Quality of Service Configuration Guidelines, Cisco IOS XE 16 (Cisco ASR 920 Series)

Quality of Service

QoS refers to the ability of a network to provide improved service to selected network traffic over various underlying technologies including ATM, Ethernet and 802.1 networks, SONET, and IP-routed networks. In particular, QoS features provide improved and more predictable network service by implementing the following services:

Supporting guaranteed bandwidth
Improving loss characteristics
Avoiding and managing network congestion
Shaping network traffic
Setting traffic priorities across the network

For more information about Quality of Service, see http://www.cisco.com/c/en/us/support/routers/asr-920-series-aggregation-services-router/products-installation-and-configuration-guides-list.html.

Quality of Service Configuration

This document provides details on the platform-dependent implementation of QoS on the Cisco ASR 920 Series Router.

For information about how to understand and configure QoS features, see http://www.cisco.com/c/en/us/support/routers/asr-920-series-aggregation-services-router/products-installation-and-configuration-guides-list.html.

QoS Support Overview

Table below provides an overview of QoS feature support on the Cisco ASR 920 Series Router. For more detail about the support for each feature, see Global QoS Limitations.

Table 1. QoS Feature Overview
Feature	Main	Service Instance	Trunk EFP	Member Link
Dynamic policy modification	3.6	3.6	3.6	3.6
EFP QoS Support	3.6	3.6	3.6
Classification
Ingress		3.5	3.5.1	3.5
Egress		3.5	3.5.1	3.5
IPv6 ¹	3.6	3.6	3.6	3.6
Match any	3.6	3.6	3.6	3.6
Marking
Ingress	3.5	3.5	3.5.1	3.5
Egress	3.6	3.6	3.6	3.6
Policing
Ingress	3.5	3.5	3.5.1	3.5
Egress	3.6	3.6	3.6	3.6
Priority policing	3.6	3.6	3.6	3.6
Shaping
Port Shaping	3.6.1	3.6.1	3.6.1
Congestion Avoidance
WRED	3.6	3.6	3.6	3.6
Multiple Priority Queues	3.7	3.7	3.7	3.7
Congestion Management
Strict Priority	3.5	3.5	3.5.1	3.5
Scheduling
Ingress
Egress	3.5	3.5	3.5.1	3.5
QoS ACLs
Ingress	3.5.1	3.5.1	3.5.1	3.5.1
Egress

¹ IPv6 based ACLs are not supported for TDM interfaces.

Table 2. QoS Support for TDM Features
Feature	Release	Interface Module	Ingress	Egress
HDLC	3.9	T1/E1	No	Yes
HDLC	3.13	OC-3	Yes	Yes
MLPPP	3.7	T1/E1	No	Yes
	3.8	OC-3	No	Yes
	3.13	T1/E1 and OC-3	Yes	Yes
POS	3.8	T1/E1	NA	NA
	3.8	OC-3	No	Yes
	3.9	OC-12	No	Yes
	3.13	OC-3	Yes	Yes
ATM QoS L2 for ATM PW	3.7	T1/E1 and OC-3	Yes	NA
QoS for CEM/ATM/IMA PW	3.9	T1/E1 and OC-3	Yes	NA

Global QoS Limitations

The following limitations apply to multiple QoS features for the Cisco ASR 920 Series Router:

Configuring 4MB buffer (in micro seconds) provides lesser buffer than 3MB. If more than 3MB of buffer is required use qlimit percent (%).
Queue-limit in percentage must be configured with caution. Oversubscribing queues and Buffering until system limits may lead to buffer exhaustion at single point of time. Also this scenario may result in punt keep-alive packet drops. Prolonged buffer exhaustion may result in system crash due to punt keep-alive failure.
When EVCs under a physical interface have a QoS policy attached, the following limitations apply:
- The port-level policy is limited to the class-default class
- Only the shape command is supported in the port-level policy
The Cisco ASR 920 Series Router supports up to 64 unique QoS classification service instances in a given bridge domain. QoS service instances refer to ports, VLAN classes, EFPs associated with a QoS classification policy.
Modification of class-map definitions while applied to an interface or Ethernet Flow Point is not supported.
Policy validation—Some QoS policy configurations are not validated until you apply the policy-map to an interface or Ethernet Flow Point. If a QoS configuration is invalid, the router rejects the configuration when you apply it to an interface. In some cases, a QoS configuration may be rejected due to hardware resource exhaustion or limitations. If you receive such an error message, detach the policy and adjust your QoS configuration.
The match-all keyword is supported only for QinQ classification.
SAToP and CESoPSN pseudowire traffic has a default MPLS Exp priority setting of 5 (high).
QoS is supported on POS interfaces on optical interface module.
Three-level QoS policies are not supported on the OC-3/OC-12 serial, MLPPP, and PoS interfaces. You can only apply QoS policies on two levels on these interfaces.
QoS does not account for CRC values on an interface and assumes that the value is 2 bytes. CRC differences can cause accuracy issues for 2 to 3% of the 128-byte traffic.
The Cisco ASR 920 Series Router supports a maximum of 128 internal and reserved labels that represent PHB (cos/dscp/exp/prec) values on a QoS policy. A label exhaustion message is displayed if a policy exceeds the maximum number of labels.
QOS does not support WRED counters for all the match conditions.
The ICMPv4 packets classification based on ACL attached on interface is not supported.
Effective with Cisco IOS XE Everest 16.5.1, if the same table-mapping is applied on multiple interfaces, the MDT index is shared across these interfaces. Thus increased scaling of table-map is possible if table-mapping is reused.
EXP based classification at the egress of PE routers is not supported by default. To achieve this, the EXP bits are required to be imposed over the cross connect at the ingress of the same PE router using an input policy.
In case of a PE router,
- For an egress policy-map to work at the access interface, it is mandatory to configure an ingress policy-map at the core interface of the same router.
- If an ingress policy-map on core port has a marking option along with a egress policy-map on the same router, then egress policy-map must be configured at the corresponding access port for the marking to be preserved.
Egress policy action at more than one level is not supported on combination of service-group and EFP.

QoS Features Using MQC Limitations

Table below lists the QoS MQC scaling limitations on router per release.

Table 3. QoS on MQC Limitations
Supported on Cisco ASR 920 Series Router	Cisco IOS XE 3.13S
No. of unique policy-maps	1024
No. of unique class-maps	4096
No. of classes per policy-map	256
No. of filters per class-map	16

Restrictions for Ingress QoS

Restrictions for Ingress QoS in the Cisco IOS Release 3.9 and later:

EC main interface
- Only policing and marking are supported.
- A class-map can have any type of filter, including the match vlan and match service instance commands.
EC EVC/TEFP
- Only policing and marking are supported.
- Match service instance is not supported.
Member links
- Only policing and marking are supported.
- Policy-map on a member link is not supported with EVC configured at the port-channel level.
Policy-map application is allowed only on the EC main interface, EC member link, or EC EVC.

Note

Ingress policer on port-channel across cylon works at twice the policer rate.

Restrictions for Egress QoS

The maximum number of classes supported on the policy map is 8, which includes class class-default; 7 user-defined classes and class class-default is supported.
The maximum number of port-channel interfaces that can be created and supported for QoS on the router is 16.

Restrictions for Egress QoS in the Cisco IOS XE Release 3.9 and later:

EC main interface
- Classification statistics for the policy-map on a port-channel main interface are not supported as no queues are allocated for a port-channel main interface.
- Policing and marking actions are only allowed in the policy-map on a port-channel main interface.
- Queuing actions are not supported.
- Egress TCAM entries are used even in the absence of member links.
EC EVC/TEFP
- Classification statistics for the policy-map on port-channel EVC/TEFP are not supported as no queues are allocated for port-channel EVC/TEFP.
- Policing and marking actions are only allowed in the policy-map on port-channel EVC/TEFP.
- Queuing actions are not supported.
- Egress TCAM entries are used even if there are no member links present.
EC Member links
- For egress match service-instance policy-map on EC member links, the same policy must be present on all other EC member links.
- Match service-instance policy-map is replicated automatically for all the member links when the first policy is applied on any of the member links.
- For non-match service-instance policy-map, the same policy-map can be applied for all member-links.
- Dynamic modification of match service-instance policy-map actions is not allowed.
- Deleting a global match service instance policy-map is not allowed if it applied to the member links.
- Policing, marking and queuing action are supported on port-channel member links.
- The running configuration displays the first member link on the first policy applied in a service-policy configuration.
- The show policy-map interface brief command only displays the policy-map applied on the running configuration.
- Applying a same policy again on other member links where a policy-map was already applied will not display any error. A differently named policy if applied again will display an error.
- For match service instance policy-map on egress member links, the policy-map statistics information is reset, when a member link is added or deleted from a port-channel either by configuration or by LACP port-bundling/unbundling action.
- There is no difference in behavior for non-match service instance policies on the member links. They continue to work in the legacy mode. There is no conservation of TCAM entries in this mode, even if the same policy is applied on all member links.
- Policy-map application is allowed only on either EC main interface, or EC member link, or EC EVC.

Routed Port-Channel

Routed Port-channel Interface

The following features are supported for the ingress policy-map on a routed port-channel interface:

Marking
Policing
Conditional marking
Marking and policing
Classification criteria is prec, dscp or ip acl.

Example for Routed Port-channel Interface


policy-map routed_pc_ingress
class prec1
set prec 2
class prec2
police cir 100m
class prec3
police cir 150m conform-action set-prec-transmit 4 exceed-action drop
class prec4
police cir 200m
set prec 0
!
end

The following features are supported for egress policy-map on routed port-channel interface:

Marking
Policing
Classification criteria is prec or dscp.

Example for egress policy-map on routed port-channel interface


policy-map pc_egress
class dscp0
set dscp 16
class dscp48
police cir 1m
!
end

Member-links on Routed Port-channel Interface

The following features are supported for ingress policy-map on member links on the routed port-channel interface:

Marking
Policing
Conditional marking
marking and policing
Classification criteria is prec, dscp or ip acl.

The following features are supported for egress policy-map on member links on the routed port-channel interface:

Shaping
Queue-limit
Bandwidth (kbps, percent)
Bandwidth remaining (ratio, percent)
WRED
Port Shaper
Low Latency Queue (LLQ or priority queue)

Example for egress policy-map on member links on the routed port-channel interface


policy-map mem_link_egress
class qos-group0
bandwidth percent 90
class qos-group67
police cir 1m
priority
class class-default
shape average 64k
!
end

Port-channel with EFP

The following features are supported for ingress policy-map on port-channel on EFP:

Marking
Policing
Conditional marking
marking and policing
The classification criteria is VLAN or EFP, Cos in child.

