA-A3-OC3 that is terminating a large number of PPPoA
>> connections. I need to do basic QoS/prioritization for voice traffic.
>> I
>> am using a subinterface per VPI with a vc-class to reference the
>> virtual-template.
>>
>> I have set up a parent/child policy-map as the documentation suggested
>> but trying to apply it doesn't work:
>>
>> router(config)#int atm4/0
>> router(config-if)#service-policy output VOICE-PARENT
>> GTS : Not supported on this interface
>
> No, this won't work. You've got several places you can apply the template:
>
> -On the sub-interface
> -On the PVC, with the outer shaper removed
> -On the virtual-access (via the virtual-template)
>
> If you're bulk-terminating a bunch of PPPoA sessions, I'd suggest that you want it applied to the virtual-access interface. You can do this by either applying it to the virtual-template (if you're sure you always want the same policy for all the users), or push it back from RADIUS as a Cisco-avpair as each virtual-access interface is cloned.

OK, something like this:

class-map match-all EVERYTHING
match access-group name EVERYTHING
class-map match-all IS-VOICE
match access-group name IS-VOICE
!
!
policy-map IS-VOICE
class IS-VOICE
priority percent 75
set dscp ef
class EVERYTHING
set dscp default

vc-class atm pppoa-1
encapsulation aal5mux ppp Virtual-Template1

interface Virtual-Template1
ip unnumbered Loopback0
ip accounting output-packets
no logging event link-status
peer default ip address pool adsl1
ppp authentication pap chap radius-ppp
ppp authorization radius-ppp
ppp link reorders
ppp multilink
ppp multilink fragment disable
service-policy output IS-VOICE

ip access-list standard EVERYTHING
permit any
!
ip access-list extended IS-VOICE
permit ip 192.168.221.0 0.0.0.63 any

I have applied this configuration but the only interfaces that show up
in show queueing are MLP bundles. The PVC's that show up after that
section all list the queueing as FIFO still:

router#show queueing
Current fair queue configuration:

Interface Discard Dynamic Reserved Link Priority
threshold queues queues queues queues
Virtual-Access180 64 256 256 8 1
Virtual-Access207 64 256 256 8 1
Virtual-Access450 64 256 256 8 1
Virtual-Access541 64 256 256 8 1
Virtual-Access573 64 256 256 8 1
Virtual-Access574 64 256 256 8 1
Virtual-Access575 64 256 256 8 1
Virtual-Access595 64 256 256 8 1
Virtual-Access597 64 256 256 8 1
Virtual-Access599 64 256 256 8 1
Virtual-Access640 64 256 256 8 1
Virtual-Access651 64 256 256 8 1
Virtual-Access654 64 256 256 8 1

Current DLCI priority queue configuration:
Current priority queue configuration:

List Queue Args
Current custom queue configuration:

VC 15/155 -
VC 15/155: Per VC queueing is FIFO.
VC 14/99 -
VC 14/99: Per VC queueing is FIFO.
VC 13/43 -
VC 13/43: Per VC queueing is FIFO.
VC 11/187 -
VC 11/187: Per VC queueing is FIFO.
VC 10/531 -
VC 10/531: Per VC queueing is FIFO.
VC 10/275 -
VC 10/275: Per VC queueing is FIFO.
VC 15/156 -
VC 15/156: Per VC queueing is FIFO.

Have I missed something else?

Thanks
dave

--
Dave Weis
515-224-9229
djweis [at] internetsolver
http://www.internetsolver.com/
Please check out our Complete Support Service
http://www.internetsolver.com/completesupport/
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

tim at pelican

Nov 24, 2009, 8:49 AM

Post #5 of 6 (751 views)
Permalink

Re: Basic QoS on ATM subinterfaces [In reply to]

Hi Dave,

> interface Virtual-Template1
> ip unnumbered Loopback0
> ip accounting output-packets
> no logging event link-status
> peer default ip address pool adsl1
> ppp authentication pap chap radius-ppp
> ppp authorization radius-ppp
> ppp link reorders
> ppp multilink
> ppp multilink fragment disable
> service-policy output IS-VOICE

That looks OK to me.

> in show queueing are MLP bundles. The PVC's that show up after that
> section all list the queueing as FIFO still:

To be honest, I've never looked at that. Does 'show policy-map interface virtual-accessNNN' give you the right policy applied, and drops happening in the right class at the right times?

(I have a feeling the underlying interface still says FIFO as it is just passing on the PPPoA packets after the virtual-access has decided which packets to encapsulate in which order, but *don't* take that as gospel).

I'd also second another poster's point that you'll want the individual VCs configured as something like vbr-nrt, not ubr, or you won't ever get the back-pressure to the service-policy.

Regards,
Tim.
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

djweis at internetsolver

Nov 24, 2009, 9:29 AM

Post #6 of 6 (755 views)
Permalink

Re: Basic QoS on ATM subinterfaces [In reply to]

Hello Tim

Tim Franklin wrote:
>> interface Virtual-Template1
>> ip unnumbered Loopback0
>> ip accounting output-packets
>> no logging event link-status
>> peer default ip address pool adsl1
>> ppp authentication pap chap radius-ppp
>> ppp authorization radius-ppp
>> ppp link reorders
>> ppp multilink
>> ppp multilink fragment disable
>> service-policy output IS-VOICE
>
> That looks OK to me.
>
>> in show queueing are MLP bundles. The PVC's that show up after that
>> section all list the queueing as FIFO still:
>
> To be honest, I've never looked at that. Does 'show policy-map interface virtual-accessNNN' give you the right policy applied, and drops happening in the right class at the right times?

If I run it against an MLP bundle with a voice device behind it, I do
see what I would expect:

router#show policy-map int Vi573
Virtual-Access573

Service-policy output: IS-VOICE

Class-map: IS-VOICE (match-all)
108729 packets, 22455543 bytes
5 minute offered rate 0 bps, drop rate 0 bps
Match: access-group name IS-VOICE
Queueing
Strict Priority
Output Queue: Conversation 264
Bandwidth 75 (%)
Bandwidth 224640 (kbps) Burst 5616000 (Bytes)
(pkts matched/bytes matched) 0/0
(total drops/bytes drops) 0/0
QoS Set
dscp ef
Packets marked 108729

Class-map: EVERYTHING (match-all)
332750 packets, 429378515 bytes
5 minute offered rate 38000 bps, drop rate 0 bps
Match: access-group name EVERYTHING
QoS Set
dscp default
Packets marked 332750

Class-map: class-default (match-any)
0 packets, 0 bytes
5 minute offered rate 0 bps, drop rate 0 bps
Match: any

If I run it on a non-MLP virtual-access interface I get no output at all
from the show policy-map command.

> (I have a feeling the underlying interface still says FIFO as it is just passing on the PPPoA packets after the virtual-access has decided which packets to encapsulate in which order, but *don't* take that as gospel).

Understood. I did find another document that I'm going to try and adapt:
http://www.cisco.com/en/US/tech/tk543/tk544/technologies_tech_note09186a0080094ad2.shtml

> I'd also second another poster's point that you'll want the individual VCs configured as something like vbr-nrt, not ubr, or you won't ever get the back-pressure to the service-policy.

Can you configure a VC as vbr-nrt without explicitly putting the rates?
All of the PVC's are varying speeds with no explicit configuration for
any one PVC.

--
Dave Weis
515-224-9229
djweis [at] internetsolver
http://www.internetsolver.com/
Please check out our Complete Support Service
http://www.internetsolver.com/completesupport/
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

Index | Next | Previous | View Threaded

Interested in having your list archived? Contact Gossamer Threads