Configuring Modular QoS Service Packet Classification

This chapter covers these topics:

Packet Classification Overview

Packet classification involves categorizing a packet within a specific group (or class) and assigning it a traffic descriptor to make it accessible for QoS handling on the network. The traffic descriptor contains information about the forwarding treatment (quality of service) that the packet should receive. Using packet classification, you can partition network traffic into multiple priority levels or classes of service. The source agrees to adhere to the contracted terms and the network promises a quality of service. Traffic policers and traffic shapers use the traffic descriptor of a packet to ensure adherence to the contract.

Traffic policers and traffic shapers rely on packet classification features, such as IP precedence, to select packets (or traffic flows) traversing a router or interface for different types of QoS service. After you classify packets, you can use other QoS features to assign the appropriate traffic handling policies including congestion management, bandwidth allocation, and delay bounds for each traffic class.

The Modular Quality of Service (QoS) command-line interface (MQC) is used to define the traffic flows that must be classified, where each traffic flow is called a class of service, or class. Subsequently, a traffic policy is created and applied to a class. All traffic not identified by defined classes fall into the category of a default class.

Traffic Class Elements

The purpose of a traffic class is to classify traffic on your router. Use the class-map command to define a traffic class.

A traffic class contains three major elements:

  • A name

  • A series of match commands - to specify various criteria for classifying packets.

  • An instruction on how to evaluate these match commands (if more than one match command exists in the traffic class)

Packets are checked to determine whether they match the criteria specified in the match commands. If a packet matches the specified criteria, that packet is considered a member of the class and is forwarded according to the QoS specifications set in the traffic policy. Packets that fail to meet any of the matching criteria are classified as members of the default traffic class.

This table shows the details of match types supported on the router.

Match Type Supported

Min, Max

Max Entries

Support for Match NOT

Support for Ranges

Direction Supported on Interfaces

IPv4 DSCP

IPv6 DSCP

DSCP

(0,63)

64

Yes

Yes

Ingress

IPv4 Precedence

IPv6 Precedence

Precedence

(0,7)

8

Yes

No

Ingress

MPLS Experimental Topmost

(0,7)

8

Yes

No

Ingress

Access-group

Not applicable

8

No

Not applicable

Ingress

QoS-group

(1,7)

7

No

No

Egress

Protocol

(0,255)

1

Yes

Not applicable

Ingress


Note

Egress queue statistics are displayed only for those classes which have a corresponding match criteria in the egress. Therefore, if you have a set qos-group x configured in the ingress, you must have a corresponding match qos-group x in the egress, in order to see the statistics in the egress side. Also, see Usage of QoS-group and Queue Selection.

Default Traffic Class

Unclassified traffic (traffic that does not meet the match criteria specified in the traffic classes) is treated as belonging to the default traffic class.

If the user does not configure a default class, packets are still treated as members of the default class. However, by default, the default class has no enabled features. Therefore, packets belonging to a default class with no configured features have no QoS functionality. These packets are then placed into a first in, first out (FIFO) queue and forwarded at a rate determined by the available underlying link bandwidth. This FIFO queue is managed by a congestion avoidance technique called tail drop.

For egress classification, match on qos-group (1-7) is supported. Match qos-group 0 cannot be configured. The class-default in the egress policy maps to qos-group 0 .

This example shows how to configure a traffic policy for the default class: 


configure
 policy-map ingress_policy1
 class class-default
  police rate percent 30
 !

Create a Traffic Class

To create a traffic class containing match criteria, use the class-map command to specify the traffic class name, and then use the match commands in class-map configuration mode, as needed.

Guidelines

  • Users can provide multiple values for a match type in a single line of configuration; that is, if the first value does not meet the match criteria, then the next value indicated in the match statement is considered for classification.

  • Use the not keyword with the match command to perform a match based on the values of a field that are not specified.

  • All match commands specified in this configuration task are considered optional, but you must configure at least one match criterion for a class.

  • If you specify match-any , one of the match criteria must be met for traffic entering the traffic class to be classified as part of the traffic class. This is the default. If you specify match-all , the traffic must match all the match criteria.