Example for Port-channel with EFP


policy-map cos_child
class cos0
set cos1
!
policy-map efp_pc_ingress
class vlan100
police cir 10m
service-policy cos_child
!
end

The following features are supported for egress policy-map on port-channel on EFP:

Marking
Policing
Conditional marking
marking and policing
The classification criteria is VLAN or EFP, Cos in child

Example for egress policy-map on port-channel on EFP


policy-map cos_child
class cos0
set cos1
!
policy-map efp_pc_ingress
class vlan100
police cir 10ms
service-policy cos_child
!
end

EFP of Port-channel with EFP Configuration

The following features are supported for ingress and egress policy-map on EFP of port-channel on EFP:
- Marking
- Policing
- Conditional marking
- marking and policing
- The classification criteria is VLAN or EFP, Cos in child.

Note

Match EFP is cannot be configured.

Member Links of Port-channel with EFP Configuration

The following features are supported for ingress and egress policy-map on member links of port-channel on EFP:
- Marking
- Policing
- Conditional marking
- marking and policing
- The classification criteria is VLAN or EFP, Match VLAN and Cos in child.

Restrictions for Hierarchical Policies

The Cisco ASR 920 Series Router supports hierarchical QoS policies with up to three levels, allowing for a high degree of granularity in traffic management.

There are limitations on the supported classification criteria at each level in the policy-map hierarchy. The following limitations apply when configuring hierarchical policy-map classification:

The topmost policy-map in a three-level hierarchy only supports classification using class-default.

Sample Hierarchical Policy Designs

The following are examples of supported policy-map configurations:

Three-Level Policy—You can only apply a three-level policy to a physical port on the router. A three-level policy consists of:
- Topmost policy: class-default
- Middle policy: match vlan
- Lowest policy: match qos-group/match prec/match cos/match dscp
  
  The following sample policy uses a flat class-default policy on the port and VLAN policies on EFP interfaces to unique QoS behavior to each EFP.

Sample Policy


Policy-map port-shaper
Class class-default
Shape average percent 70
Service-policy Vlan_set

Policy-map Vlan_set
Class vlan100
Bandwidth percent 20
Shape average 200m
Service-policy child1
Class vlan200_300
Bandwidth percent 75
Service-policy child2

Policy-map child1
Class prec2
Shape average percent 40

Policy-map child2
Class prec4
Police cir percent 50

Two-Level Policy
- Topmost policy: match vlan
- Lowest policy: match qos-group/match prec/match cos/match dscp
Two-Level Policy
- Topmost policy: class-default
- Lowest policy: match vlan
Two-Level Policy
- Topmost policy: class-default
- Lowest policy: match mpls experimental topmost
Flat policy: match ip dscp
Flat policy: match vlan inner
Flat policy: class-default

Ingress and Egress Hierarchical Policing

In releases before Cisco IOS XE Release 3.9, policing was supported only at one level in the ingress and egress policy. It was only at the PHB or class level.

Effective with Cisco IOS XE Release 3.9, policing is supported at two levels of the policy-map.

Ingress policing
- Port and EFP level
- EFP and Class level
- Port and Class level

Egress policing

EFP and Class level

Port and Class level

Note

Egress hierarchical policing is supported on two levels but one of the levels must be Class level.

If an Ingress hierarchal policy is configured on the interface, the show Ethernet service instance interface command does not display the service instance statistics.

The class-level in an Egress hierarchal policy is configured internally as shaper.

Dissimilar PHB Support for MPLS and VPLS Interfaces

Effective with Cisco IOS XE Release 3.11S, dissimilar per-hop behavior (PHB) match on exp is supported for Ingress and Egress on MPLS and VPLS interfaces.

In earlier releases prior to Cisco IOS XE Release 3.11S, when qos-group or discard-class based on exp classification was configured, Egress based classification was not allowed on any other classification except Ingress set qos-group or discard-class. This was due to the PHB security model.

With Cisco IOS Release 3.11, only EVC based tunnel type configuration with either Layer2 VPN or Layer3 VPN is supported.

As pipe mode and uniform mode are supported, when qos-group or discard class (pipe-mode) is matched again on the Egress interface, all qos-groups in tunnel types (such as Layer2 VPN, Layer3 VPN, and MPLS VPN ) are supported only if the tunnel type exists on the EFP. The qos-group entries in the TCAM are matched on the tunnel type. Thus, dissimilar PHB match at the egress is supported for both Ingress and Egress simultaneously on the router.

For example, the qos-group configured at Layer2 terminating Egress interface is matched against the Layer2 VPN tunnel type. This enables the dscp (uniform-mode) on the Layer3 VPN terminating Egress interface to match with the Layer3 VPN Egress interface.

Restrictions for Dissimilar PHB Support for MPLS and VPLS Interfaces

Supported for qos-group or discard-class and dscp dissimilar matches for Egress PE (VPN terminating) and not for regular EVCs.
If match discard-class policy is applied at the interface level (the policy is applied to the Layer3 interface), the match dscp policies on the other Layer3 VPN interfaces cannot be applied.

We recommend that the policy is applied at the EVC level on individual Layer2 or Layer3 interfaces instead of at port-level. Alternatively, configuring a match EFP policy to match qos-group or discard-class classification on the EFP for Layer 2 VPN, and match dscp on the EFP for Layer3 VPN is recommended.
If both Layer2 VPN and Layer3 VPN configurations exists on an interface, and the port-based policy has match qos-group or discard-class, then two match dscp classifications are not supported on the match dscp.

MPLS VPN QoS Mapping

Table below summarizes the default MPLS mappings for the Cisco ASR 920 Series Router.

Table 4. Default MPLS QoS Mapping
Feature	Imposition	Disposition
L3VPN, MPLS	IP Prec bit copied to MPLS Exp bit.	IP Prec bit is unchanged. If a VLAN tag is pushed at egress, CoS bit is set to 0.
L2VPN  (EoMPLS, VPLS)	MPLS Exp bit is set to 0	IP Prec bit is unchanged. If a VLAN tag is pushed at egress, CoS bit is set to 0.
MPLS-TP
CESoPN	MPLS Exp bit is set to 5	IP Prec bit is unchanged.
SAToP	MPLS Exp bit is set to 5	IP Prec bit is unchanged.
6VPE, 6PE	Prec bit value is copied to the MPLS Exp bit	IP Prec bit is unchanged. If a VLAN tag is pushed at egress, CoS bit is set to 0.

Note

You can modify the default mapping behaviors using explicit marking policies.

QoS Policer and Shaper Calculation

Table below summarizes the packet accounting information used to make policer and shaper calculations on the Cisco ASR 920 Series Router.

Table 5. QoS Accounting Calculation
Feature	Direction	Traffic Type	Values Counted
Policing	Ingress	IPv4/L3VPN	L2 overhead, VLAN tag, CRC
Shaping	Egress	IPv4/L3VPN	L2 Ethernet overhead, VLAN tag, CRC, preamble, IPG
Policing	Egress	IPv4/L3VPN	Layer 2 Ethernet overhead, VLAN
Policing	Ingress	L2VPN	Layer 2 Ethernet overhead, VLAN tag, CRC
Shaping	Egress	L2VPN	Layer 2 Ethernet overhead, VLAN tag, CRC, preamble, IPG
Policing	Egress	L2VPN	Layer 3 payload (without CRC)

The following considerations also apply when understanding QoS policer and shaper calculations:

Egress shaping is applied at layer 1.
Ingress packet length accounting is performed at egress.
Egress shaping and policing do not account for newly pushed VLAN tags and MPLS labels.
If two policers are configured at egress, the statistics on the child PHB or PQ level are not displayed.

Service Groups

The Service Group feature (aggregate policing) introduced in Cisco IOS XE Release 3.11 allows you create service groups, add service instances to those service groups, and apply service policies to the newly created groups. The service policies contain the aggregate features such as traffic policing that can applied to the groups.

A service group can be configured with certain match conditions and police traffic can flow via multiple targets at PHB level, EFP level or multiple ports.

The following features are supported in Cisco IOS XE Release 3.11:

Policing is only supported for service groups. Both Ingress and Egress policies are supported.
Service groups are supported on EFPs and Trunk EFPs.
Service groups are supported at both Ingress and Egress.
A service group policy can be configured as hierarchical policy of user-defined classes.
EFPs support both regular and service group policies.
At Ingress, only two level policer is allowed on the service group policy.
At Egress, only one level policer is allowed on the service group policy.

Restrictions for Service Groups

Service groups is not supported on port and port-channels.
Queuing and Marking on service group policy is not supported in Cisco IOS XE Release 3.11.
Classification based on match input vlan, match input interface, and match service instance is not supported.
Service group policies is not supported on EFPs configured on port channel.
The same EFP cannot be configured as members of multiple service groups.
If an EFP is a member of service group with a policy-map present in service group, the same policy-map cannot be applied on that EFP. The same policy-map applied on an EFP cannot be used in a service group.
Limited support for statistics counters is provided.
The no policy-map command cannot be executed for policies attached to service groups or to the policies attached to service instance which are configured as member of service groups.
Policer percentage policy-maps are not supported on service groups.
If dynamic modification is performed on the service-group policy or policy attached to EFP that is part of service-group, you have to exit of the policy-map sub-mode for the changes to take effect
Service-group can have EFP members present across ports. If these ports are present across the ASIC in the router, then aggregate policing cannot work. For aggregate policing to work correctly, ports has to be present on same ASIC.
QoS service groups for port channel sub-interface (BDI) is not supported.

Merging Service Groups and EFP Policies

If an EFP is a member of a service group without any policy configuration, and a service group policy is configured, then the service group policy is internally attached to the EFP.
If an EFP which is member of service-group has a policy configured, then the service-group policy and EFP policy are merged to form a new internal policy that is attached to the EFP.
The show policy-map interface interface_num service instance id command displays the statistics for the policy configured on the service instance. This command output deviates when service groups are configured.
- If the service instance is a member of a service group without a policy configured, it displays statistics for the service group policy.

