New and Changed Information

Table 1. New Features and Changed Behavior in Cisco APIC

Cisco APIC Release Version

Feature or Update

Where documented

Release 6.2(2)

ECN Buffer Tuning

You can now configure ECN buffer threshold parameters at the fabric or node level. This allows fine-tuning of congestion management to improve throughput under bursty traffic conditions.

For more information, see ECN Buffer Tuning in ACI Fabric

Release 6.1(4)

Support for DSCP-based PFC

Cisco ACI QoS Level Settings

Release 6.0(2)

View the QoS Interface Statistics using the APIC GUI

 Troubleshooting Cisco APIC QoS Policies

Release 5.1(3)

Added support for Nexus 9300-FX3 platform switches for ROCEv2

 RoCEv2 QoS

Release 5.1(3)

ICMP replies with the same Class of Service (CoS) value that was sent in the request.

 CoS Preservation Guidelines and Limitations

Release 5.0(1)

Support for custom SR-MPLS QoS policies.

SR-MPLS QoS

Release 4.0(2)

Additional information on QoS behavior in Cisco ACI fabrics.

Cisco ACI QoS Overview

Release 4.0(1)

Support for new QoS settings to enable RoCEv2 technology in Cisco APIC environment.

Support for additional custom QoS levels and L3Out configuration.

RoCEv2 QoS

Custom QoS

QoS for L3Outs

Multipod QoS

Release 3.1(2)

Enhanced QoS policy enforcement on L3Out ingress traffic.

QoS for L3Outs

Release 2.1(1)

Support for Cisco ACI fabric to classify the traffic for devices that classify the traffic based only on the CoS value

Custom QoS

Release 2.0(2)

Support for CoS preservation and DSCP Multipod QoS settings was added for multipod topologies.

Multipod QoS

Cisco ACI QoS Overview

Cisco ACI Quality of Service (QoS) feature allows you to classify the network traffic in your fabric and then to prioritize and police the traffic flow to help avoid congestion in your network. When traffic is classified within the fabric, it is assigned a QoS Priority Level, which is then used throughout the fabric to provide the most desirable flow of packets through the network.

Any traffic for which the QoS features are enabled undergoes the following stages:

  • Classification – identification of the traffic type and assignment of a Cisco ACI QoS Level based on it.

  • Policing – control of the traffic based on its classification.

  • Marking – tagging of network packets based on the configured policing rules and its behavior.

  • Queuing and Scheduling - prioritization and/or isolation of network packets based on their QoS Level and markings.

The following sections provide more detailed information on each of the QoS process flow stages.

Classification and Marking

Traffic classification is used to partition traffic in your Cisco ACI fabric into QoS Levels based on several criteria such as ingress packet headers (DSCP or CoS), source EPGs, or EPG Contracts.

The values used to classify traffic are called match criteria. When you configure a QoS Level for a type of traffic, you can specify one or more of these criteria to match, you can choose to exclude a particular criteria, or you can determine the traffic class by matching any or all criteria. Traffic that fails to match any class is assigned to a default class (Level3) of traffic.

When packets first ingress the Cisco ACI fabric, two values can be use to classify the traffic into the proper QoS Level:

  • Class of Service (CoS): Also referred to as "dot1p value", a QoS feature developed by the 802.1p group that uses a 3-bit Priority Code Point (PCP) inside the Layer-2 Ethernet frames to differentiate traffic.

  • Differentiated Services Code Point (DSCP): A Layer-3 alternative to CoS that uses a 6-bit value in the IP packet header to classify traffic.

Marking

After traffic is classified, the packets are marked by adding the QoS class ID to the outer header of each packet. Traffic classification and marking happens on the ingress leaf switches only; the spine and egress leaf switches only map the packets to proper class of service based on the CoS value.

Policing

While Cisco ACI fabrics are non-blocking if properly sized and there are no oversubscription concerns, a leaf interface may still be shared between multiple EPGs. Applying proper QoS policies can prevent one EPG from monopolizing the link.

One of the common use-cases is to classify the traffic coming from a given server to EPGs, for example as data, backup, or vMotion. Following the classification, you can then police the ingress traffic for each EPG to ensure that backup traffic does not consume too much bandwidth and interfere with the data traffic. Using this type of ingress per-EPG policing, we can provision different limits for data EPG, backup EPG, and vMotion EPG on any given leaf switch interface.

When you configure QoS policing in your fabric, the following rules apply:

  • The policies can be applied on interfaces or EPGs.

    Interface policies are defined at the tenant level and can be applied in both, ingress and egress directions. Because these policies are attached to a port, they are enforced globally with no concept of individual EPGs.

    EPG policies are defined at the tenant level and can be applied only in the ingress direction. Because these policies are attached to an EPG, they are enforced at the physical interface level per EPG. You can configure a single policer instance to be used by all EPG members or a dedicated policer for each member.

  • Policies can be applied from the fabrics access (infra) or the tenant (fvTenant) portions of the fabric.

  • If any traffic exceeds the limits configured in the policies, the packets can be either dropped or marked.

Queuing and Scheduling

After the traffic packets have been classified (or re-classified based on markings) and assigned a QoS Level, they are subject to being queued for transmission. Multiple queues can be used based on the packet's priority and a scheduling algorithm is used to determine which queue's packet is to be transmitted next.

Cisco ACI uses a Deficit Weighted Round Robin (DWRR) scheduling algorithm. This scheduling algorithm allows packets of variable sizes and provides a deficit counter to dynamically adjust queue priorities. The queuing and scheduling policy is a fabric-wide configuration and applies to all nodes. The same policy is applied within each node whenever packet queuing takes place, which simplifies the configuration and ensures consistent end-to-end compatibility with standard QoS, such as in the NX-OS-mode switches.

Cisco ACI fabric supports a number of user-configurable QoS levels as well as levels reserved for fabric control traffic, SPAN, and traceroute traffic. Cisco APIC release 4.0(1) supports six user-configurable QoS levels, while earlier releases supported three.

The following table lists the user-configurable QoS levels:

Table 2. Cisco APIC User-Configurable QoS Levels

Class of Service

QoS Group Used by DCBX (ETS configuration and ETS recommendation)1

Traffic Type

Dot1p (CoS) Marking in VXLAN Header

DEI Bit2

0

0

Level 3 (default)

0

0

1

1

Level 2

1

0

2

2

Level 1

2

0

4

7

Level 6

2

1

5

6

Level 5

3

1

6

5

Level 4

5

1

3

3

APIC Controller

3

0

9

Not Advertised

SPAN

4

0

8 (SUP)

4

Control

5

0

8 (SUP)

4

Traceroute

6

0

7

Not Advertised

Copy Service

7

0

1 In the IEEE DCBX PFC configuration LLDP TLV, the priority value is the associated CoS value regardless of which PFC level (1 through 6) is enabled.

2 The Drop Eligible Indicator (DEI) bit is a 1-bit field that is used to indicate frames eligible to be dropped during traffic congestion. The CoS value (3 bits) + DEI value (1 bit) represents the QoS class.

The following table lists the reserved QoS levels, with each level being mapped to a hardware queue and configured at the fabric level:

Table 3. Cisco APIC Reserved QoS Groups

Traffic Type

Description

APIC Controller

Strict priority queue, includes all traffic to and from APIC.

SPAN

Best effort traffic. A Deficit Weighted Round Robin (DWRR) queue with least weight. SPAN and ERSPAN traffic has lower priority than data traffic and will be dropped in case of congestion.

Control

Strict priority queue, includes all SUP-generated traffic and control traffic, such as LACP, ISIS, BGP, and COOP.

Traceroute

Best effort traffic.

Scheduling and Congestion Avoidance

If at any point the network becomes congested, a congestion avoidance algorithm can be used to determine which packets to transmit, queue, or drop. Cisco APIC deploys two different congestion avoidance algorithms for user-configurable QoS Levels:

  • Tail Drop (TD): In case of congestion, any new incoming packets (tail end of a queue) are dropped. Tail Drop uses single threshold per queue.