  • For the match access-group command, QoS classification based on the packet length or TTL (time to live) field in the IPv4 and IPv6 headers is not supported.

  • For the match access-group command, when an ACL list is used within a class-map, the deny action of the ACL is ignored and the traffic is classified based on the specified ACL match parameters.

  • The match qos-group , traffic-class , and discard-class are supported only in egress direction, and these are the only match criteria supported in egress direction.

  • The egress default class implicitly matches qos-group 0.

  • Multicast takes a system path that is different than unicast on router, and they meet later on the egress in a multicast-to-unicast ratio of 20:80 on a per interface basis. This ratio is maintained on the same priority level as that of the traffic.

  • Egress QoS for multicast traffic treats traffic classes 0-5 as low-priority and traffic classes 6-7 as high priority. Currently, this is not user-configurable.

  • Egress shaping does not take effect for multicast traffic in the high priority (HP) traffic classes. It only applies to unicast traffic.

  • If you set a traffic class at the ingress policy and do not have a matching class at egress for the corresponding traffic class value, then the traffic at ingress with this class will not be accounted for in the default class at the egress policy map.

  • Only traffic class 0 falls in the default class. A non-zero traffic class assigned on ingress but with no assigned egress queue, falls neither in the default class nor any other class.

  • Also, see Usage of QoS-group and Queue Selection.

Configuration Example

You have to accomplish the following to complete the traffic class configuration:

  1. Creating a class map

  2. Specifying the match criteria for classifying the packet as a member of that particular class

    (For a list of supported match types, see Traffic Class Elements.) 


Router# configure
Router(config)# class-map match-any qos-1
Router(config-cmap)# match qos-group 1
Router(config-cmap)# end-class-map
Router(config-cmap)# commit

Use this command to verify the class-map configuration: 


Router#show class-map qos-1
1) ClassMap: qos-1    Type: qos
    Referenced by 2 Policymaps

Also see, Running Configuration.

Also see, Verification.

Related Topics

Associated Commands

Traffic Policy Elements

A traffic policy contains three elements:

  • Name

  • Traffic class

  • QoS policies

After choosing the traffic class that is used to classify traffic to the traffic policy, the user can enter the QoS features to be applied to the classified traffic.

The MQC does not necessarily require that the users associate only one traffic class to one traffic policy.

The order in which classes are configured in a policy map is important. The match rules of the classes are programmed into the TCAM in the order in which the classes are specified in a policy map. Therefore, if a packet can possibly match multiple classes, only the first matching class is returned and the corresponding policy is applied.

The router supports 32 classes per policy-map in the ingress direction and 8 classes per policy-map in the egress direction.

This table shows the supported class-actions on the router.

Supported Action Types

Direction supported on Interfaces

minimum-bandwidth

egress

bandwidth-remaining

egress

mark

(See Packet Marking)

police

ingress

priority

egress (level 1 to level 7)

queue-limit

egress

shape

egress

wred

egress

WRED supports default and discard-class options; the only values to be passed to the discard-class being 0 and 1.

Create a Traffic Policy

The purpose of a traffic policy is to configure the QoS features that should be associated with the traffic that has been classified in a user-specified traffic class or classes.

To configure a traffic class, see Create a Traffic Class.

After you define a traffic policy with the policy-map command, you can attach it to one or more interfaces to specify the traffic policy for those interfaces by using the service-policy command in interface configuration mode. With dual policy support, you can have two traffic policies, one marking and one queuing attached at the output. See, Attach a Traffic Policy to an Interface.

Configuration Example

You have to accomplish the following to complete the traffic policy configuration:

  1. Creating a policy map that can be attached to one or more interfaces to specify a service policy

  2. Associating the traffic class with the traffic policy

  3. Specifying the class-action(s) (see Traffic Policy Elements) 


Router# configure
Router(config)# policy-map  test-shape-1
Router(config-pmap)# class qos-1

/* Configure class-action ('shape' in this example). 
Repeat as required, to specify other class-actions */
Router(config-pmap-c)# shape average percent 40
Router(config-pmap-c)# exit

/* Repeat class configuration as required, to specify other classes */

Router(config-pmap)# end-policy-map
Router(config)# commit

See, Running Configuration.