Example

policy-map ex
class phb
police cir 50000000
class class-default
!
service-group 1
service-policy input ex
end
interface GigabitEthernet0/0/1
no ip address
negotiation auto
service instance 1 ethernet
encapsulation dot1q 100
group 1
bridge-domain 100
!
End

Router(config)# show policy-map interface gigabitEthernet 0/0/1 service instance

GigabitEthernet0/0/1: EFP 1
Service-policy input: ex
Class-map: phb (match-all)
2210042 packets, 3315063000 bytes
5 minute offered rate 82024000 bps, drop rate 77782000 bps
Match: cos 1
police:
cir 50000000 bps, bc 1562500 bytes
conformed 112980 packets, 169470000 bytes; actions:
transmit
exceeded 2097062 packets, 3145593000 bytes; actions:
drop
conformed 4191000 bps, exceeded 77782000 bps
Class-map: class-default (match-any)
0 packets, 0 bytes
5 minute offered rate 0000 bps, drop rate 0000 bps
Match: any

If the service instance is configured with a policy, it displays statistics for the merged policy.

Example

policy-map efpp
class efpc
set dscp 2
!
end
interface GigabitEthernet0/0/1
no ip address
negotiation auto
service instance 1 ethernet
encapsulation dot1q 100
group 1
service-policy input efpp
bridge-domain 100
!
End

Router(config)# show policy-map interface gigabitEthernet 0/0/1 service instance

GigabitEthernet0/0/1: EFP 1
Service-policy input: ex+efpp
Class-map: phb (match-all)
0 packets, 0 bytes
5 minute offered rate 0000 bps, drop rate 0000 bps
Match: cos 1
police:
cir 50000000 bps, bc 1562500 bytes
conformed 0 packets, 0 bytes; actions:
transmit
exceeded 0 packets, 0 bytes; actions:
drop
conformed 0000 bps, exceeded 0000 bps
Class-map: efpc (match-all)
0 packets, 0 bytes
5 minute offered rate 0000 bps, drop rate 0000 bps
Match: dscp 1
QoS Set
dscp 2
Marker statistics: Disabled
Class-map: class-default (match-any)
0 packets, 0 bytes
5 minute offered rate 0000 bps, drop rate 0000 bps
Match: any

Restrictions for Merging Service Groups and EFP Policies

If the physical port of a member EFP has a policy, the policy cannot be attached to a service group.
If the physical port of a member EFP has an egress policy, attaching the egress policy on the service group or adding a member after attaching a policy is not supported on the service group.
If EFP has rewrite push configured, the EFP cannot be a member to a service group with any policy configured.
Internally created (merged) policies cannot be used for configuring of interfaces. Modifications to these policies is not supported.
Policer actions with similar class names on both policies is not allowed.
Conditional and non-conditional marking simultaneously in same class not allowed.
Marking at more than one level is not supported.
Only single level policer is supported at Egress.
3 level Ingress policer is not supported.
Attach queuing-based child policy to a non-queuing based class is not supported.
Match EXP and, match L4 port type is not supported in a single policy-map.
The maximum number of PHB level classes cannot exceed 8 in an Egress policy.
The following commands are not supported:
- show policymap interface interface_name service group service_group_id
- show policy-map target service-group
- show service-group traffic-stats

Creating a Service Group

Procedure

Step 1	enable Example: `Device> enable` Enables privileged EXEC mode. Enter your password if prompted.
Step 2	configure terminal Example: `Device# configure terminal` Enters global configuration mode.
Step 3	service-group `service-group-identifier` Example: `Device(config)# service-group 20` Creates a service group and enters service-group configuration mode. `service-group-identifier` —Service group number.
Step 4	description `descriptive-text` Example: `Device(config-service-group)# description subscriber account number 105AB1` (Optional) Creates a description of the service group. `descriptive-text` — Additional information about the service group. Descriptions can be a maximum of 240 characters.
Step 5	service-policy (input \| output) `policy-map-name` Example: `Device(config-service-group)# service-policy input policy1` (Optional) Attaches a policy map to the service group, in either the ingress (input) or egress (output) direction. `policy-map-name` —previously created policy map.
Step 6	end Example: `Device(config-service-group)# end` (Optional) Returns to privileged EXEC mode.

Adding Service Instance Members to the Service Group

Procedure

Step 1

enable

Example:


Device> enable

Enables privileged EXEC mode.

Enter your password if prompted.

Step 2

configure terminal

Example:


Device# configure terminal

Enters global configuration mode.

Step 3

interface gigabitethernet slot/subslot/port

Example:


Device(config)# interface gigabitEthernet 0/1/5

Specifies the Gigabit Ethernet or Ten Gigabit Ethernet interface to configure and enters interface configuration mode, where:

slot/subslot/port —The location of the interface.

slot —The chassis slot number where the interface module is installed.

Note

The interface module slot number is always 0.

subslot —The subslot where the interface module is installed. Interface module subslots are numbered from 0 to 5, from bottom to top.
port —The number of the individual interface port on an interface module.

Step 4

service instance number ethernet [name]

Example:


Device(config-if)# service instance 200 ethernet

Configure an EFP (service instance) and enter service instance configuration) mode.

The number is the EFP identifier, an integer from 1 to 4000.
(Optional) ethernet name is the name of a previously configured EVC. You do not need to use an EVC name in a service instance.

Step 5

group service-group-identifier

Example:

Device(config-if-srv)# group 20

Creates a service group.

service-group-identifier —Service group number.

Step 6

exit

Example:


Device(config-if-srv)# exit

(Optional) Returns to interface configuration mode.

Step 7

end

Example:


Device(config-if-srv)# end

(Optional) Returns to privileged EXEC mode.

Deleting a Service Group

Procedure

Step 1

enable

Example:


Device> enable

Enables privileged EXEC mode.

Enter your password if prompted.

Step 2

configure terminal

Example:


Device# configure terminal

Enters global configuration mode.

Step 3

no service-group service-group-identifier

Example:


Device(config)# service-group 20

Deletes a service group and deletes all members from the service group.

service-group-identifier —Service group number.

Note

When you delete a service group, all members of the service group are automatically removed from the service group.

Step 4

end

Example:


Device(config)# end

(Optional) Returns to privileged EXEC mode.

Configuration Examples

This example shows policing action on the service group:

policy-map qos-group-in
class cos1
police cir 64000
policy-map qos-group-out
class cos2
police cir 64000
policy-map qos-member-out1
class cos3
police cir 64000
policy-map qos-member-out2
class cos4
police cir 64000

This example shows both Ingress and Egress policies supported on the service group:

service-group 1
service-policy in qos-group-in
service-policy out qos-group-out
int gigabitEthernet1/0/0
service instance 101 ethernet
group 1
service-policy out qos-member-out1
service instance 102 ethernet
group1
service-policy out qos-member-out2
int gigabitEthernet1/0/1
service instance 200 ethernet
group 1
service-policy out qos-member-out1
service instance 300 ethernet
service-policy out qos-member-out2

Verifying the Service Group

Use the show running-config service group command to verify the service groups configuration:

Router# show running-config service-group 1
```
Building configuration...
Current configuration:
service-group 1
service-policy input co1
end
```

Use the show platform software uea-qos service-group stats command to verify the statistics of service groups:

Router# show platform software uea-qos service-group 1 stats

Service Group 1
Service-policy input: co1
class-map: co1:
policy name co1, parent class , parent policy
conformed 54645 packets, 3497280 bytes
exceeded 3705853 packets, 237174592 bytes
violated 0 packets, 0 bytes
conformed 93000 bps, exceeded 6300000 bps

MPLS Diffserv Tunneling Modes Implementation

The MPLS specification defines three Diffserv operation modes:

Uniform—There is only one DiffServ marking that is relevant for a packet when traversing the MPLS network.
Pipe—The network uses two markings for traffic: the original marking value, which is used once the packets leave the MPLS network, and a separate marking value that is used while the traffic crosses intermediate nodes on the LSP span. The second marking value is dropped when traffic exits the MPLS network.
Short-Pipe—the egress LSR uses the original packet marking instead of using the marking used by the intermediate LSRs.

The following sections describe how to implement these modes on the Cisco ASR 920 Series Router using QoS policies.

Implementing Uniform Mode

Use the following guidelines to implement uniform mode on the Cisco ASR 920 Series Router:

Imposition

To copy the diffserv value to the MPLS Exp bit, create a QoS configuration as follows:

Option 1
- Classify based on Prec bit or DSCP bit at ingress.
- Set the qos-group.
- Classify on qos-group.
- Set the MPLS exp value.
Option 2
- Classify based on Prec bit or DSCP bit at ingress.
- Set the mpls Exp bit at imposition.

Tag-to-tag Transfer

To ensure that outer tag values are copied to inner tags, explicitly mark the outer Exp value on the inner Exp bit.

Disposition

To copy the MPLS Exp bit to the diffserv/IP prec bit, create a QoS configuration as follows:

Classify based on MPLS Exp bit on the ingress interface.
Set the qos-group value.
Classify based on qos-group on the egress interface.
Mark the IP prec or DSCP bit.

Implementing Pipe Mode

Use the following guidelines to implement pipe mode on the Cisco ASR 920 Series Router:

Imposition

To set the MPLS Exp bit by policy, create a QoS configuration as follows:

Option 1
- Set the qos-group on the egress interface.
- Classify based on qos-group on the egress interface.
- Set the MPLS Exp value.
Option 2
- Apply the set mpls exp imposition command at ingress.
  Disposition

To preserve the original IP Prec or diffserv value so that egress queuing is based on MPLS exp value, create a QoS configuration as follows:

Classify on MPLS Exp value on the ingress interface.
Set the qos-group on the egress interface.
Classify based on qos-group value on the egress interface.

Implementing Short-Pipe Mode

Use the following guidelines to implement short-pipe mode on the Cisco ASR 920 Series Router:

Disposition

To preserve the original IP Prec or diffserv value so that egress queuing is based on MPLS Prec or diffserv value, create a QoS configuration as follows:

Classify based on IP prec or DSCP value on the egress interface.
Mark the IP prec or DSCP bit.

Classification

Classifying network traffic allows you to organize packets into traffic classes or categories on the basis of whether the traffic matches specific criteria. Classifying network traffic (used in conjunction with marking network traffic) is the foundation for enabling many quality of service (QoS) features on your network.

Table below summarizes the QoS Classification limitations for the Cisco ASR 920 Series Router. In the table, I represents Ingress and E represents Egress.