  • Weighted Random Early Detection (WRED): Provides an early detection mechanism, which allows for low priority packets to be preemptively dropped in order to protect higher priority queues from congestion. WRED uses one or more thresholds per queue with each queue associated with DSCP or CoS values.

Switch Roles in QoS Flow

When you enable QoS features, the fabric's switches perform the following QoS-related tasks:

Table 4.

Switch

Task

Ingress leaf switch

  • Classification

  • Marking

  • Buffering

  • Queuing

  • Ingress Policing

Spine switch

  • Buffering

  • Queuing

Egress leaf switch

  • Buffering

  • Queuing

  • Egress Policing

Cisco ACI QoS Policy Precedence

Once traffic has been classified, you can use the QoS classes to prioritize flow within you fabric by assigning a QoS level to EPG traffic as described in more detail in the following sections. However, keep in mind that if multiple QoS policies are configured and could apply for any given traffic, only one policy is applied using the following precedence:

  • QoS policy for EPG Contract

    If QoS is enabled in the Contract between EPGs, the QoS class specified in the contract is used.

  • QoS policy for source EPG

    If QoS is not enabled in the Contract, but custom QoS is enabled at the source EPG level, the custom QoS class is used and traffic is classified based on DSCP or 802.1p values.

  • Default QoS class

    If no QoS class is specified, the traffic is assigned Level3 QoS class by default.

Cisco ACI QoS Level Settings

Cisco ACI provides a number of user-configurable QoS levels. Cisco APIC Release 4.0(1) supports six user-configurable QoS levels, while earlier releases supported three. The following sections describe how to configure specific settings for each of these levels.

Configuring Cisco ACI QoS Level Settings Using the Cisco APIC GUI

This section describes how to configure specific settings for each Cisco ACI QoS level.

Procedure

Step 1

From the main menu bar, select Fabric > Access Policies.

Step 2

In the left-hand navigation pane, select Policies > Global > QOS Class > <level>.

You can configure the following settings for each QoS Level:

Name Description

Admin State

The policy administrative state. The state can be:

  • Enabled (Default)

  • Disabled

MTU

This field is unused. You cannot modify the MTU value in a QoS policy.

Minimum Buffers

The minimum number of reserved buffers. The number can be between 0 and 3.

The default value is 0.

Congestion Algorithm

The congestion algorithm used for this QoS Level. The congestion algorithm can be:

  • Tail Drop

  • Weighted random early detection

Congestion Notification

(Weighted random early detection algorithm only)

Indicates the state of Explicit Congestion Notification (ECN) setting. Enabling Congestion Notification causes the packets that would be dropped to be ECN-marked instead. The state can be:

  • Enabled

  • Disabled

The default setting is Disabled.

This setting is used for the RoCEv2 feature described in more detail in RoCEv2 and the Required APIC QoS Settings.

Min Threshold (percentage)

(Weighted random early detection algorithm only)

The minimum queue threshold as a percentage of the maximum queue length for WRED algorithm.

If the average queue size is below the minimum threshold value, the arriving packets are queued immediately.

This setting is used for the RoCEv2 feature described in more detail in RoCEv2 and the Required APIC QoS Settings.

Max Threshold (percentage)

(Weighted random early detection algorithm only)

The maximum queue threshold as a percentage of the maximum queue length for WRED algorithm.

If the average queue size is greater than the maximum threshold value, the arriving packets are dropped.

This setting is used for the RoCEv2 feature described in more detail in RoCEv2 and the Required APIC QoS Settings.

Probability (percentage)

(Weighted random early detection algorithm only)

The probability value for WRED algorithm.

The probability determines whether the packet is dropped or queued when the average queue size is between the minimum and the maximum threshold values.

This setting is used for the RoCEv2 feature described in more detail in RoCEv2 and the Required APIC QoS Settings.

Weight

(Weighted random early detection algorithm only)

The weight value for WRED algorithm.

Weight has a range of 0 to 7 and is used to calculate average queue length. Lower weight prioritizes current queue length, while higher weight prioritizes older queue lengths.

This setting is used for the RoCEv2 feature described in more detail in RoCEv2 and the Required APIC QoS Settings.

Scheduling algorithm

The scheduling algorithm used for this QoS Level. The scheduling algorithm can be:

  • Strict priority

  • Weighted round robin (Default): Use this algorithm if your configuration requires IEEE ETS (Enhanced Transmission Selection) that is specified by the IEEE 802.1Qaz standard. IEEE ETS is required for Azure Stack HCI support.

Bandwidth allocated (in %)

The percentage of total bandwidth allocated to this QoS Level. The value can be between 0 and 100.

The default value is 20.

PFC Admin State

The administrative state of the Priority Flow Control policy applied to FCoE traffic. The state can be:

  • Enabled: Enables Priority Flow Control for FCoE traffic.

  • Disabled: Disables Priority Flow Control for FCoE traffic.

No-Drop-CoS

The CoS level to impose no drop FCoE packet handling even in case of FCoE traffic congestion. The options are:

  • cos 0

  • cos 1

  • cos 2

  • cos 3

  • cos 4

  • cos 5

  • cos 6

  • cos 7

  • Unspecified: Used to avoid imposing no drop.

Scope

The Priority Flow Control (PFC) scope.

  • Fabric-wide PFC: Across the entire fabric.

  • IntraTor PFC: Across the leaf switch only.

PFC No-Drop DSCP

Enable the No-Drop dscp value at the Qos class Level. This is a global configuration for the Qos class level to impose no drop packet handling in case of congestion. Match values range from DSCP 0 to 63. Multiple DSCP matches are supported at same time. For example, you could select DSCP0, DSCP1, and DSCP60.as DSCP values from the drop-down box.

Note

 

You can configure up to two No-drop classes per interface. You must ensure that both the No-drop classes have the No-drop DSCP policy enabled or both not enabled.

If the DSCP policy is configured globally at the interface level and PFC No-Drop-DSCP value is also configured but the packet matches with the CoS No-Drop value, the Cos classification will not work as expected. However if packets comes without IP header then the CoS match will work.

Step 3

Click Submit to save the changes.

Configuring the Interface No-Drop DSCP Match Policy

You can use the Cisco APIC GUI to configure the required QoS options to enable support for RoCEv2 in your fabric. Use this option to enable the PFC on each interface to enable No-Drop DSCP match policy for front-panel ports (TOR). The No-Drop DSCP match policy setting can be set to either on or off.

Procedure

Step 1

Log in to your Cisco APIC GUI.

Step 2

Navigate to Fabric > Access Policies > Policies > Interface > No-drop DSCP Match Policy.

Step 3

In the Properties panel, to enable the DSCP policy, specify the following information:

  • Name: Provide the name of the No Drop DSCP match policy.

  • (Optional) Description: Provide an optional description of the No Drop DSCP match policy.

  • No-Drop DSCP Match Mode: The No-drop DSCP Match Policy setting can either be set to Off or On.

Step 4

Click Submit to save the changes.

Step 5

To attach the custom No-Drop PFC policy:

  1. Navigate to Fabric > Access Policies > Interfaces> Policy Groups> > Leaf Access Port / PC Interface/ VPC Interface>.

  2. Attach the custom No-Drop PFC policy you created.

  3. Click Submit to save the changes.

Configuring Cisco ACI QoS Level Settings Using the NX-OS-Style CLI

This section describes how to configure specific settings for each Cisco ACI QoS level.

Procedure

Step 1

Enter configuration mode.

Example:

apic1# config

Step 2

Choose the QoS Level you want to configure.

In the following command, replace level2 with the QoS Level you want to configure:

Example:

apic1(config)# qos parameters level2

Step 3

Configure one or more settings for the QoS Level.