See, Verification.

Related Topics

Associated Commands

Attach a Traffic Policy to an Interface

After the traffic class and the traffic policy are created, you must attach the traffic policy to interface, and specify the direction in which the policy should be applied.


Note

Hierarchical policies are not supported.

When a policy-map is applied to an interface, the transmission rate counter of each class is not accurate. This is because the transmission rate counter is calculated based on the exponential decay filter.

Configuration Example

You have to accomplish the following to attach a traffic policy to an interface:

  1. Creating a traffic class and the associated rules that match packets to the class (see Create a Traffic Class )

  2. Creating a traffic policy that can be attached to one or more interfaces to specify a service policy (see Create a Traffic Policy )

  3. Associating the traffic class with the traffic policy

  4. Attaching the traffic policy to an interface, in the ingress or egress direction 


Router# configure
Router(config)# interface HundredGigE 0/6/0/18
Router(config-int)# service-policy output test-shape-1
Router(config-int)# commit

Running Configuration 


/* Class-map configuration */

class-map match-any traffic-class-1
 match traffic-class 1
 end-class-map
!
- - -
- - -

/* Traffic policy configuration */
policy-map test-shape-1
 class traffic-class-1
  shape average percent 40
 !
 class class-default
 !
 end-policy-map
!
- - -
- - -

/* Attaching traffic policy to an interface in egress direction */ 
interface HundredGigE0/6/0/18
 service-policy output test-shape-1
 !

Verification


Router# show qos interface hundredGigE 0/6/0/18 output 

NOTE:- Configured values are displayed within parentheses Interface HundredGigE0/6/0/18 ifh 0x30001f8  -- output policy
NPU Id:                        3
Total number of classes:       2
Interface Bandwidth:           100000000 kbps
VOQ Base:                      11112
VOQ Stats Handle:              0x88430698
Accounting Type:               Layer1 (Include Layer 1 encapsulation and above)
------------------------------------------------------------------------------
Level1 Class                             =   qos-1
Egressq Queue ID                         =   11113 (LP queue)
Queue Max. BW.                           =   40329846 kbps (40 %)
Queue Min. BW.                           =   0 kbps (default)
Inverse Weight / Weight                  =   1 / (BWR not configured)
Guaranteed service rate                  =   40000000 kbps
TailDrop Threshold                       =   50069504 bytes / 10 ms (default)
WRED not configured for this class

Level1 Class                             =   class-default
Egressq Queue ID                         =   11112 (Default LP queue)
Queue Max. BW.                           =   101803495 kbps (default)
Queue Min. BW.                           =   0 kbps (default)
Inverse Weight / Weight                  =   1 / (BWR not configured)
Guaranteed service rate                  =   50000000 kbps
TailDrop Threshold                       =   62652416 bytes / 10 ms (default)
WRED not configured for this class

Related Topics

Associated Commands

Packet Marking

The packet marking feature provides users with a means to differentiate packets based on the designated markings. The router supports egress packet marking. match on discard-class on egress, if configured, can be used for a marking policy only.

The router also supports L2 ingress marking.

For ingress marking:

Ingress traffic— For the ingress pop operation, re-marking the customer VLAN tag (CoS, DEI) is not supported.

Egress traffic— The ingress ‘pop VLAN’ is translated to a ‘push VLAN’ for the egress traffic, and (CoS, DEI) marking is supported for newly pushed VLAN tags. If two VLAN tags are pushed to the packet header at the egress side, both inner and outer VLAN tags are marked. For example:

1. rewrite ingress tag pop 1 symmetric

2. rewrite ingress tag pop 2 symmetric

3. rewrite ingress tag translate 2-to-1 dot1q/dot1ad <> symmetric

Limitation

The statistics and counters for the egress marking policy cannot be viewed on the router.

Supported Packet Marking Operations

This table shows the supported packet marking operations.

Supported Mark Types

Range

Support for Unconditional Marking

Support for Conditional Marking

set dscp

0-63

ingress

No

set qos-group

0-7

ingress

No

Class-based Unconditional Packet Marking

The packet marking feature allows you to partition your network into multiple priority levels or classes of service, as follows:

  • Use QoS unconditional packet marking to set the IP precedence or IP DSCP values for packets entering the network. Routers within your network can then use the newly marked IP precedence values to determine how the traffic should be treated.