Table 6. QoS Classification Limitations
Match	Main Layer 3 Interface		EFP Interface		Trunk EFP		L3 Ether- channel		L2 Port Channel		L3 Port Channel Member		L2 Port Channel Member		OC-3		OC-12		T1/E1		MLPPP
Features	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E
Multiple match statements	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.7.1	3.13	3.7.1
access -group	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.7.1	3.13	3.7.1
all	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.7.1	X	3.7.1
any	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.7.1	X	3.7.1
class -map	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
cos	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	X	X	X	X	X	X	X
cos inner	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	X	X	X	X	X	X	X
destination -address	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
discard-class	X	3.5	X	3.5	X	3.5	X	3.9	X	3.9	X	3.5	X	3.9	X	3.9	X	3.9	X	3.7.1	X	3.7.1
dscp (IPv4)	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.7.1	3.13	3.7.1
dscp (IPv6)	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.7.1	X	3.7.1
flow pdp	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
frde	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
frdlci	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
ip dscp	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.7.1	X	3.7.1
ip precedence (IPv4)	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.7.1	X	3.7.1
ip precedence (IPv6)	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.7.1	X	3.7.1
ip rtp	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
mpls experimental	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.7.1	X	3.7.1
mpls experimental topmost	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.7.1	X	3.7.1
not	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
packet length	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
precedence (IPv4)	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.7.1	3.13	3.7.1
precedence (IPv6)	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.7.1	X	3.7.1
protocol	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
qos-group	X	3.5	X	3.5	X	3.5	X	3.9	X	3.9	X	3.5	X	3.9	X	3.9	X	3.9	X	3.7.1	X	3.7.1
service instance ethernet	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
source- address	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
vlan	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.7.1	X	3.7.1
vlan inner	3.5	3.5	3.5	3.5	3.5	3.5	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.7.1	X	3.7.1

Ingress Classification Limitations

The following limitations apply to QoS classification on the Cisco ASR 920 Series Router:

If you configure egress classification for a class of traffic affected by an input policy-map, you must use the same classification criteria on the ingress and egress policy-maps.

Egress Classification Limitations

Egress policy-map with police action is supported on port-channel interface(LAG).
When applying a QoS policy to a link aggregation group (LAG) bundle, you must assign the policy to a physical link within the bundle; you cannot apply the policy to the LAG bundle or the port channel interface associated with the bundle.
MPLS Pipe Mode Limitations—When you configure pipe mode for Time to Live (TTL), the router enables pipe mode for QoS as well. When pipe mode is enabled, you cannot enable egress classification based on the header on an egress interface. For example, you cannot classify based on egress DSCP value for MPLS IP packets when the router is in pipe mode.
MPLS classification using EXP values in an egress policy are applied to normal IP packets in an MPLS core network. When Egress classification for EXP values are converted to equivalent IP precedence values, the first 5 bits in the DSCP values will be used to classify the MPLS packets. However, normal IP packets will be classified as well.

It is recommended to move the EXP classification to ingress policy and egress classification to be moved to the QoS group set from ingress policy to avoid classification of normal IP packets.

Traffic Classifying on MLPPP Interfaces

The Cisco ASR 920 Series Router supports the following match commands in a QoS class-map applied to an egress MLPPP interface.

match discard-class
match dscp
match precedence
match qos-group

Release 3.13 introduces support for ingress QoS on MLPPP interfaces. The Cisco ASR 920 Series Router supports the following match commands in a QoS class-map applied to a ingress MLPPP interface.

match access-group
match dscp
match precedence

The Cisco router supports service-policy input policy-name command on the ingress and egress QoS interface.

Additional Classification Limitations

The topmost policy-map in a three-level hierarchy only supports classification using class-default.

Classifying Traffic using an Access Control List

You can classify inbound packet based on an IP standard or IP extended access control list (ACL). By default, TCAM optimization or expansion method is used. Both Security ACL and QoS ACL can be configured on the same interface. Follow these steps to classify traffic based on an ACL:

Create an access list using the access-list or ip access-list commands
Reference the ACL within a QoS class map using the match access-group configuration command
Attach the class map to a policy map

Limitations and Usage Guidelines

The following limitations and usage guidelines apply when classifying traffic using an ACL:

QoS ACLs are supported only for IPv4 and IPv6 traffic
QoS ACLs are supported only for ingress traffic
You can use QoS ACLs to classify traffic based on the following criteria:
- Source and destination host
- Source and destination subnet
- TCP source and destination
- UDP source and destination
Named and numbered ACLs are supported.
You can apply QoS ACLs only to the third level class (bottom-most).
The following range of numbered access lists are supported:
- 1-99—IP standard access list
- 100-199—IP extended access list
- 1300-1999—IP standard access list (expanded range)
- 2000-2699—IP extended access list (expanded range)
You must create an ACL before referencing it within a QoS policy.
Deny statements within an ACL are ignored for the purposes of classification.
Classifying traffic based on TCP flags using an ACL is not supported.
Classifying traffic using multiple mutually exclusive ACLs within a match-all class-map is not supported.
Classifying traffic on a logical/physical level using an ACL is not supported.
Applying QoS ACLs to MAC addresses is not supported.
Port matching with the neq keyword is only supported for a single port.
A given command can consume multiple matching operations if you specify a source and destination port, as shown in the following examples:
- permit tcp any lt 1000 any —Uses one port matching operation
- permit tcp any lt 1000 any gt 2000 —Uses two port matching operations
- permit tcp any range 1000 2000 any 400 500 —Uses two port matching operations

Only the following combination of matches are currently supported for Ingress policies:

Combination A: DSCP, Outer COS, UDP/TCP Source and Destination port number, IP SA/DA
Combination B: IP SA/DA, Outer COS, Inner COS, DSCP, MPLS EXP

Combination C: MAC DA, Outer COS, Inner COS, DSCP, MPLS Exp

Note

Policy with match on L4 ACL and MPLS EXP together is currently not supported.

For more information about configuring QoS, see the Quality of Service section at http://www.cisco.com/c/en/us/support/routers/asr-920-series-aggregation-services-router/products-installation-and-configuration-guides-list.html.

For more information about configuring access control lists, see the Security section at http://www.cisco.com/c/en/us/support/routers/asr-920-series-aggregation-services-router/products-installation-and-configuration-guides-list.html.

Configuring Multiple Match Statements

Support for a single match or match-any command in a given QoS class-map is shown in the following example:

Example for a Single match or match-any Class Map


class-map match-any my-restrict-class_00
   match ip prec

class-map match-any my-restrict-class_01
   match qos-group 2

class-map match-any my-restrict-class_03
   match cos 3

Support for multiple match or match-any commands in a given QoS class-map is shown in the following example:

Example for a Multiple match or match-any Class Map


class-map match-any my-class
   match ip prec 1
   match qos-group 2
   match cos 3

The router treats the statements as a logical OR operation and classifies traffic that matches any match statement in the class map.

Traffic Classification Using Match EFP Service Instance Feature

Service Provider configurations have various service instances on the PE. QoS policy-maps are applied on these service instances or group of service instances. Cisco IOS XE Release 3.9S introduces the Match EFP Service Instance feature. The benefits of this feature are:

Identify the various types of service-instances like EFP, Trunk EFPs
Apply policies on these service instances at the port
Manage bandwidth and priority across the service instances on the port and across classes within the service instance
Apply policies on a group of transport service instances such as applying similar policies to a group of EFPs.

Restrictions for Configuring Match Service Instances

Ethernet service instances configured under the interface can be classified in a class of a policy-map. The class can match on a group or set of match service instance statements.
```
class-map match-any policeServiceInstance
    match service instance ethernet  100
    match service instance ethernet 200
```
Match service instance supported at both Ingress and Egress level.
match service instance and match PHB per flows classification are defined at respective levels in the policy hierarchy under the port.
The number of EFPs supported per group is 256. Only 256 match statements are supported per class.
Match EFP policy-map can be configured only on the port and not under the service instance.

Example for Configuring Match Service Instances


interface GigabitEthernet0/3/4
 no ip address
 negotiation auto
 service-policy output BTS_Total
 service instance 10 ethernet
  encapsulation dot1q 100
  rewrite ingress tag pop 1 symmetric
  bridge-domain 100
 !
 service instance trunk 20 ethernet
  encapsulation dot1q 20-29
  rewrite ingress tag pop 1 symmetric
  bridge-domain from-encapsulation
 !
 service instance 30 ethernet
  encapsulation dot1q 30
  xconnect 192.44.32.21 101 encapsulation mpls


class-map match-any service-instance-group-with-BMG
match service instance ethernet 10
match service instance ethernet 20

class-map service-instance-30
match service instance ethernet 30

class-map service-instance-20
match service instance ethernet 20

class-map VOICE
match qos-group 0

class-map SIGNALING
match qos-group 1

class-map match-any DATA
match qos-group 2
match qos-group 4

policy-map child-X
class VOICE
priority level 1
police cir 20m
class SIGNALING
priority level 2
police cir 30m
class DATA
shape average 90m
random-detect cos-based


policy-map BTS_OUT_Bi
class service-instance-group-with-BMG
shape average 100m
service-policy child-X
class service-instance-30
shape average 200m
service-policy child-X

policy-map BTS_Total
class class-default
shape average 250m
service-policy BTS_OUT_Bi

QoS Marking

QoS marking allows you to set a desired value on network traffic to make it easy for core devices to classify the packet.

Table below summarizes the QoS Marking limitations for the Cisco ASR 920 Series Router. In the table, I represents Ingress and E represents Egress.

Table 7. Marking QoS Limitations
Features	Main Layer 3 Interface		EFP Interface		Trunk EFP		L3 Ether- channel		L2 Port Channel		L3 Port Channel Member		L2 Port Channel Member		OC-3		OC-12		T1/E1		MLPPP
set	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E
atm- clp	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
cos	3.5	3.6	3.5	3.6	3.5.1	3.6	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	X	X	X	X	X	X	X
cos inner	3.5	3.6	3.5	3.6	3.5.1	3.6	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	X	X	X	X	X	X	X
discard- class	X	3.6	X	3.6	X	3.6	X	3.9	X	3.9	X	3.5	X	3.9	X	3.9	X	3.9	X	3.9	3.13	3.7.1
dscp	3.6	3.6	3.6	3.6	3.6	3.6	3.9	3.9	3.9	3.9	3.6	3.6	3.9	3.9	X	3.9	X	3.9	X	3.9	3.13	3.7.1
dscp- transmit	3.5	X	3.5	X	3.5.1	X	3.9	X	3.9	X	3.5	X	3.9	X	X	X	X	X	X	X	X	X
ip dscp	3.5	3.6	3.5	3.6	3.5.1	3.6	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.9	3.13	3.7.1
ip prece- dence	3.5	3.6	3.5	3.6	3.5.1	3.6	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.9	3.13	3.7.1
mpls experi- mental
mpls experi- mental impo- sition	3.5	X	3.5	X	3.9	X	3.9	X	3.9	X	3.5	X	3.9	X	3.9	X	3.9	X	3.9	X	3.9	X
mpls experi- mental impo- sition qos- group	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
mpls experi- mental topmost	3.5	3.6	3.5	3.6	3.5.1	3.6	3.9	3.9	3.9	3.9	3.5	3.5	3.9	3.9	X	3.9	X	3.9	X	3.9	3.13	3.9
prece- dence	3.6	3.6	3.6	3.6	3.6	3.6	3.9	3.9	3.9	3.9	3.6	3.6	3.9	3.9	X	3.9	X	3.9	X	3.9	3.13	3.7.1
prec- transmit	3.5	X	3.5	X	3.5.1	X	3.9	X	3.9	X	3.5	X	3.9	X	X	X	X	X	X	X	X	X
qos- group	3.6	X	3.6	X	3.6	X	3.9	X	3.9	X	3.5	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1	3.13

Overview of Marking

The Cisco ASR 920 Series Router supports the following parameters with the set command:

set cos
set discard-class
set dscp
set precedence
set ip dscp
set ip precedence
set mpls experimental imposition (ingress marking)
set mpls experimental topmost
set qos-group

CoS Marking Limitations

The following limitations apply when configuring CoS marking:

set cos —This set action has no effect unless there is a egress push action to add an additional header at egress. The COS value set by this action will be used in the newly added header as a result of the push rewrite. If there are no push rewrite on the packet, the new COS value will have no effect.
The set cos inner command is not supported.

