cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
768
Views
0
Helpful
1
Replies

QoS and Nesting

NETAD
Level 4
Level 4

Hello, can someone review and explain this QoS policy for me especially the nesting part of it and why do we do it this way. I'm seeing this configuration very common. Also is 60% a lot of bandwidth for voice on a 2GB circuit? 

 

!
!
class-map match-any STREAMING-VIDEO
match dscp cs5
match dscp af31
match dscp af32
match dscp af33
class-map match-any INTERACTIVE-VIDEO
match dscp cs4
match dscp af41
match dscp af42
match dscp af43
!
class-map match-any CRITICAL-APPS
description match Business Critical Apps in access-list CRITICAL-APPS
match dscp cs2
match dscp af11
match dscp af12
match dscp af13
match dscp af21
match dscp af22
match dscp af23
!
class-map match-any VOICE-CONTROL
description matches cs3 or af31, or access-list VOICE-CONTROL
match ip dscp cs3 af31
!
class-map match-all VOICE-MEDIA
description matches from 16384-32767, marked EF or access-list VOICE-MEDIA
match ip dscp ef
match ip rtp 16384 16363
!
policy-map QoS
description Gives voice 60%, signaling 15%, 10% critical apps, and queues the rest
!
class INTERACTIVE-VIDEO
bandwidth remaining percent 30
random-detect dscp-based
set dscp tunnel af41
!
class STREAMING-VIDEO
bandwidth remaining percent 10
random-detect dscp-based
set dscp tunnel af31
!
class VOICE-MEDIA
priority percent 60
set dscp ef
!
class VOICE-CONTROL
bandwidth percent 15
set dscp cs3
queue-limit 256 packets
!
class CRITICAL-APPS
bandwidth percent 10
random-detect
class class-default
queue-limit 512 packets
!
policy-map Flagler_SHAPE
description Shapes the Ethernet circuit to 6GB, and applies voice QoS to the 6GB
class class-default
shape average 2000000000
service-policy QoS
!
!
!
!
interface Te0/0/0
bandwidth 2000000
service-policy output SHAPE
!
!
!

1 Reply 1

Joseph W. Doherty
Hall of Fame
Hall of Fame
Nesting, with a hiearchal policy, where the parent uses a shaper, is commonly used where the physical port bandwidth is higher than some later downstream point where you have no control over the QoS.

For example, given
rtr A <gig> rtr B <FE> rtr C

If sending traffic from rtr A to rtr C, the FE between rtrs B and C will congest if traffic from rtr A to rtr B exceeds 100 Mbps. Assuming we don't control QoS policy on rtr B, we can effectively control it by shaping rtr A's output to 100 Mbps.

"Also is 60% a lot of bandwidth for voice on a 2GB circuit?"

Probably, but if so, it's more about it being 60% rather than 2 Gbps. Cisco recommends to allocate no more than 1/3 of the bandwidth to LLQ.