QoS nested policy map

[tedauction]

Hello, if I had the following policy-map on a WAN uplink port, how does the port know what the total capacity of the WAN uplink is, in order to divide it in to percentages ?

I understand you can make this a nested policy and use a 'shape' command to define the actual bandwidth of your WAN uplink. However I have seen many configs like this were there is no nested shape command.

policy-map QUEUES

class Multi-Media

priority

class IA-Low

bandwidth remaining percent 10

class DataTrans

bandwidth remaining percent 60

class IA-High

bandwidth remaining percent 15

class class-default

bandwidth remaining percent 15

[Joseph W. Doherty]

When applied directly to a physical interface, it will use what it believes is the bandwidth of the interface.  For example if applied to a triple speed Ethernet interface running at 100 Mbps, it would use 100 Mbps.  If applied to a serial fractional T1, running at 512 Kbps, it will use 512 Kbps.  Also the interface bandwidth statement will, I recall, preempt what QoS will use for its bandwidth.

[tedauction]

Thank you Joseph. What happens if the WAN interface is negotiating at 100Mbps but the service provider has only actually provisioned you with 30Mbps.

Will the policy map above be useless without a nested 'shape 30000' policy-map or interface bandwidth statement ?

i.e is this how it should always be (or equally valid, don't use the 'policy-map WAN' but instead put the command 'bandwidth 3000' on the output interface.) ?

policy-map QUEUES
class Multi-Media
priority
class IA-High
bandwidth remaining percent 15
class IA-Low
bandwidth remaining percent 10
class DataTrans
bandwidth remaining percent 60
class class-default
bandwidth remaining percent 15

policy-map WAN
class class-default
shape peak 30000000
service-policy QUEUES

[Joseph W. Doherty · Accepted solution]

Will the policy map above be useless without a nested 'shape 30000' policy-map or interface bandwidth statement ?

In a case like this, w/o a shaper, you would send at full physical rate, so your queuing would only come into play when there was interface congestion (at 100 Mbps), even if you set the bandwidth statement 30 Mbps.

Except for shapers and policers, QoS bandwidth statements really only determine bandwidth allowance ratios between classes.

I.e. generally you would be the same result for:

class a
 bandwidth 8000
class b
 bandwidth 8000

class a
 bandwidth 8000000
class b
 bandwidth 8000000

class a
 bandwidth percent 1
class b
 bandwidth percent 1

class a
 bandwidth percent 50
class b
 bandwidth percent 50

class a
 bandwidth remaining percent 15
class b
 bandwidth remaining percent 15

Classes A and B get an equal share of the bandwidth.