Ingress Marking Limitations

The following limitations apply to QoS marking on the Cisco ASR 920 Series Router:

The Cisco ASR 920 Series Router does not support hierarchical marking.
You can configure marking and policing for any number of classes on any one of the three levels of the policy-map hierarchy. If you configure marking on one level, you can configure policing without marking (transmit, drop) on another level.
In the flow of the packet, if both ingress and egress markings are needed, you must classify the packet with the ingress marked phb class at egress and remark it to preserve the ingress marking. Marking in class-default of the ingress marked packets will not preserve the ingress markings.

Egress Marking Limitations

IOS XE Release 3.6 introduces support for egress marking. The following limitations apply when configuring marking on egress interfaces:

The set cos inner command is not supported.
The set mpls experimental imposition command is supported.
The set mpls experimental topmost command is supported for marking MPLS Exp bits; other commands for marking MPLS Exp bits are not supported.

CoS Marking for Pseudowires

The Outer-CoS set in the transport VLAN of the MPLS PW packet, egressing the NNI based on the incoming CoS of the packet coming in on the UNI. With the existing support, a per-EFP or interface QoS policy is applied on the pseudowire originating on the cross-connect on the incoming UNI, to mark the MPLS EXP imposition using the per-EFP or interface policy. By supporting a default EXP to CoS mapping for all pseudowire (L2VPN), the traffic in the transport L2 ring gets the same priority as the ingress policy in the MPLS network.

The default marking of COS from EXP imposition impacts all the pseudowires initiating from the router which are configured with EXP marking policy, that is, the policy to mark imposition EXP marks COS as well.
Egress set cos using egress policy overwrites the S-COS.
If the topmost EXP is changed through ingress marking, the modified EXP is propagated to the egress outer S-COS. Egress set cos can overwrite S-COS.
If the topmost EXP is changed through egress marking, the modified EXP is propagated to the egress outer S-COS. Egress set cos can overwrite S-COS.
The implicit mapping of this EXP to MPLS transport VLAN COS is notsupported. This is applicable only for L2VPN traffic for which the EXP value is derived from the user configured policy.

Example

In the following configuration example, the SVI MPLS is configured between PE1 and P routers. MPLS in physical interfaces is configured between P and PE 2 routers. The EFP X-connect is configured on the Access side.

Topology

ixia---(g0/0/1)PE1(teng0/0/2)---(teng0/2)P(g0/7)---(g0/7)PE2(g0/1)---ixia

PE1 Router

interface Loopback0 
ip address 1.1.1.1 255.255.255.255 
  
vlan 2 
interface vlan2 
no shut 
ip address 20.0.0.1 255.255.255.0 
mpls ip 
mpls label protocol ldp 
  
router ospf 10 
network 1.1.1.1 0.0.0.0 area 0 
network 20.0.0.1 0.0.0.0 area 0 
  
policy-map ingress 
class class-default 
set mpls experimental imposition 4 
  
interface GigabitEthernet 0/0/1 
load-interval 30 
media-type rj45 
service-policy input ingress 
service instance 2 ethernet 
encapsulation dot1q 2 
xconnect 2.2.2.2 10 encapsulation mpls 
  
class-map match-all cos4 
match cos 4 
  
policy-map egress 
class cos4 
  
interface TenGigabitEthernet 0/0/2 
load-interval 30 
service-policy output egress
service instance 2 ethernet
encapsulation dot1q 2
rewrite ingress tag pop 1 symmetric
bridge-domain2

Verifying PE 1 Router

show policy-map interface gig0/1 input GigabitEthernet 0/0/1 

Service-policy input: ingress 
Target association type: DEFAULT 
Class-map: class-default (match-any) 
2000 packets, 128000 bytes 
30 second offered rate 4000 bps, drop rate 0000 bps 
Match: any 
set mpls exp imposition 4 
  
show policy-map interface teng0/2 output TenGigabitEthernet 0/0/2 
Service-policy output: egress 
Target association type: DEFAULT 
Class-map: cos4 (match-all) 
2000 packets, 128000 bytes 
30 second offered rate 4000 bps 
Match: cos 4 
  
Class-map: class-default (match-any) 
0 packets, 0 bytes 
30 second offered rate 0000 bps, drop rate 0000 bps 
Match: any

P router

class-map match-all cos4 
match cos 4 
  
policy-map ingress 
class cos4 
  
interface TenGigabitEthernet 0/2 
load-interval 30 
service-policy input ingress 
  
interface Vlan2 
ip address 20.0.0.2 255.255.255.0 
mpls ip 
mpls label protocol ldp 
  
router ospf 10 
network 20.0.0.2 0.0.0.0 area 0 
network 30.0.0.2 0.0.0.0 area 0 
  
class-map match-all exp4 
match mpls experimental topmost 4 
  
policy-map egress 
class exp4 
  
interface GigabitEthernet 0/7 
ip address 30.0.0.2 255.255.255.0 
media-type rj45 
mpls ip 
mpls label protocol ldp 
service-policy output egress

Verifying P Router

Router# show policy-map interface teng0/2 input TenGigabitEthernet0/2 

Service-policy input: ingress 
Target association type: DEFAULT 
Class-map: cos4 (match-all) 
2000 packets, 188000 bytes 
30 second offered rate 6000 bps 
Match: cos 4 
Class-map: class-default (match-any) 
181 packets, 13992 bytes 
30 second offered rate 1000 bps, drop rate 0000 bps 
Match: any 
  
Router# show policy-map interface gig0/1 output GigabitEthernet 0/7 
Service-policy output: egress 
Target association type: DEFAULT 
Class-map: exp4 (match-all) 
2000 packets, 144000 bytes 
5 minute offered rate 0000 bps 
Match: mpls experimental topmost 4 
Class-map: class-default (match-any) 
4 packets, 216 bytes 
5 minute offered rate 0000 bps, drop rate 0000 bps 
Match: any

PE 2 Router

class-map match-all cos3 
match cos 3 
  
class-map match-all exp4 
match mpls experimental topmost 4 
  
policy-map ingress 
class exp4 
  
policy-map egress 
class cos3 
  
interface Loopback0 
ip address 2.2.2.2 255.255.255.255 
  
interface GigabitEthernet 0/7 
no switchport 
ip address 30.0.0.1 255.255.255.0 
media-type rj45 
mpls ip 
mpls label protocol ldp 
service-policy input ingress 
  
router ospf 10 
network 2.2.2.2 0.0.0.0 area 0 
network 30.0.0.1 0.0.0.0 area 0 
  
interface GigabitEthernet 0/1 
load-interval 30 
media-type rj45 
service-policy output egress 
service instance 2 ethernet 
encapsulation dot1q 2 
xconnect 1.1.1.1 10 encapsulation mpls

Verifying PE2 Route

show policy-map interface gig0/7 input GigabitEthernet 0/7 

Service-policy input: ingress 
Target association type: DEFAULT 
Class-map: exp4 (match-all) 
2000 packets, 172000 bytes 
5 minute offered rate 0000 bps 
Match: mpls experimental topmost 4 
Class-map: class-default (match-any) 
47 packets, 4222 bytes 
5 minute offered rate 0000 bps, drop rate 0000 bps 
Match: any 
  
show policy-map interface gig0/1 output GigabitEthernet0/0/1 
Service-policy output: egress 
Target association type: DEFAULT 
Class-map: cos3 (match-all) 
2000 packets, 128000 bytes 
30 second offered rate 4000 bps 
Match: cos 3 
Class-map: class-default (match-any) 
0 packets, 0 bytes 
30 second offered rate 0000 bps, drop rate 0000 bps 
Match: any

Traffic Marking on MLPPP Interfaces

Egress QoS on MLPPP interfaces.

The Cisco ASR 920 Series Router supports the following parameters with the set command on egress MLPPP interfaces:

set ip dscp
set ip precedence

Ingress QoS on MLPPP interfaces.

The Cisco ASR 920 Series Router supports the following parameters with the set command on the ingress MLPPP interfaces:

set dscp
set precedence
set ip dscp
set ip precedence
set mpls exp imposition
set mpls exp topmost
set qos-group
set discard-class

IPv6 Traffic Marking

The Cisco ASR 920 Series Router supports the following commands for marking both IPv4 and IPv6 packets:

set dscp
set precedence

For more information about IPv6 QoS, see:

Additional Marking Limitations

The following additional marking usage guidelines apply in Release 3.9:

Release 3.9 introduces support for ingress MPLS Exp marking on pseudowire CEM and ATM interfaces, including SAToP, CESoPSN, ATM IMA, and ATMoMPLS.
Marking is supported on Etherchannel interfaces and individual member links; however, you cannot configure marking on both interface levels at once.

CoS Marking for Local Traffic

The injected control packets that are generated from CPU, are marked to a specific CoS value for the list of supported protocols based on the user configuration. The remaining unsupported protocols carry CoS marking as per the existing platform behavior.

Note

The following list of protocols is marked as COS-0 for packets that are generated from RSP2 devices, by default.

The default marking value for NTP is 7.
The default marking value for the other supported protocols is 0.