The following example shows how to configure congestion notification and congestion detection algorithm for a QoS level:

Example:

apic1(config-qos)# algo wred
apic1(config-qos-algo)# ecn enabled
apic1(config-qos-algo)# maxthreshold 60
apic1(config-qos-algo)# minthreshold 40
apic1(config-qos-algo)# probability 0
apic1(config-qos-algo)# weight 1
apic1(config-qos-algo)# exit

The following example shows how to configure no-drop CoS:

Example:

apic1(config-qos)# pause no-drop cos 1 fabric
apic1(config-qos-algo)#

Configuring Cisco ACI QoS Level Settings Using the REST API

This section describes how to configure specific settings for each Cisco ACI QoS level.

Procedure

Configure settings for a QoS level.

In the following example, replace level2 with the QoS class you want to configure.

POST URL: https://<apic-ip>/api/node/mo/uni.xml

Example:

<qosClass admin="enabled" dn="uni/infra/qosinst-default/class-level2" prio="level2">
    <qosCong algo="wred" wredMaxThreshold="60" wredMinThreshold="40" wredProbability="0"
             ecn="enabled"/>
    <qosPfcPol name="default" noDropCos="cos0" adminSt="yes" enableScope="fabric"/>
</qosClass>

Custom QoS Policy and Ingress/Egress Markings

You can create a custom QoS policy in Cisco APIC by translating the DSCP and CoS values of the ingressing traffic to a QoS priority level to be used inside the Cisco ACI fabric. Translation is supported only if the DSCP values are present in the IP packet and CoS values are present in the Ethernet frames.

As an example, custom QoS policies allow you to classify traffic coming into the Cisco ACI fabric traffic from devices that classify the traffic based only on the CoS value, such as Layer-2 packets which do not have an IP header.

Custom QoS Guidelines and Limitations

If you create custom QoS policies based on both, CoS and DSCP, values and both values are present in an ingressing packet but are matched to different QoS priority levels, the DSCP mapping takes precedence.

Custom QoS policies based on DSCP value translation require 5 continuous chunks of TCAM memory space per DSCP translation policy. If continuous memory space is not available, the DSCP translation policy will fail to program in the hardware and a fault will be generated on the APIC. You can verify available TCAM space using the following command on the switch: 

show system internal aclqos qos policy detail

Alternatively, you can verify available TCAM space using the following vsh_lc shell command:

vsh_lc -c 'show system internal aclqos qos policy detail'

If you create a custom QoS policy based on CoS value, you must first enable the global fabric CoS preservation policy, as described in Class of Service (CoS) Preservation for Ingress and Egress Traffic.

If you are running a release prior to release 4.0(1), CoS translation is not supported on external Layer 3 interfaces.

CoS translation is supported only if the egress frame is 802.1Q encapsulated.

CoS translation is not supported when the following configuration options are enabled:

  • Contracts are configured that include QoS.

  • The outgoing interface is on a Fabric Extender (FEX) because FEX follows a static mapping table based on VLAN CoS or (dot1p value) received. Use Cisco ACI Multi-Pod instead of CoS Preservation Policy for egress QoS classification on FEX Host Interfaces (HIF) ports. For more information, see Multi-Pod QoS and DSCP Translation Policy.

  • Multipod QoS using a DSCP policy is enabled.

    For more information about Multipod and DSCP policy, see Multi-Pod QoS and DSCP Translation Policy.

  • Dynamic packet prioritization is enabled.

  • If an EPG is configured with intra-EPG endpoint isolation enforced.

  • If an EPG is configured with microsegmentation enabled.

  • Starting with release 4.0(1), all DPP prioritized traffic has CoS 3 marked in spite of custom QoS configuration.

    Only in the 4.0 releases, when these packets are ingressing and egressing same leaf switch, the CoS value is retained, leading to the frames leaving the Fabric with CoS 3 marking.

Creating Custom QoS Policy Using Cisco APIC GUI

This section describes how to create a custom QoS policy and associate it with an EPG using the Cisco APIC GUI.

Before you begin

You must have created the tenant, application, and EPGs that will consume the custom QoS policy.

Procedure

Step 1

Log in to your Cisco APIC GUI.

Step 2

From the horizontal navigation bar, select Tenants > <tenant-name>.

Step 3

In the left-hand navigation pane, expand <tenant-name> > Policies > Protocol > Custom QoS.

Step 4

Right click the Custom QoS and choose Create Custom QoS Policy.

Step 5

Provide the name and an optional description of the custom QoS policy information.

Step 6

Create a DSCP mapping for one or more QoS priority levels.

The DSCP mapping allows you to map ingress DSCP values to a QoS priority level as well as the egress DSCP and CoS value for traffic that leaves the ACI fabric. For each mapping, you can specify the following fields:

Name Description
Priority The QoS priority level to which the DSCP values will be mapped.
DSCP Range From The start of the DSCP range.
DSCP Range To The end of the DSCP range.
DSCP Target The DSCP value for egressing traffic.
Target Cos The CoS value for the egressing traffic.

Step 7

Create a CoS mapping for one or more QoS priority levels.

The CoS mapping allows you to map ingress CoS values to a QoS priority level as well as the egress DSCP and CoS value for traffic that leaves the ACI fabric. For each mapping, you can specify the following fields:

Name Description
Priority The QoS priority level to which the DSCP values will be mapped.
Dot1P Range From The start of the CoS range.
Dot1P Range To The end of the CoS range.
DSCP Target The DSCP value for egressing traffic.
Target Cos The CoS value for the egressing traffic.

Step 8

Click Submit to save the changes.

Step 9

Attach the custom QoS policy you created to an EPG.

  1. Navigate to Tenants > <tenant-name> > Application Profiles > <application-profile-name> > Application EPGs > <application-epg-name>.

  2. In the main window pane, select Policy > Generalthe custom QoS policy you created.

  3. In the main window pane, select the custom QoS policy you created from the Custom QoS drop down menu.

  4. Click Submit to save the changes.

Creating Custom QoS Policy Using NX-OS Style CLI

This section describes how to create a custom QoS policy and associate it with an EPG using the NX-OS style CLI.

Before you begin

You must have created the tenant, application, and EPGs that will consume the custom QoS policy.

Procedure

Step 1

Enter configuration mode.

apic1# configure

Step 2

Enter tenant configuration mode.

apic1(config)# tenant <tenant-name>

Step 3

Create QoS policy.

apic1(config-tenant)# policy-map type qos <qos-policy-name>

Step 4

Set DCSP range and target QoS priority level.

apic1(config-tenant-pmap-qos)# match dscp AF23 AF31 set-cos 6

Step 5

Return to tenant configuration mode. 

apic1(config-tenant-pmap-qos)# exit

Step 6

Create or edit an application profile.

apic1(config-tenant)# application <application-name>

Step 7

Create or edit an EPG in the application profile.

To create a normal EPG:

apic1(config-tenant-app)# epg <epg-name>

To create an external Layer-2 EPG:

apic1(config-tenant)# external-l2 epg <ext-l2-epg-name>

Step 8

Associate the QoS policy with the EPG.

The system prompt may be different depending on whether you create a normal EPG or an external EPG.
apic1(config-tenant-app-epg)# service-policy <qos-policy-name>

Step 9

Return to the tenant configuration mode.

apic1(config-tenant-app-epg)# exit

Creating Custom QoS Policy Using REST API

This section describes how to create a custom QoS policy and associate it with an EPG using the REST API.

Before you begin

You must have created the tenant, application, and EPGs that will consume the custom QoS policy.

Procedure

Step 1

Create a custom QoS policy.

<qosCustomPol name="vrfQos001" dn="uni/tn-t001/qoscustom-vrfQos001">
    <qosDscpClass to="AF31" targetCos="6" target="unspecified"
                  prio="unspecified" from="AF23"/>
    <qosDot1PClass to="1" targetCos="6" target="unspecified"
                   prio="unspecified" from="0"/>
</qosCustomPol>

Step 2

Associate the policy with an EPG that will consume it.