    On ingress direction, after matching the traffic based on either the IP Precedence or DSCP value, you can set it to a particular discard-class. Weighted random early detection (WRED), a congestion avoidance technique, thereby uses discard-class values to determine the probability that a packet is dropped.

  • Use QoS unconditional packet marking to assign MPLS packets to a QoS group. The router uses the QoS group to determine how to prioritize packets for transmission. To set the QoS group identifier on MPLS packets, use the set qos-group command in policy map class configuration mode.


    Note
    Setting the QoS group identifier does not automatically prioritize the packets for transmission. You must first configure an egress policy that uses the QoS group.

Note

  • Conditional packet marking is not supported.

  • Unless otherwise indicated, the class-based unconditional packet marking for Layer 3 physical interfaces applies to bundle interfaces.

Bundle Traffic Policies

A policy can be bound to bundles. When a policy is bound to a bundle, the same policy is programmed on every bundle member (port). For example, if there is a policer or shaper rate, the same rate is configured on every port. Traffic is scheduled to bundle members based on the load balancing algorithm.

Both ingress and egress traffic is supported. Percentage-based policies are supported.

For details, see Configure QoS on Link Bundles.

Restrictions


Note

The router has a single core,hence the per core scale is applicable.

Example: For Default Configuration, which is Normal (2 counter mode) QoS Mode & 32 Class Map-Size, you can configure 128 interfaces with Ingress Policy per core.

Other restrictions to follow:

  • If you have a set traffic class statement explicitly configured in ingress service policy, it is mandatory to have a corresponding match traffic class on egress for the traffic to be correctly matched and the stats to be accounted in show policy-map interface <> output command. To match the ingress traffic to egress class-default, traffic class should be set to 0 on ingress.

  • If you have a set traffic class configured in Ingress service policy, and no corresponding match traffic class on egress, the traffic will not go to class default and the stats for this traffic flow will not be seen in show policy-map interface <> output command.

  • If you do not have any set traffic class statement in ingress, then traffic will hit the default-class on egress.

  • If you have a set discard-class statement configured in ingress service policy, it is mandatory to have a corresponding match discard-class on egress for the traffic to be correctly matched and the stats to be accounted in show policy-map interface <> output command.

  • If you have a set discard-class statement configured in ingress service policy and do not have a corresponding match discard-class on egress, the traffic will not hit the class-default and the stats for this flow will not be accounted in show policy-map interface <> output command.

  • The system does not support class-map size on peering mode.

Restrictions for QoS on BVI

  • The system does not support egress QoS policy on BVI.

  • If you apply L3 ingress QoS policy on L2 interface, which is a part of the same bridge-domain as BVI, the classification might not work if packets are destined to the BVI MAC address.

  • If a QoS policy is attached to BVI, the policy is inherited by the L2 interfaces, which are part of the same bridge-domain. Hence, any other policy cannot be applied on the L2 interfaces. Similarly, if a QoS policy is attached to any of the L2 interfaces, any QoS policy cannot be applied on the BVI, which is part of the same bridge-domain.

In-Place Policy Modification

The In-Place policy modification feature allows you to modify a QoS policy even when the QoS policy is attached to one or more interfaces. A modified policy is subjected to the same checks that a new policy is subject to when it is bound to an interface. If the policy-modification is successful, the modified policy takes effect on all the interfaces to which the policy is attached. However, if the policy modification fails on any one of the interfaces, an automatic rollback is initiated to ensure that the pre-modification policy is in effect on all the interfaces.

You can also modify any class map used in the policy map. The changes made to the class map take effect on all the interfaces to which the policy is attached.


Note

  • The QoS statistics for the policy that is attached to an interface are lost (reset to 0) when the policy is modified.

  • When a QoS policy attached to an interface is modified, there might not be any policy in effect on the interfaces in which the modified policy is used for a short period of time.

  • The system does not support the show policy-map statistics for marking policies.

  • An in-place modification of an ACL does not reset the policy-map statistics counter.