Following are the supported protocols:

ARP
ISIS
SNMP
NTP
TELNET
SSH
TFTP
Syslog
FTP
DNS
TACACS
ICMP

Example

The following example shows how to enable CoS marking on protocols.

platform cos-mark protocol protocol cos-value cos-value

platform cos-mark protocol snmp cos-value 7 mark

Note

In CoS_mark double tagged packets, if the values remain same then the outer tag CoS value is also copied to inner tag. For the supported protocols, user configured CoS value is marked in outer tag and inner tag.

Limitations

Following are the CoS marking limitations for protocols:

Marking a higher or lower CoS value does not modify egressing queuing decisions on the device where CoS marking is enabled. Locally generated packets exhibit default behavior of egressing on a high-priority queue or a low-priority queue. The CoS value is only marked again on a user configured packet.
For control packets on a class-default queue, if the class-default has set action, then class-default marking overrides the configured protocol marking.

Traffic Policing

Traffic policing allows you to control the maximum rate of traffic sent or received on an interface, and to partition a network into multiple priority levels or class of service (CoS). This section describes the policing limitations and configuration guidelines for the router.

The router supports the following policing types:

Single-rate policer with two color marker (1R2C) (color-blind mode)
Two-rate policer with three color marker (2R3C) (color-blind mode)

Table below summarizes the QoS policing limitations for the Cisco ASR 920 Series Router. In the table, I represents Ingress and E represents Egress.

Policing QoS Limitations


Features	Main Layer 3 Interface		EFP Interface		Trunk EFP		L3 Ether- channel		L2 Port Channel		L3 Port Channel Member		L2 Port Channel Member		OC-3		OC-12		T1/E1		MLPPP
	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E
Rate Limiting
One rate	3.5	3.6	3.5	3.6	3.6	3.6	3.9	3.9	3.9	3.9	3.5	3.6	3.9	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
One rate and two marking	3.5	3.6	3.5	3.6	3.6	3.6	3.9	3.9	3.9	3.9	3.5	3.6	3.9	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
Two rates and two Marking	3.5	X	3.5	X	3.6	X	3.9	X	3.9	X	3.5	X	3.9	X	X	X	X	X	X	X	X	X
Two rates and three actions	3.5	X	3.5	X	3.6	X	3.9	X	3.9	X	3.5	X	3.9	X	X	X	X	X	X	X	X	X
Color Aware policing
Commands
bandwidth	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
bandwidth remaining ratio	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
bandwidth percent	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
police (percent)	X	3.6	X	3.6	X	3.6	X	3.6	X	3.6	X	3.6	X	3.6	X	3.9	X	3.9	X	3.9	X	3.7.1
police (policy map)	X	3.6	X	3.6	X	3.6	X	3.6	X	3.6	X	3.6	X	3.6	X	3.9	X	3.9	X	3.9	X	3.7.1
police (policy map class)	X	3.6	X	3.6	X	3.6	X	3.6	X	3.6	X	3.6	X	3.6	X	3.9	X	3.9	X	3.9	X	3.7.1
police (two rates)	X	3.6	X	3.6	X	3.6	X	3.6	X	3.6	X	3.6	X	3.6	X	3.9	X	3.9	X	3.9	X	3.7.1
priority	X	3.6	X	3.6	X	3.6	X	3.6	X	3.6	X	3.6	X	3.6	X	3.9	X	3.9	X	3.9	X	3.7.1
Supported actions
drop	3.5	X	3.5	X	3.5.1	X	3.9	X	3.9	X	3.5	X	3.9	X	X	X	X	X	X	X	X	X
set- qos- transmit	3.5	X	3.5	X	3.5.1	X	3.9	X	3.9	X	3.5	X	3.9	X	X	X	X	X	X	X	3.13	X
set- cos- transmit	3.5	X	3.5	X	3.5.1	X	3.9	X	3.9	X	3.5	X	3.9	X	X	X	X	X	X	X	3.13	X
set- dscp- transmit	3.5	X	3.5	X	3.5.1	X	3.9	X	3.9	X	3.5	X	3.9	X	X	X	X	X	X	X	3.13	X
set- prec- transmit	3.5	X	3.5	X	3.5.1	X	3.9	X	3.9	X	3.5	X	3.9	X	X	X	X	X	X	X	3.13	X
set- discard- class- transmit	3.5	X	3.5	X	3.5.1	X	3.9	X	3.9	X	3.5	X	3.9	X	X	X	X	X	X	X	3.13	X
set- mpls- experi- mental- topmost- transmit	3.5	3.6	3.5	3.6	3.5.1	3.6	3.9	3.9	3.9	3.9	3.5	3.6	3.9	3.9	X	3.9	X X	3.9	X	3.9	3.13	3.9
set- mpls- experi- mental- impo- sition- transmit	3.5	X	3.5	X	3.5.1	X	3.9	X	3.9	X	3.5	X	3.9	X	X	X	X	X	X	X	3.13	X
transmit	3.5	X	3.5	X	3.5.1	X	3.9	X	3.9	X	3.5	X	3.9	X	X	X	X	X	X	X	X	X

Supported Commands

The Cisco ASR 920 Series Router supports the following policing commands on ingress interfaces:

police (percent)—police cir percent percentage [burst-in-msec] [bc conform-burst-in-msec ms] [be peak-burst-in-msec ms] [pir percent percentage] [conform-action action [exceed-action action [violate-action action]]]
police (policy map)—police cir bps [[bc] normal-burst-bytes [maximum-burst-bytes | [be] [burst-bytes]]] [pir bps [be burst-bytes]] [conform-action action [exceed-action action [violate-action action]]]
police (two rates)—police cir cir [bc conform-burst] [pir pir] [be peak-burst] [conform-action action [exceed-action action [violate-action action]]]

The Cisco ASR 920 Series Router supports the following policing commands on egress interfaces:

bandwidth (policy-map class)—bandwidth {bandwidth-kbps | remaining percent percentage | percent percentage} [account {qinq | dot1q} aal5 subscriber-encapsulation]
bandwidth remaining ratio —bandwidth remaining ratio ratio [account {qinq | dot1q} [aal5] {subscriber-encapsulation | user-define d offset}]
police (policy map)—police cir bps [[bc] normal-burst-bytes [maximum-burst-bytes | [be] [burst-bytes]]] [pir bps [be burst-bytes]] [conform-action action [exceed-action action [violate-action action]]]
priority—priority {bandwidth-kbps | percent percentage} [burst]

Several restrictions apply when using egress policing; see the Egress Policing Limitations section for more information.

Supported Actions

The Cisco ASR 920 Series Router supports the following policing actions on ingress interfaces:

transmit
drop
set-qos-transmit
set-cos-transmit
set-dscp-transmit
set-prec-transmit
set-discard-class-transmit
set-mpls-experimental-topmost-transmit
set-mpls-experimental-imposition-transmit

Percentage Policing Configuration

The router calculates percentage policing rates based on the maximum port PIR rate. The PIR rate is determined as follows:

Default—Port line rate
Speed command applied—Operational rate
Port shaping applied to port—Shaped rate

Ingress Policing Limitations

The following limitations apply to QoS policing on the Cisco ASR 920 Series Router:

If you configure a policer rate or burst-size that the router cannot achieve within 1% accuracy, the configuration is rejected. The command output presents recommendations for the closest possible lower and higher configuration value.
You can configure marking and policing for any number of classes on any one of the three levels of the policy-map hierarchy. If you configure marking on one level, you can configure policing without marking (transmit, drop) on another level.
If you configure marking using the set command, you can only configure policing on that level using the transmit and drop command.
If you configure a policer using a set command, you cannot use the set command at other levels of the hierarchical policy-map.

Egress Policing Limitations

The Cisco ASR 920 Series Router supports the bandwidth and bandwidth-remaining commands on egress interfaces under the following conditions:

Mixed bandwidth types are not supported in the same policy. For example, you cannot configure a policy containing both the bandwidth remaining percent command and bandwidth remaining ratio command.
The bandwidth and bandwidth-remaining commands are not supported in a class containing the priority command. The bandwidth and bandwidth-remaining commands must be configured on classes of the same level.

If you want to create a configuration that uses the bandwidth or bandwidth-remaining commands and the priority command, you must include a police statement in the QoS class.

The following is a sample supported configuration:


Router# show policy-map
  Policy Map PHB
    Class cos1
     police cir 200000 bc 8000
       conform-action transmit
       exceed-action drop
      priority
    Class cos2
      bandwidth 100
      bandwidth remaining percent 40
    Class cos3
      bandwidth 200
      bandwidth remaining percent 50

The priority and police commands must be applied on a single class.

The following is a sample supported configuration:


Router# show policy-map
  Policy Map PHB
    Class cos1
     police cir 200000 bc 8000
       conform-action transmit
       exceed-action drop
      priority
    Class cos2
      bandwidth 100
    Class cos3
      bandwidth 200

Egress MLPPP interfaces support a single-rate policer with two color marker (1R2C) (color-blind mode) at the LLQ level.

Egress port-level policing is supported with ingress EFP policy on the router.

The following is a sample supported configuration:


Policy-map ingress_policy
  Class cos3
    Set cos 5
Policy-map egress policy
  Class cos5
    Shape average 30m

###Ingress
interface GigabitEthernet0/4/0
no ip address
negotiation auto
service instance 100 ethernet
  encapsulation dot1q 100
  service-policy input ingress_policy  >>>> Ingress policy in EFP
  bridge-domain 100

###Egress
interface GigabitEthernet0/4/0
no ip address
negotiation auto
service-policy output egress_policy   >>>>>Egress policy on Port
service instance 100 ethernet
  encapsulation dot1q 100
  bridge-domain 100

QoS features on egress policing is supported on MLPPP interfaces using the police command. Egress MLPPP interfaces support a single-rate policer with two color marker (1R2C) (color-blind mode) at the LLQ level.
Police and Set in same policy class-map

police and set commands can be configured together in the egress policy class-map. In prior releases, a error message was displayed when both police and set commands were configured.

Sample example displaying the error message:
```
Router(config)#policy-map egress
Router(config-pmap)#class p1
Router(config-pmap-c)#police cir 200m
Router(config-pmap-c-police)#set prec 2
Qos:Configuration failed - Set and police not allowed in same class p1 of policy egress
QoS: Configuration failed. Invalid set
```
Egress Policing on Non Priority Queue

Policing is supported at the egress on non priority queues.