<fvAEPg prio="unspecified" prefGrMemb="exclude" pcEnfPref="unenforced"
        name="ep2" matchT="AtleastOne" isAttrBasedEPg="no" fwdCtrl=""
        dn="uni/tn-t001/ap-ap2/epg-ep2">
    <fvRsDomAtt tDn="uni/vmmp-VMware/dom-vs1" resImedcy="lazy" 
                primaryEncap="unknown" netflowPref="disabled"
                instrImedcy="lazy" encapMode="auto" encap="unknown"
                delimiter="" classPref="encap"/>
    <fvRsCustQosPol tnQosCustomPolName="vrfQos001"/>
    <fvRsBd tnFvBDName="default"/>
</fvAEPg>

Class of Service (CoS) Preservation for Ingress and Egress Traffic

When traffic enters the Cisco ACI fabric, each packet's priority is mapped to a Cisco ACI QoS level. These QoS levels are then stored in the CoS field and DEI bit of the packet's outer header while the original headers are discarded.

If you want to preserve the original CoS values of the ingressing packets and restore it when the packet leaves the fabric, you can enable the 802.1p Class of Service (CoS) preservation using a global fabric QoS policy as described in this section.

The CoS preservation is supported in single pod and multipod topologies. However, in multipod topologies, CoS preservation can be used only when you are not concerned with preserving the settings in the IPN between pods. To preserve the CoS values of the packets as they are transiting the IPN, use the DSCP translation policy as described in Multi-Pod QoS and DSCP Translation Policy.

Class of Service Preservation and DSCP

The following table shows the overlay differentiated services code point (DSCP) values and corresponding Cisco Application Centric Infrastructure (ACI) quality of service (QoS) and ingress class of service (CoS) values:

Table 5. Overlay DSCP, QoS, and CoS

Overlay DSCP Value

ACI QoS Level

Ingress CoS

0

Level 3

0

1

Level 2

0

2

Level 1

0

8

Level 3

1

9

Level 2

1

10

Level 1

1

11

Level 4

1

12

Level 5

1

13

Level 6

1

14

Level 2

7

15

Level 1

7

16

Level 3

2

17

Level 2

2

18

Level 1

2

19

Level 4

2

20

Level 5

2

21

Level 6

2

22

Level 4

0

23

Level 6

7

24

Level 3

3

25

Level 2

3

26

Level 1

3

27

Level 4

3

28

Level 5

3

29

Level 6

3

30

Level 4

7

32

Level 3

4

33

Level 2

4

34

Level 1

4

35

Level 4

4

36

Level 5

4

37

Level 6

4

38

Level 5

0

40

Level 3

5

41

Level 2

5

42

Level 1

5

43

Level 4

5

44

Level 5

5

45

Level 6

5

46

Level 5

7

48

Level 3

6

49

Level 2

6

50

Level 1

6

51

Level 4

6

52

Level 5

6

53

Level 6

6

54

Level 6

0

56

Level 3

7
Table 6. Control Plane Overlay DSCP, QoS, and CoS

Overlay DSCP Value

ACI QoS Level

Ingress CoS

3

DPP

x

4

SPAN

x

5

Sup

x

6

Traceroute

x

7

Copy Service

x

57

Multicast HREP Drop

x

58

Multi-Site Endpoint Announce

x

59

Flood in Encapsulation

x

60

Q-in-Q

x

61

Proxy ARP

x

62

Copy Service

x

63

Isolated EPG

x

CoS Preservation Guidelines and Limitations

The following guidelines and limitations apply for Class of Service (CoS) preservation:

  • Only the CoS value within a VLAN header is preserved, the DEI bit is not preserved.

  • For VXLAN encapsulated packets, the CoS value contained in the outer header is not preserved.

  • CoS values are not preserved when the following configuration options are enabled:

    • Contracts are configured that include QoS.

    • The outgoing interface is on a Fabric Extender (FEX) because FEX follows a static mapping table based on VLAN CoS or (dot1p value) received. Use Cisco ACI Multi-Pod instead of CoS Preservation Policy for egress QoS classification on FEX Host Interfaces (HIF) ports. For more information, see Multi-Pod QoS and DSCP Translation Policy.

    • Traffic is flowing from an EPG with isolation enforced to an EPG without isolation enforced.

    • A DSCP QoS policy is configured on a VLAN EPG and the packet has an IP header.

      DSCP marking can be set at the filter level on the following with the precedence order from the innermost to the outermost:

      • Contract

      • Subject

      • In Term

      • Out Term


      Note

      When specifying vzAny for a contract, external EPG DSCP values are not honored because vzAny is a collection of all EPGs in a VRF and EPG-specific configuration cannot be applied. If EPG-specific target DSCP values are required, then the external EPG should not use vzAny.

  • Beginning in Cisco APIC release 5.1(3), ICMP replies with the same Class of Service (CoS) value that was sent in the request.

  • For no-drop-CoS, only use 0-5 and 7 as no-drop. Do not use CoS level 6 as FIP packets will not get forwarded to peer device and FLOGI will not get established.

Enabling Global No-Drop DSCP Match Using GUI

This section describes how to enable the global no-drop DSCP match to ensure that QoS priority settings are handled the same for traffic entering and transiting a single-pod fabric as for traffic entering one pod and egressing another in a multipod fabric.

Procedure

Step 1

From the main menu bar, select Fabric > Access Policies.

Step 2

In the left-hand navigation pane, select Policies > Global > QOS Class.

Step 3

In the Global - QOS Class main window pane, check the No-Drop DSCP match control: enable nodrop dscp match checkbox.

Step 4

Click Submit to save the changes..

Enabling CoS Preservation Using GUI

This section describes how to enable CoS preservation to ensure that QoS priority settings are handled the same for traffic entering and transiting a single-pod fabric as for traffic entering one pod and egressing another in a multipod fabric.


Note

Enabling CoS preservation applies a default CoS-to-DSCP mapping to the various traffic types.

Procedure

Step 1

From the main menu bar, select Fabric > Access Policies.

Step 2

In the left-hand navigation pane, select Policies > Global > QOS Class.

Step 3

In the Global - QOS Class main window pane, check the Preserve COS: Dot1p Preserve checkbox.

Step 4

Click Submit to save the changes..

Enabling CoS Preservation Using CLI

This section describes how to enable CoS preservation to ensure that QoS priority settings are handled the same for traffic entering and transiting a single-pod fabric as for traffic entering one pod and egressing another in a multipod fabric.


Note

Enabling CoS preservation applies a default CoS-to-DSCP mapping to the various traffic types.

Procedure

Step 1

Enter configuration mode.

apic1# configure

Step 2

Enables CoS preservation.

apic1(config)# qos preserve cos

Enabling CoS Preservation Using REST API

This section describes how to enable CoS preservation to ensure that QoS priority settings are handled the same for traffic entering and transiting a single-pod fabric as for traffic entering one pod and egressing another in a multipod fabric.


Note

Enabling CoS preservation applies a default CoS-to-DSCP mapping to the various traffic types.

Procedure

Enable CoS preservation.

POST https://<apic-ip>/api/node/mo/uni/infra/qosinst-default.xml

<qosInstPol name="default" dn="uni/infra/qosinst-default" ctrl="dot1p-preserve"/>

Disable CoS preservation.

<qosInstPol name="default" dn="uni/infra/qosinst-default" ctrl=""/>

Multi-Pod QoS and DSCP Translation Policy

When traffic is sent and received within the Cisco ACI fabric, the QoS Level is determined based on the CoS value of the VXLAN packet's outer header. In Multi-Pod topologies, where devices that are not under Cisco APIC's management may modify the CoS values in the transiting packets, you can preserve the QoS Level setting by creating a mapping between the Cisco ACI and the DSCP value within the packet.