Note

  • For QOS EXP-Egress marking applied on L3 interface, there is a limit of 3 unique policy-maps per NPU. When the maximum limit for policy-maps is reached and you try to modify a policy-map which is shared between different interfaces, you may get an error.

  • For QOS egress marking (CoS, DEI) applied on L2 interface, there is a limit of 13 unique policy-maps per NPU. When the maximum limit for policy-maps is reached and you try to modify a policy-map which is shared between different interfaces, you may get an error

Verification

If unrecoverable errors occur during in-place policy modification, the policy is put into an inconsistent state on target interfaces. No new configuration is possible until the configuration session is unblocked. It is recommended to remove the policy from the interface, check the modified policy and then re-apply accordingly.

References for Modular QoS Service Packet Classification

Specification of the CoS for a Packet with IP Precedence

Use of IP precedence allows you to specify the CoS for a packet. You can create differentiated service by setting precedence levels on incoming traffic and using them in combination with the QoS queuing features. So that, each subsequent network element can provide service based on the determined policy. IP precedence is usually deployed as close to the edge of the network or administrative domain as possible. This allows the rest of the core or backbone to implement QoS based on precedence.

Figure 1. IPv4 Packet Type of Service Field

You can use the three precedence bits in the type-of-service (ToS) field of the IPv4 header for this purpose. Using the ToS bits, you can define up to eight classes of service. Other features configured throughout the network can then use these bits to determine how to treat the packet in regard to the ToS to grant it. These other QoS features can assign appropriate traffic-handling policies, including congestion management strategy and bandwidth allocation. For example, queuing features such as LLQ can use the IP precedence setting of the packet to prioritize traffic.

IP Precedence Bits Used to Classify Packets

Use the three IP precedence bits in the ToS field of the IP header to specify the CoS assignment for each packet. You can partition traffic into a maximum of eight classes and then use policy maps to define network policies in terms of congestion handling and bandwidth allocation for each class.

Each precedence corresponds to a name. IP precedence bit settings 6 and 7 are reserved for network control information, such as routing updates. These names are defined in RFC 791.

IP Precedence Value Settings

By default, the routers leave the IP precedence value untouched. This preserves the precedence value set in the header and allows all internal network devices to provide service based on the IP precedence setting. This policy follows the standard approach stipulating that network traffic should be sorted into various types of service at the edge of the network and that those types of service should be implemented in the core of the network. Routers in the core of the network can then use the precedence bits to determine the order of transmission, the likelihood of packet drop, and so on.

Because traffic coming into your network can have the precedence set by outside devices, we recommend that you reset the precedence for all traffic entering your network. By controlling IP precedence settings, you prohibit users that have already set the IP precedence from acquiring better service for their traffic simply by setting a high precedence for all of their packets.

The class-based unconditional packet marking and LLQ features can use the IP precedence bits.

IP Precedence Compared to IP DSCP Marking

If you need to mark packets in your network and all your devices support IP DSCP marking, use the IP DSCP marking to mark your packets because the IP DSCP markings provide more unconditional packet marking options. If marking by IP DSCP is undesirable, however, or if you are unsure if the devices in your network support IP DSCP values, use the IP precedence value to mark your packets. The IP precedence value is likely to be supported by all devices in the network.

You can set up to 8 different IP precedence markings and 64 different IP DSCP markings.

Usage of QoS-group and Queue Selection

The router supports up to 8 CoSQs for each egress interface, in the range of 0 through 7, with 0 being the default CoSQ. The qos-group value is used to select a CoSQ and eventually a virtual output queue (VOQ).

In order to designate the traffic class to a certain CoSQ other than CoSQ 0, in the ingress policy-map, you must explicitly configure set qos-group x command in the class-map, where 'x' is the CoSQ value.

In the egress policy-map, a class-map with a corresponding match qos-group x allows further QoS actions to be applied to the traffic class.

For example, 

class-map prec1
  match prec 1
 
policy-map test-ingress
  class prec1
    set qos-group 1
    police rate percent 50
  
class-map qg1
  match qos-group 1
 
policy-map test-egress
  class qg1
    set cos 1