Sample configuration:


Router#sh policy-map testp
  Policy Map testp
    Class cos1
      priority
     police cir 20000000 bc 625000
       conform-action transmit
       exceed-action drop
    Class cos2
     police cir 20000000 bc 625000
       conform-action transmit
       exceed-action drop
    Class cos4
     police cir 50000000 bc 1562500
       conform-action transmit
       exceed-action drop

Traffic Policing on MLPPP Interfaces

Release 3.13 introduces support for egress QoS on MLPPP interfaces. 1R2C color-blind policer is supported for egress QoS on MLPPP interfaces.

Release 3.13 introduces support for ingress QoS on MLPPP interfaces. The Cisco ASR 920 Series Router supports the following parameters with the set command on the ingress MLPPP interfaces:

set dscp transmit
set prec transmit
set mpls exp imposition transmit
set mpls exp topmost transmit
set qos transmit
set discard-class transmit

1R2C color-blind policer and 2R3C color-blind policer is supported for ingress QoS on MLPPP interfaces.

Traffic Shaping

Traffic shaping allows you to control the speed of traffic that is leaving an interface in order to match the flow of traffic to the speed of the receiving interface. Percentage-based policing allows you to configure traffic shaping based on a percentage of the available bandwidth of an interface.Configuring traffic shaping in this manner enables you to use the same policy map for multiple interfaces with differing amounts of bandwidth.

This section describes the shaping limitations and configuration guidelines for the Cisco ASR 920 Series Router.

Table below summarizes the QoS shaping limitations for the Cisco ASR 920 Series Router; an X indicates support.

Table 8. Shaping Limitations by Interface
	GigE		10 GigE		EFP		Trunk EFP		Port Channel		Member Link		OC-3		OC-12		T1/E1		Serial		MLPPP		ACL
	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E
Features
Class- Based
Distri- buted
Frame
Generic
Shape
adaptive		X				X		X			X
average		X				X		X			X											X
fecn- adapt		X				X		X			X
max- buffers		X				X		X			X
peak		X				X		X			X

Additional Shaping Limitations

The following additional shaping usage guidelines apply in Release 3.9:

Policies using shaping are supported only on individual member links of an etherchannel. Applying a shaping policy directly on an etherchannel interface is not supported.
Class-based shaping is supported at all levels.

Configuring Egress Shaping on EFP Interfaces

Configuring an EFP port shaper allows you to shape all EFPs on a port using a port policy with a class-default shaper configuration, as in the following partial sample configuration:


policy-map port-policy
  class class-default
    shape average percent 50
policy-map efp-policy
  class class-default
    shape average percent 25
    service-policy child-policy
policy-map child-policy
  class phb-class
    <class-map actions>

The following configuration guidelines apply when configuring an EFP port shaping policy:

When the configuration specifies a shaper rate using a percentage, the router calculates the value based on the operational speed of a port. The operational speed of a port can be the line rate of the port or the speed specified by the speed command.
The rates for bandwidth percent and police percent commands configured under a port-shaper are based on the absolute rate of the port-shaper policy.
You can combine a port shaper policy (a flat shaper policy with no user-defined classes) with an egress EFP QoS shaping policy.
Configure the port shaper policy before configuring other egress QoS policies on EFP interfaces; when removing EFP QoS configurations, remove other egress EFP QoS policies before removing the port shaper policy.

Congestion Management

Congestion management features allow you to control congestion by determining the order in which packets are sent out an interface based on priorities assigned to those packets. Congestion management entails the creation of queues, assignment of packets to those queues based on the classification of the packet, and scheduling of the packets in a queue for transmission.

This section describes the classification limitations and configuration guidelines for the Cisco ASR 920 Series Router.

Table below summarizes the QoS congestion management and queuing limitations for the Cisco ASR 920 Series Router. In the table, I represents Ingress and E represents Egress.

Table 9. Congestion Management QoS Limitations
Features	Main Layer 3 Interface		EFP Interface		Trunk EFP		L3 Ether- channel		L2 Port Channel		L3 Port Channel Member		L2 Port Channel Member		OC-3		OC-12		T1/E1		MLPPP
	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E
CBWFQ	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.8	X	3.9	X	3.9	X	3.7.1
D-CBWFQ
IP RTP priority	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
Frame Relay IP RTP	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
Frame Relay PVC Interface Priority Queuing	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
LLQ	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.8	X	3.9	X	3.9	X	3.7.1
D-LLQ
LLQ-FR
Custom queuing
Priority Queuing
Commands
bandwidth (kbps)	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
bandwidth percent	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
bandwidth remaining percent	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
bandwidth remaining ratio	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
compression header ip	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
drop	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
fair-queue	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
priority	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
priority (kbps)	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
priority (without queue-limit)
priority percent	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
queue-limit (cells)	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
queue-limit (packets)	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X

Ingress Queuing Limitations

The Cisco ASR 920 Series Router does not support queuing on ingress interfaces.

Egress Queuing Limitations

The Cisco ASR 920 Series Router supports tail drop queuing on egress interfaces using the queue-limit command. The following limitations apply to egress queuing:

If you configure a queue size that the router cannot achieve within 1% accuracy, the configuration is rejected. The command output presents recommendations for the closest possible lower and higher configuration value.
Egress policy-map with queuing action is not supported on port-channel interface(LAG). The policy must be applied to the policy-maps on the member links.
Release 3.8 extends the maximum bytes value of the queue-limit number-of-packets [bytes | ms | packets] command. The previous maximum value was 491520 bytes; the new value is 2 MB.
Release 3.8 enhances the show policy-map interface command to display the default queue-limit.
Release 3.8 introduces support for the queue-limit percent command.

Support for Queuing Features on MLPPP Interfaces

Release 3.7(1) introduces support for QoS features on egress MLPPP interfaces. The following queuing features are supported on egress MLPPP interfaces:

Tail drop queuing uses the queue-limit command.
Hierarchical QoS (2 level) is supported on MLPPP.
3-level policies are not supported on MLPPP interfaces.

Support for Low Latency Queuing on Multiple EFPs

The Cisco ASR 920 Series Router supports QoS policies that allow for low-latency queuing (LLQ) across multiple EFPs. For more information about this feature, see http://www.cisco.com/en/US/docs/ios-xml/ios/qos_plcshp/configuration/xe-3s/qos-plcshp-ehqos-pshape.html.

Additional Queuing Limitations

The following additional queuing usage guidelines:

The router supports QoS policies that allow for low-latency queuing (LLQ) across multiple EFPs. For more information about this feature, see http://www.cisco.com/en/US/docs/ios-xml/ios/qos_plcshp/configuration/xe-3s/qos-plcshp-ehqos-pshape.html.
CBWFQ is supported on second and third level classes.
Queue-limit is supported only in leaf-level (per-hop behavior) classes.
Queue-limit can not be configured without first configuring a scheduling action (bandwidth, shape average, or priority).
Queue-limit can not co-exist with queue-limit percent.
Queue-limit percent and WRED can not be configured in the same class.
Queue-limit percent can not be configured with different qualifiers (cos, dscp, prec, and Qos-group).
Queue-limit policy can be applied only on egress interface.
Queue-limit can be configured in bytes or microseconds, or percent per class in the egress-policy.
Default queue-limits for 1 and 10 G are 80 and 120 KB, repectively.
Maximum queue-limit that can be configured in bytes is 2 MB.

Congestion Avoidance

Congestion avoidance techniques monitor network traffic loads in an effort to anticipate and avoid congestion at common network bottlenecks. Congestion avoidance is achieved through packet dropping. Among the more commonly used congestion avoidance mechanisms is Random Early Detection (RED), which is optimum for high-speed transit networks. Cisco IOS QoS includes an implementation of RED that, when configured, controls when the router drops packets. If you do not configure Weighted Random Early Detection (WRED), the router uses the cruder default packet drop mechanism called tail drop.

Table below summarizes the QoS congestion avoidance limitations for the Cisco ASR 920 Series Router. In the table, I represents Ingress and E represents Egress.

Table 10. Congestion Avoidance QoS Limitations
Features	Main Layer 3 Interface		EFP Interface		Trunk EFP		L3 Ether- channel		L2 Port Channel		L3 Port Channel Member		L2 Port Channel Member		OC-3		OC-12		T1/E1		MLPPP
	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E	I	E
Tail Drop (default)
RED	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
random- detect	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
WRED	not supported on ingress ifcs
random- detect discard- class
random- detect discard- class- based	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
random- detect expon- ential- weight- ing- constant	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
fair- queue	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X
prece- dence
random- detect cos- based	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
random- detect dscp- based	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
random- detect prec- edence- based	X	3.6	X	3.6	X	3.6	X	X	X	X	X	3.6	X	3.9	X	3.9	X	3.9	X	3.9	X	3.7.1
shape
D-WRED
Flow- based WRED
Diffserv WRED

Congestion Avoidance Configuration

The following sections describe the supported congestion avoidance features on the Cisco ASR 920 Series Router:

Supported Commands

The Cisco ASR 920 Series Router supports the following commands for WRED:

random-detect cos-based —Outer CoS
random-detect dscp-based — IPv4 DSCP
random-detect precedence-based — IPv4 Precedence bit

Supported Interfaces

WRED is supported at the PHB level but not on logical or physical interfaces. You can apply WRED policies on the following interface types:

Main Layer 3 interface
Port-channel Layer 3 member-links
Service instances
Trunk EFPs

Verifying the Configuration

You can use the show policy-map interface command to display the number of WRED drops and tail drops.

For more information about configuring congestion avoidance, see the following documents:

Ingress Congestion Avoidance Limitations

WRED is not supported on ingress interfaces.

Egress Congestion Avoidance Limitations

The following limitations apply when configuring congestion avoidance on the Cisco ASR 920 Series Router:

Queuing feature to support WRED in a class such as shape or bandwidth are supported.
You must apply WRED within a policy map.
WRED is only supported on egress interfaces.
WRED is not supported in priority queues.
WRED is supported in the class-default class if there are no other user-defined classes in the policy-map.
You can configure a maximum of 2 WRED curves per class.
Fair-queue is not supported. You can configure WRED with either the shape or the fair-queue (CBWFQ) commands.
The default value for exponential-weighting-constant is 9.
The default value for mark-probability is 10.
You can specify the minimum-threshold and maximum-threshold in terms of bytes or microseconds. Setting threshold values in terms of packets is not supported.

Egress Congestion Avoidance on MLPPP Interfaces

Release 3.7(1) introduces support for the following egress congestion features on MLPPP interfaces:

RED queuing using the random-detect command
WRED queuing using the random-detect command. You can apply WRED to:
- DSCP
- Precedence
- Discard-class

MLPPP egress queuing is supported only on the 3rd level classes (bottom-most).