If you are not concerned with preserving the QoS settings in the IPN traffic between pods, but would like to preserve the original CoS values of the packets ingressing and egressing the fabric, see Class of Service (CoS) Preservation for Ingress and Egress Traffic instead.

Figure 1. Multi-Pod Topology

As illustrated in this figure, traffic between pods in a Multi-Pod topology passes through an IPN, which may contain devices that are not under Cisco APIC's management. When a network packet is sent from a spine or a leaf switch in POD1, the devices in the IPN may modify the 802.1p value in the packet. In this case, when the frame reaches a spine or a leaf switch in POD2, it would have an 802.1p value that was assigned by the IPN device, instead of the Cisco ACI QoS Level value assigned at the source in POD1.

In order to preserve the proper QoS Level of the packet and avoid high priority packets from being delayed or dropped, you can use a DSCP translation policy for traffic that goes between multiple PODs connected by an IPN. When a DSCP translation policy is enabled, Cisco APIC converts the QoS Level value (represented by the CoS value of the VXLAN packet) to a DSCP value according to the mapping rules you specify. When a packet sent from POD1 reaches POD2, the mapped DSCP value is translated back into the original CoS value for the appropriate QoS Level.

DSCP Translation Guidelines

  • Prior to Cisco APIC release 4.0(1), custom DSCP values could be assigned to User Levels 1 through 3.

  • Starting with Cisco APIC release 4.0(1), values can be selected for Levels 4 through 6 as well.

  • Starting with Cisco APIC release 4.0(1), if a multipod DSCP translation policy is enabled and the fabric hardware includes spine switch models that are earlier than the -EX switches, the CoS value for the traceroute policy must not overlap with the user traffic.

  • In addition to the values you configure in the DSCP translation policy, DSCP values 57-63 are used by the ACI Control Plane traffic across IPN.


Note

For traffic passing through the IPN, do not map any DSCP values to CS6.

The following table provides definitions for the DSCP and ToS settings used in DSCP policies and maps:

DSCP or ToS Level

Description

AF11

Assured Forwarding Class 1, low probability of dropping

AF12

Assured Forwarding Class 1, medium probability of dropping

AF13

Assured Forwarding Class 1, high probability of dropping

AF21

Assured Forwarding Class 2, low probability of dropping

AF22

Assured Forwarding Class 2, medium probability of dropping

AF23

Assured Forwarding Class 2, high probability of dropping

AF31

Assured Forwarding Class 3, low probability of dropping

AF32

Assured Forwarding Class 3, medium probability of dropping

AF33

Assured Forwarding Class 3, high probability of dropping

AF41

Assured Forwarding Class 4, low probability of dropping

AF42

Assured Forwarding Class 4, medium probability of dropping

AF43

Assured Forwarding Class 4, high probability of dropping

CS0

TOS Class Selector value 0 (the default)

CS1

TOS Class Selector value 1 (typically used for streaming traffic)

CS2

TOS Class Selector value 2 (typically used for OAM traffic such as SNMP, SSH, and Syslog)

CS3

TOS Class Selector value 3 (typically used for signalling traffic)

CS4

TOS Class Selector value 4 (typically used for Policy Plane traffic and to priority queue)

CS5

TOS Class Selector value 5 (typically used for broadcast video traffic)

CS6

TOS Class Selector value 6 (typically used for Network control traffic)

CS7

TOS Class Selector value 7

Expedited Forwarding

EF is dedicated to low-loss, low-latency traffic

Voice Admit

Similar to EF, but also admitted through CAC

Creating DSCP Translation Policy Using Cisco APIC GUI

This section describes how to create a DSCP translation policy to guarantee QoS Level settings across multiple PODs connected by an IPN.

Procedure

Step 1

Navigate to Tenants > infra.

Step 2

In the Navigation pane, expand Policies > Protocol > DSCP class-cos translation policy for L3 traffic.

Step 3

In the Properties panel, click Enabled to enable the DSCP policy.

Step 4

Check the No-Drop DSCP match control: enable nodrop dscp match checkbox to enable DSCP match for no-drop traffic for PFC classification on all spines including IPN interfaces.

Step 5

Map each traffic stream to one of the available levels.

Note

 

Each QoS Level must be mapped to a unique value.

Step 6

Click Submit to save the changes.

Creating DSCP Translation Policy Using NX-OS Style CLI

This section describes how to create a DSCP translation policy to guarantee QoS Level settings across multiple PODs connected by an IPN.

Procedure

Step 1

Enters configuration mode.

apic1# configure

Step 2

Enters tenant configuration mode for the infra tenant. 

apic1(config)# tenant infra

Step 3

Create the DSCP translation map.

apic1(config-tenant)# qos dscp-map default

Step 4

Configure the DSCP translation mappings.

Note

 

All mappings must be unique within a DSCP translation map and you must not map any QoS level to CS6. 

apic1(config-qos-cmap# set dscp-code control CS3
apic1(config-qos-cmap# set dscp-code span CS5
apic1(config-qos-cmap# set dscp-code level1 CS0
apic1(config-qos-cmap# set dscp-code level2 CS1
apic1(config-qos-cmap# set dscp-code level3 CS2
apic1(config-qos-cmap# set dscp-code level4 CS3
apic1(config-qos-cmap# set dscp-code level5 CS4
apic1(config-qos-cmap# set dscp-code level6 CS5
apic1(config-qos-cmap# set dscp-code policy CS4
apic1(config-qos-cmap# set dscp-code traceroute CS5

Step 5

Enable the DSCP translation.

apic1(config-qos-cmap)# no shutdown

Creating DSCP Translation Policy Using REST API

This section describes how to create a DSCP translation policy to guarantee QoS Level settings across multiple PODs connected by an IPN.

Procedure

Step 1

Enable and configure a DSCP translation policy.

POST https://<apic-ip>/api/node/mo/uni/tn-infra/dscptranspol-default.xml
<qosDscpTransPol dn="uni/tn-infra/dscptranspol-default" adminSt="enabled"
                 traceroute="AF43" span="AF42" policy="AF22" level3="AF13"
                 level2="AF12" level1="AF11" control="AF21" />

Step 2

Disable the DSCP translation policy.

POST https://<apic-ip>/api/node/mo/uni/tn-infra/dscptranspol-default.xml
<qosDscpTransPol dn="uni/tn-infra/dscptranspol-default" adminSt="disabled"
                 traceroute="AF43" span="AF42" policy="AF22" level3="AF13"
                 level2="AF12" level1="AF11" control="AF21"/>

Configuring QoS On IPN Devices

This section describes how to configure QoS on the IPN devices to map traffic to the different classes specified as part of the DSCP translation policy described in the previous section.

Before you begin

You must have configured Multipod.

Procedure

Step 1

On the APIC, match the QoS Class Policy-Level 1, QoS Class Policy-Level 2, and QoS Class Policy-Level 3 according to the policy determined in the IP network (IPN to IPN).

  1. On the menu bar, click Fabric > Access Policies.

  2. In the Policies pane, click Global Policies > QOS Class Policies > Level 1.

  3. In the QOS Class Policy - Level1 panel, update the Scheduling Algorithm and Bandwidth Allocated (in%) drop-down list.

  4. Click Submit.

  5. Repeat the steps for QoS Class Policy-Level 2 and QoS Class Policy-Level 3.

Step 2

On the APIC, create a DSCP policy to enable guaranteeing QoS priority settings in a multipod topology and configure DSCP mappings for various traffic streams in the fabric.

  1. On the menubar, click TENANTS > infra.

  2. In the Navigation pane, expand Protocol Policies > DSCP class-cos translation policy for L3 traffic.

  3. In the Properties panel, click Enabled to enable the DSCP policy.

  4. Map each traffic stream to one of the available levels. They must all be unique.

    The traffic in the IP network (from IPN to IPN) is treated as priority traffic.

    For traffic passing through the IPN, do not map any DSCP value to CS6.

    For example:

    • User Level 1 traffic is mapped to Expedited Forwarding, since it carries voice and real time traffic.

    • User Level 2 traffic is mapped to CS3, as it is often used for traffic marked for precedence 3 treatment.

    • User Level 3 traffic is mapped to CS0, as it is the default traffic.

    • User Level 4

    • User Level 5

    • User Level 6

    • Control Plane Traffic is mapped with CS7 and to priority queue.

    • Policy Plane Traffic is mapped with CS4 and to priority queue.

    • Span Traffic is mapped with CS1, as it is traditionally treated as background or scavenger class traffic.

    • Traceroute Traffic is mapped with CS5.

  5. Click Submit.

Step 3

On each IPN device, configure the following:

  1. Create class maps to match the markings configured on the APIC.

    class-map type qos match-all UserLevel1
  match dscp 46
class-map type qos match-all UserLevel2
  match dscp 24
class-map type qos match-all UserLevel3
  match dscp 0
class-map type qos match-all SpanTraffic
  match dscp 8
class-map type qos match-all iTraceroute
  match dscp 40
class-map type qos match-all CONTROL-TRAFFIC
  match dscp 32,56

  1. Create a policy map to label the ingress Control Plane and Policy Plane traffic with a QoS group. 

    policy-map type qos ACI-CLASSIFICATION
  class CONTROL-TRAFFIC
    set qos-group 7
  class UserLevel1
    set qos-group 6
  class UserLevel2
    set qos-group 3
  class UserLevel3
    set qos-group 0
  class SpanTraffic
    set qos-group 1
  class iTraceroute
    set qos-group 5

  2. Configure priority queue for the QoS group.

    policy-map type queuing IPN-8q-out-policy
  class type queuing c-out-8q-q7
    priority level 1
  class type queuing c-out-8q-q6
    priority level 2
  class type queuing c-out-8q-q5
    bandwidth remaining percent 0
  class type queuing c-out-8q-q4
    bandwidth remaining percent 0
  class type queuing c-out-8q-q3
    bandwidth remaining percent 40
  class type queuing c-out-8q-q2
    bandwidth remaining percent 0
  class type queuing c-out-8q-q1
    bandwidth remaining percent 1
  class type queuing c-out-8q-q-default
    bandwidth remaining percent 58

  3. Apply the policy map to system level QoS.

    system qos
  service-policy type queuing output IPN-8q-out-policy

  4. Associate the interfaces connected to the spine switch with the service policy.

    interface Ethernet1/49.4
  description POD2-Spine-401 e1/5
  encapsulation dot1q 4
  vrf member IPNACISJC
  service-policy type qos input ACI-CLASSIFICATION
  ip address 10.149.195.106/30
  ip ospf network point-to-point
  ip router ospf IPNACISJC area 0.0.0.0
  ip pim sparse-mode
  ip dhcp relay address 10.0.0.1 
  ip dhcp relay address 10.0.0.2 
  ip dhcp relay address 10.0.0.3 
  no shutdown

interface Ethernet1/50.4
  description POD2-Spine-402 e1/5
  encapsulation dot1q 4
  vrf member IPNACISJC
  service-policy type qos input ACI-CLASSIFICATION
  ip address 10.149.195.110/30
  ip ospf network point-to-point
  ip router ospf IPNACISJC area 0.0.0.0
  ip pim sparse-mode
  ip dhcp relay address 10.0.0.1 
  ip dhcp relay address 10.0.0.2 
  ip dhcp relay address 10.0.0.3 
  no shutdown

Step 4

(Optional) Verify the ingress interface on IPN.

You can verify the ingress interface settings as described in Verifying IPN Ingress Interface Settings.

Step 5

(Optional) Verify the egress interface on IPN.

You can verify the egress interface settings as described in Verifying IPN Egress Interface Settings.

L3Outs QoS

L3Out QoS can be configured using Contracts applied at the external EPG level. Starting with Release 4.0(1), L3Out QoS can also be configured directly on the L3Out interfaces.


Note

If you are running Cisco APIC Release 4.0(1) or later, we recommend using the custom QoS policies applied directly to the L3Out to configure QoS for L3Outs.

Packets are classified using the ingress DSCP or CoS value so it is possible to use custom QoS policies to classify the incoming traffic into Cisco ACI QoS queues. A custom QoS policy contains a table mapping the DSCP/CoS values to the user queue and to the new DSCP/CoS value (in case of marking). If there is no mapping for a specific DSCP/CoS value, the user queue is selected by the QoS priority setting of the ingress L3Out interface if configured.

L3Outs QoS Guidelines and Limitations

The following guidelines and limitations apply to configuring QoS for L3Outs:

  • A custom QoS policy is not supported for Layer 3 multicast traffic sourced from outside the Cisco Application Centric Infrastructure (ACI) fabric (received from L3Out).

  • When configuring the QoS policy by using contracts to be enforced on the border leaf switch where the L3Out is located, the VRF table must be in egress mode (Policy Control Enforcement Direction must be "Egress").

    A custom QoS setting can be configured directly on an L3Out and applied for the traffic coming from the border leaf switch, as such, the VRF table does not need to be in egress mode.

  • To enable the QoS policy to be enforced, the VRF Policy Control Enforcement Preference must be "Enforced."

  • When configuring the contract that controls communication between the L3Out and other EPGs, include the QoS class or target DSCP in the contract or subject.


    Note

    Only configure a QoS class or target DSCP in the contract, not in the external EPG (l3extInstP).

  • When creating a contract subject, you must choose a QoS priority level. You cannot choose Unspecified.


    Note

    The exception is with custom QoS policies, as a custom QoS policy will set the DSCP/CoS value even if the QoS class is set to Unspecified. When the QoS level is unspecified, the level is treated as 3 by default.

  • On generation 2 switches, QoS supports levels 4, 5, and 6 configured under Global policies, EPG, L3Out, custom QoS, and Contracts. The following limitations apply:

    • The number of classes that can be configured with the strict priority is increased to 5.

    • The 3 new classes are supported only with generation 2 switches.

    • If traffic flows between generation 1 switches and generation 2 switches, the traffic will use QoS level 3.

    • For communicating with FEX for new classes, the traffic carries a Layer 2 CoS value of 0.

    Generation 1 switches can be identified by the lack of "EX," "FX, "FX2," "GX," or later suffix at the end of the name. For example, N9K-9312TX. Generation 2 and later switches can be identified by the "EX," "FX, "FX2," "GX," or later suffix at the end of the name. For example N9K-93108TC-EX or N9K-9348GC-FXP.

  • You can configure QoS class or create a custom QoS policy to apply on an L3Out interface.

Configuring QoS Directly on L3Out Using GUI

This section describes how to configure QoS directly on an L3Out. This is the preferred way of configuring L3Out QoS starting with Cisco APIC Release 4.0(1).

Procedure

Step 1

From the main menu bar, select Tenants > <tenant-name> .

Step 2

In the left-hand navigation pane, expand Tenant <tenant-name> > Networking > L3Outs > <routed-network-name> > Logical Node Profiles > <node-profile-name> > Logical Interface Profiles > <interface-profile-name>.

You may need to create new network, node profile, and interface profile if none exists.

Step 3

In the main window pane, configure custom QoS for your L3Out.

You can choose to configure a standard QoS level priority using the QoS Priority drop-down list. Alternatively, you can set an existing or create a new custom QoS policy from the Custom QoS Policy dropdown.

Configuring QoS Directly on L3Out Using CLI

This section describes how to configure QoS directly on an L3Out. This is the preferred way of configuring L3Out QoS starting with Cisco APIC Release 4.0(1).

You can configure QoS for L3Out on one of the following objects:

  • Switch Virtual Interface (SVI)

  • Sub Interface

  • Routed Outside

Procedure

Step 1

Configure QoS priorities for a L3Out SVI.