Class-based Weighted Fair Queuing (CBWFQ) using the bandwidth and bandwidth percent commands. CBWFQ is supported on 2nd and 3rd level classes.
Class-based Shaping using the shape average and shape average percent commands. Class-based shaping is supported at all levels.
Class-based excess bandwidth scheduling using the bandwidth remaining percent and bandwidth remaining ratio commands. Class-based excess bandwidth scheduling is supported on 2nd and 3rd level QoS classes.

Additional Congestion Avoidance Limitations

You can specify the minimum-threshold and maximum-threshold in terms of bytes or microseconds. Setting threshold values in terms of packets is not supported.
Policies using Class-based Weighted Fair Queuing (CBWFQ) and WRED are supported only on individual member links of an etherchannel. Applying a CBWFQ policy directly on an etherchannel interface is not supported.
Aggregate-WRED is not supported. However, multiple random-detect statements with the same curve are supported in the same class.

Verifying the Configuration

You can use the show policy-map interface command to display the number of WRED drops and tail drops.

For more information about configuring congestion avoidance, see the following documents:

Scheduling

This section describes the scheduling limitations and configuration guidelines for the Cisco ASR 920 Series Router.

Ingress Scheduling Limitations

The Cisco ASR 920 Series Router does not support scheduling on ingress interfaces.

Egress Scheduling Limitations

If you configure a CIR, PIR, or EIR rate that the router cannot achieve within 1% accuracy, the configuration is rejected. The command output presents recommendations for the closest possible lower and higher configuration value.
You can only configure one priority value on each parent class applied to a QoS class or logical interface.
You can only configure priority on one class in a QoS policy.
You can not configure priority value and a policer in the same class.

The following limitations apply when configuring a 3-level scheduling policy on an egress interface configured as an EFP:

Only two of the three levels can contain scheduling actions such as bandwidth , shape , or priority .
Class-based excess bandwidth scheduling is supported on 2nd and 3rd level QoS classes.
One of the levels containing scheduling actions must be the class (bottom) level.

Egress Scheduling on MLPPP Interfaces

Release 3.7(1) introduces support for QoS features on egress MLPPP interfaces including scheduling. The following scheduling features are supported:

Strict priority using the priority command; strict priority is supported on 2nd and 3rd level classes.

Multi-level priority using the priority level command. You can configure two priority levels; the feature is supported on 3rd level classes. The following is the sample configuration of multi-level priority.


policy-map eg_pri_queuing
class eg_mul1_prec1
  set precedence 6
  priority level 1 percent 20
class eg_mul1_prec2
  set precedence 5
  priority level 2 percent 18
class eg_mul1_prec3
  set precedence 4
  shape average 4000000
class class-default
  set precedence 3
  shape average 2000000
!

The following limitations apply when configuring a 3-level scheduling policy on an egress interface configured as an EFP:

Only two of the three levels can contain scheduling actions such as bandwidth , shape , or priority .
QoS policies using the priority command are supported only on individual member links.
MLPPP is not supported on port-channel (etherchannel).

Egress QoS for IPSLA

Starting from Cisco IOS XE 16.8.1, IPSLA packet classification is enabled in egress QoS. This feature enables you to apply classification and queueing on the egress interface for IPSLA packets. The egress interface can be either a Layer 2 interface under BDI or a Layer 3 physical interface.

The following table provides IPSLA packet classification based on QoS and egress interface.

Table 11. IPSLA QoS based on Egress Interface
Egress interface	QoS Classification
Layer 2 interface under BDI	Based on VLAN
Layer 3 physical interface	Based on DSCP

The egress QoS classification for IPSLA packets is disabled by default and a global level command has to be enabled to support IPSLA classification on the egress interface.

To enable egress classification for IPSLA packets, use the platform ipsla classify cpu packets command at the global configuration mode, and then reload the router.

You can verify if the egress QoS is enabled for IPSLA packets using the show romvar | i IPSLA command.

The TOS value set from IPSLA UDP jitter configurations is used as QoS value in the packet.

Limitations for IPSLA QoS

IPSLA UDP jitter packets support only the version 1 of operations, and the remaining version 2 and version 3 are not supported.
The QoS operations on IPSLA traffic with respect to COS values is not supported.
The marking of IPSLA UDP jitter packets is not supported on the ingress interface of the responder.
The egress QoS classification for IPSLA packets is supported only in the legacy mode and legacy IPSLA flows.

Configuring QoS for IPSLA

To enable egress Qos for IPSLA at the global configuration mode using the following command and then reload the router:

router#configure terminal
router(config)#platform ipsla classify cpu packets
router(config)#end

To verify if the egress QoS for IPSLA is enabled, use the following command:

router#show romvar | i IPSLA
IPSLA_QOS = 1

To perform global configuration on the IPSLA source for egress QoS, use the following commands:

 router#configure terminal
router(config)#ip sla 1
router(config-ip-sla)#udp-jitter 192.0.2.2 1024 source-ip 192.0.2.1 num-packets 241 interval 30
router(config-ip-sla)#frequency 15
router(config-ip-sla)#verify-data
router(config-ip-sla)#tos 32           <------------ QoS value in the IPSLA traffic 
router(config-ip-sla)#timeout  5000
router(config-ip-sla)#threshold  5000
router(config-ip-sla)#end

To perform global configuration on the IPSLA responder, use the following commands:

router#configure terminal
router(config)#ip sla responder 
router(config)#end

To start IPSLA UDP jitter traffic, use the following command:

router#configure terminal
router(config)#ip sla schedule SLA-ID start-time now
router(config)#end

To perform basic QoS configuration on the interfacecs for the IPSLA traffic being sent, use the following commands:

router#configure terminal
router(config)#class-map cs1
router(config-cmap)#match ip dscp cs1
router(config-cmap)#policy-map match_cs1
router(config-pmap)#class cs1
router(config-pmap-c)#end

To apply service policy on egress interface, use the following commands:

router#configure terminal
router(config)#interface GigabitEthernet0/0/4
router(config-if)#service-policy output match_cs1
router(config-if)#end

To verify egress QoS for IPSLA, use the following command:

router#show policy-map interface

Verifying Egress QoS for IPSLA Configuration

To verify egress QoS for IPSLA configuration, use the show policy-map interface command:

router# show policy-map interface
GigabitEthernet0/0/4 

  Service-policy output: match_cs1

    Class-map: cs1 (match-all)  
      2169 packets, 177858 bytes
      5 minute offered rate 5000 bps
      Match: ip dscp cs1 (8)

    Class-map: class-default (match-any)  
      9 packets, 666 bytes
      5 minute offered rate 0000 bps, drop rate 0000 bps
      Match: any

Additional References

Related Topic	Document Title
Cisco IOS commands	https://www.cisco.com/c/en/us/td/docs/ios-xml/ios/mcl/allreleasemcl/all-book.html

Standards and RFCs


Standard/RFC	Title
No specific Standards and RFCs are supported by the features in this document.	—

MIBs


MIB	MIBs Link
—	To locate and download MIBs for selected platforms, Cisco IOS releases, and feature sets, use Cisco MIB Locator found at the following URL: http://www.cisco.com/go/mibs

Technical Assistance


Description	Link
The Cisco Support website provides extensive online resources, including documentation and tools for troubleshooting and resolving technical issues with Cisco products and technologies. To receive security and technical information about your products, you can subscribe to various services, such as the Product Alert Tool (accessed from Field Notices), the Cisco Technical Services Newsletter, and Really Simple Syndication (RSS) Feeds. Access to most tools on the Cisco Support website requires a Cisco.com user ID and password.	http://www.cisco.com/cisco/web/support/index.html

Feature Information for Configuring QoS

The following table provides release information about the feature or features described in this module. This table lists only the software release that introduced support for a given feature in a given software release train. Unless noted otherwise, subsequent releases of that software release train also support that feature.

Use Cisco Feature Navigator to find information about platform support and Cisco software image support. To access Cisco Feature Navigator, go to www.cisco.com/go/cfn. An account on Cisco.com is not required.

Table 12. Feature Information for QoS
Feature Name	Releases	Feature Configuration Information
Configuring QoS	Cisco IOS XE Release 3.13.0S	This feature was introduced on the Cisco ASR 920 Series Aggregation Services Router (ASR-920-12CZ-A, ASR-920-12CZ-D, ASR-920-4SZ-A, ASR-920-4SZ-D).

Bias-Free Language

Results

Chapter: Quality of Service Configuration Guidelines

Quality of Service Configuration Guidelines

Quality of Service

Quality of Service Configuration

QoS Support Overview

Global QoS Limitations

QoS Features Using MQC Limitations

Restrictions for Ingress QoS

Restrictions for Egress QoS

Routed Port-Channel

Example for egress policy-map on member links on the routed port-channel interface

Port-channel with EFP

Example for egress policy-map on port-channel on EFP

EFP of Port-channel with EFP Configuration

Member Links of Port-channel with EFP Configuration

Restrictions for Hierarchical Policies

Sample Hierarchical Policy Designs

Sample Policy

Ingress and Egress Hierarchical Policing

Dissimilar PHB Support for MPLS and VPLS Interfaces

Restrictions for Dissimilar PHB Support for MPLS and VPLS Interfaces

MPLS VPN QoS Mapping

QoS Policer and Shaper Calculation

Service Groups

Restrictions for Service Groups

Merging Service Groups and EFP Policies

Example

Example

Restrictions for Merging Service Groups and EFP Policies

Creating a Service Group

Procedure

Example:

Example:

Example:

Example:

Example:

Example:

Adding Service Instance Members to the Service Group

Procedure

Example:

Example:

Example:

Example:

Example:

Example:

Example:

Deleting a Service Group

Procedure

Example:

Example:

Example:

Example:

Configuration Examples

Verifying the Service Group

MPLS Diffserv Tunneling Modes Implementation

Implementing Uniform Mode

Implementing Pipe Mode

Implementing Short-Pipe Mode

Classification

Ingress Classification Limitations

Egress Classification Limitations

Traffic Classifying on MLPPP Interfaces

Additional Classification Limitations

Classifying Traffic using an Access Control List

Limitations and Usage Guidelines

Configuring Multiple Match Statements

Traffic Classification Using Match EFP Service Instance Feature

Restrictions for Configuring Match Service Instances

Example for Configuring Match Service Instances

QoS Marking

Overview of Marking

CoS Marking Limitations

Ingress Marking Limitations

Egress Marking Limitations

CoS Marking for Pseudowires

Example

Traffic Marking on MLPPP Interfaces