Example:

interface vlan 19
    vrf member tenant DT vrf dt-vrf
    ip address 107.2.1.252/24
    description  'SVI19'
    service-policy type qos VrfQos006   // for custom QoS attachment
    set qos-class level6                // for set QoS priority
    exit

Step 2

Configure QoS priorities for a sub-interface.

Example:

interface ethernet 1/48.10
    vrf member tenant DT vrf inter-tentant-ctx2 l3out L4_E48_inter_tennant
    ip address 210.2.0.254/16
    service-policy type qos vrfQos002
    set qos-class level5

Step 3

Configure QoS priorities for a routed outside.

Example:

interface ethernet 1/37
    no switchport
    vrf member tenant DT vrf dt-vrf l3out L2E37
    ip address 30.1.1.1/24
    service-policy type qos vrfQos002
    set qos-class level5
    exit

Configuring QoS Directly on L3Out Using REST API

This section describes how to configure QoS directly on an L3Out. This is the preferred way of configuring L3Out QoS starting with Cisco APIC Release 4.0(1).

You can configure QoS for L3Out on one of the following objects:

  • Switch Virtual Interface (SVI)

  • Sub Interface

  • Routed Outside

Procedure

Step 1

Configure QoS priorities for a L3Out SVI.

Example:

<l3extLIfP descr="" dn="uni/tn-DT/out-L3_4_2_24_SVI17/lnodep-L3_4_E2_24/lifp-L3_4_E2_24_SVI_19"
           name="L3_4_E2_24_SVI_19" prio="level6" tag="yellow-green">
    <l3extRsPathL3OutAtt addr="0.0.0.0" autostate="disabled" descr="SVI19" encap="vlan-19"
                         encapScope="local" ifInstT="ext-svi" ipv6Dad="enabled" llAddr="::"
                         mac="00:22:BD:F8:19:FF" mode="regular" mtu="inherit"
                         tDn="topology/pod-1/protpaths-103-104/pathep-[V_L3_l4_2-24]"
                         targetDscp="unspecified">
        <l3extMember addr="107.2.1.253/24" ipv6Dad="enabled" llAddr="::" side="B"/>
        <l3extMember addr="107.2.1.252/24" ipv6Dad="enabled" llAddr="::" side="A"/>
    </l3extRsPathL3OutAtt>
    <l3extRsLIfPCustQosPol tnQosCustomPolName="VrfQos006"/>
</l3extLIfP>

Step 2

Configure QoS priorities for a sub-interface.

Example:

<l3extLIfP dn="uni/tn-DT/out-L4E48_inter_tenant/lnodep-L4E48_inter_tenant/lifp-L4E48"
           name="L4E48" prio="level4" tag="yellow-green">
    <l3extRsPathL3OutAtt addr="210.1.0.254/16" autostate="disabled" encap="vlan-20"
                         encapScope="local" ifInstT="sub-interface" ipv6Dad="enabled" llAddr="::"
                         mac="00:22:BD:F8:19:FF" mode="regular" mtu="inherit"
                         tDn="topology/pod-1/paths-104/pathep-[eth1/48]" targetDscp="unspecified"/>
    <l3extRsNdIfPol annotation="" tnNdIfPolName=""/>
    <l3extRsLIfPCustQosPol annotation="" tnQosCustomPolName=" vrfQos002"/>
</l3extLIfP>

Step 3

Configure QoS priorities for a routed outside.

Example:

<l3extLIfP dn="uni/tn-DT/out-L2E37/lnodep-L2E37/lifp-L2E37OUT"
           name="L2E37OUT" prio="level5" tag="yellow-green">
    <l3extRsPathL3OutAtt addr="30.1.1.1/24" autostate="disabled" encap="unknown"
                         encapScope="local" ifInstT="l3-port" ipv6Dad="enabled"
                         llAddr="::" mac="00:22:BD:F8:19:FF" mode="regular"
                         mtu="inherit" targetDscp="unspecified"
                         tDn="topology/pod-1/paths-102/pathep-[eth1/37]"/>
    <l3extRsNdIfPol annotation="" tnNdIfPolName=""/>
    <l3extRsLIfPCustQosPol tnQosCustomPolName="vrfQos002"/>
</l3extLIfP>

Configuring QoS Contracts for L3Outs Using GUI

This section describes how to configure QoS for L3Outs using Contracts.


Note

Starting with Release 4.0(1), we recommend using custom QoS policies for L3Out QoS as described in Configuring QoS Directly on L3Out Using GUI instead.

Configuring QoS classification using a contract as described in this section will take priority over any QoS policies configured directly on the L3Out.

Procedure

Step 1

Configure the VRF instance for the tenant consuming the L3Out to support QoS to be enforced on the border leaf switch that is used by the L3Out.

  1. From the main menu bar, choose Tenants > <tenant-name> .

  2. In the Navigation pane, expand Networking, right-click VRFs, and choose Create VRF.

  3. Enter the name of the VRF.

  4. In the Policy Control Enforcement Preference field, choose Enforced.

  5. In the Policy Control Enforcement Direction choose Egress

    VRF enforcement must be set to Egress when the QoS classification is done in the contract.

  6. Complete the VRF configuration according to the requirements for the L3Out.

Step 2

When configuring filters for contracts to enable communication between the EPGs consuming the L3Out, include a QoS class or target DSCP to enforce the QoS priority in traffic ingressing through the L3Out.

  1. On the Navigation pane, under the tenant that that will consume the L3Out, expand Contracts, right-click Filters and choose Create Filter.

  2. In the Name field, enter a filter name.

  3. In the Entries field, click + to add a filter entry.

  4. Add the Entry details, click Update and Submit.

  5. Expand the previously created filter and click on a filter entry.

  6. Set the Match DSCP field to the desired DSCP level for the entry, for example, EF.

Step 3

Add a contract.

  1. Under Contracts, right-click Standard and choose Create Contract.

  2. Enter the name of the contract.

  3. In the QoS Class field, choose the QoS priority for the traffic governed by this contract. Alternatively, you can choose a Target DSCP value.

    Configuring QoS classification using a contract as described in this section will take priority over any QoS policies configured directly on the L3Out.

  4. Click the + icon on Subjects to add a subject to the contract.

  5. Enter a name for the subject.

  6. In the QoS Priority field, choose the desired priority level. You cannot choose Unspecified.

  7. Under Filter Chain, click the + icon on Filters and choose the filter you previously created, from the drop down list.

  8. Click Update.

  9. On the Create Contract Subject dialog box, click OK.

Configuring QoS Contract for L3Out Using CLI

This section describes how to configure QoS for L3Outs using Contracts.


Note

Starting with Release 4.0(1), we recommend using custom QoS policies for L3Out QoS as described in Configuring QoS Directly on L3Out Using CLI instead.

Procedure

Step 1

Configure the VRF for egress mode and enable policy enforcement to support QoS priority enforcement on the L3Out.

apic1# configure
apic1(config)# tenant t1
apic1(config-tenant)# vrf context v1
apic1(config-tenant-vrf)# contract enforce egress
apic1(config-tenant-vrf)# exit
apic1(congig-tenant)# exit
apic1(config)#

Step 2

Configure QoS.

When creating filters (access-list), include the match dscp command with target DSCP level.
When configuring contracts, include the QoS class for traffic ingressing on the L3Out. Alternatively, you can define a target DSCP value. QoS policies are supported on either the contract or the subject

VRF enforcement must be ingress, for QoS or custom QoS on L3out interface, VRF enforcement need be egress, only when the QOS classification is going to be done in the contract for traffic between EPG and L3out or L3out to L3out.

Note

 

If QoS classification is set in the contract and VRF enforcement is egress, then contract QoS classification would override the L3Out interface QoS or Custom QoS classification. 

apic1(config)# tenant t1
apic1(config-tenant)# access-list http-filter
apic1(config-tenant-acl)# match ip
apic1(config-tenant-acl)# match tcp dest 80
apic1(config-tenant-acl)# match dscp EF
apic1(config-tenant-acl)# exit
apic1(config-tenant)# contract httpCtrct
apic1(config-tenant-contract)# scope vrf
apic1(config-tenant-contract)# qos-class level1
apic1(config-tenant-contract)# subject http-subject
apic1(config-tenant-contract-subj)# access-group http-filter both 
apic1(config-tenant-contract-subj)# exit
apic1(config-tenant-contract)# exit
apic1(config-tenant)# exit
apic1(config)#

Configuring QoS Contract for L3Out Using REST API

This section describes how to configure QoS for L3Outs using Contracts.


Note

Starting with Release 4.0(1), we recommend using custom QoS policies for L3Out QoS as described in Configuring QoS Directly on L3Out Using REST API instead.

Procedure

Step 1

When configuring the tenant, VRF, and bridge domain, configure the VRF for egress mode (pcEnfDir="egress") with policy enforcement enabled (pcEnfPref="enforced"). Send a post with XML similar to the following example:

Example:

<fvTenant  name="t1">
    <fvCtx name="v1" pcEnfPref="enforced" pcEnfDir="egress"/>
    <fvBD name="bd1">
        <fvRsCtx tnFvCtxName="v1"/>
        <fvSubnet ip="44.44.44.1/24" scope="public"/>
        <fvRsBDToOut tnL3extOutName="l3out1"/>
    </fvBD>"/>
</fvTenant>

Step 2

When creating the filters and contracts to enable the EPGs participating in the L3Out to communicate, configure the QoS priority.

The contract in this example includes the QoS priority, level1, for traffic ingressing on the L3Out. Alternatively, it could define a target DSCP value. QoS policies are supported on either the contract or the subject.

The filter also has the matchDscp="EF" criteria, so that traffic with this specific TAG received by the L3out processes through the queue specified in the contract subject.

Note

 

VRF enforcement should be ingress, for QOS or custom QOS on L3out interface, VRF enforcement need be egress, only when the QOS classification is going to be done in the contract for traffic between EPG and L3out or L3out to L3out.

Note

 

If QOS classification is set in the contract and VRF enforcement is egress, then contract QOS classification would override the L3out interface QOS or Custom QOS classification, So either we need to configure this one or the new one.

Example:

<vzFilter name="http-filter">
    <vzEntry  name="http-e" etherT="ip" prot="tcp" matchDscp="EF"/>
</vzFilter>
<vzBrCP name="httpCtrct" prio="level1" scope="context">
    <vzSubj name="subj1">
        <vzRsSubjFiltAtt tnVzFilterName="http-filter"/>
    </vzSubj>
</vzBrCP>

SR-MPLS QoS

Starting with Release 5.0(1), Cisco ACI fabric supports QoS classification and marking for MPLS Segment Routing (SR-MPLS) traffic ingressing and egressing the fabric.

You can use custom QoS policies to define how traffic coming from an MPLS network is prioritized within ACI fabric. You can also use these policies to re-mark the traffic when it leaves the fabric via an MPLS L3Out.

When configuring a custom QoS policy, you define the following two rules that are applied on the border leaf switch:

  • Ingress rules: These rules are applied for traffic that is ingressing the ACI fabric from an MPLS network and are used to map incoming packet's experimental bits (EXP) values to ACI QoS levels, as well as to set differentiated services code point (DSCP) values in the VXLAN header for the packet while it's inside the ACI fabric.

    The values are derived at the border leaf using a custom QoS translation policy. The original DSCP values for traffic coming from SR-MPLS without any remarking. If a custom policy is not defined or not matched, default QoS Level (Level3) is assigned.

  • Egress rules: These rules are applied for the traffic that is leaving the ACI fabric via an MPLS L3Out and are used to map the packet's IPv4 DSCP value to the MPLS packet's EXP value as well as the internal ethernet frame's CoS value.

    Classification is done at the non-border leaf switch based on existing policies used for EPG and L3Out traffic. If a custom policy is not defined or not matched, the default EXP value of 0 is marked on all labels. EXP values are marked in both, default and custom policy scenarios, and are done on all MPLS labels in the packet.

    Custom MPLS egress policy can override existing EPG, L3out, and Contract QoS policies.

The following two figures summarize when the ingress and egress rules are applied as well as how the internal ACI traffic may remark the packets' QoS fields while inside the fabric.

Figure 2. Ingress QoS
Figure 3. Ingress QoS

SR-MPLS QoS Guidelines and Limitations

The following guidelines apply when configuring QoS policies for SR-MPLS traffic:

  • Matching Exp value in Contract-level custom QoS policy is not supported. Any custom QoS policy configured at the Contract level will override the global MPLS QoS policy.

  • Dynamic Packet Prioritization (DPP) is supported for both, ingress and egress traffic.

  • An MPLS interface can act as a SPAN source, but it cannot be configured as a monitor port.

Creating SR-MPLS Custom QoS Policy Using the GUI

SR-MPLS Custom QoS policy defines the priority of the packets coming from an SR-MPLS network while they are inside the ACI fabric based on the incoming MPLS EXP values defined in the MPLS QoS ingress policy. It also marks the CoS and MPLS EXP values of the packets leaving the ACI fabric through an MPLS interface based on IPv4 DSCP values defined in MPLS QoS egress policy.

If no custom ingress policy is defined, the default QoS Level (Level3) is assigned to packets inside the fabric. If no custom egress policy is defined, the default EXP value of 0 will be marked on packets leaving the fabric.

Procedure

Step 1

From the top menu bar, navigate to Tenants > infra.

Step 2

In the left pane, select infra > Policies > Protocol > MPLS Custom QoS.

Step 3

Right click the MPLS Custom QoS folder and choose Create MPLS Custom QoS Policy.

Step 4

In the Create MPLS Custom QoS Policy window that opens, provide the name and description of the policy you're creating.

Step 5

In the MPLS Ingress Rule area, click + to add an ingress QoS translation rule.

Any traffic coming into the border leaf (BL) connected to the MPLS network will be checked for the MPLS EXP value and if a match is found, the traffic is classified into an ACI QoS Level and marked with appropriate CoS and DSCP values.

  1. In the Priority field, select the priority for the ingress rule.

    This is the QoS Level you want to assign for the traffic within ACI fabric, which ACI uses to prioritize the traffic within the fabric.. The options range from Level1 to Level6. The default value is Level3. If you do not make a selection in this field, the traffic will automatically be assigned a Level3 priority.

  2. In the EXP Range From and EXP Range To fields, specify the EXP range of the ingressing MPLS packet you want to match.

  3. In the Target DSCP field, select the DSCP value to assign to the packet when it's inside the ACI fabric.

    The DSCP value specified is set in the original traffic received from the external network, so it will be re-exposed only when the traffic is VXLAN decapsulated on the destination ACI leaf node.

    The default is Unspecified, which means that the original DSCP value of the packet will be retained.

  4. In the Target CoS field, select the CoS value to assign to the packet when it's inside the ACI fabric.

    The CoS value specified is set in the original traffic received from the external network, so it will be re-exposed only when the traffic is VXLAN decapsulated on the destination ACI leaf node.

    The default is Unspecified, which means that the original CoS value of the packet will be retained, but only if the CoS preservation option is enabled in the fabric.

  5. Click Update to save the ingress rule.

  6. Repeat this step for any additional ingress QoS policy rules.

Step 6

In the MPLS Egress Rule area, click + to add an egress QoS translation rule.

When the traffic is leaving the fabric out of the border leaf's MPLS interface, it will be matched based on the DSCP value of the packet and if a match is found, the MPLS EXP and CoS values will be set based on the policy.

  1. Using the DSCP Range From and DSCP Range To dropdowns, specify the DSCP range of the ACI fabric packet you want to match for assigning the egressing MPLS packet's priority.

  2. From the Target EXP dropdown, select the EXP value you want to assign to the egressing MPLS packet.

  3. From the Target CoS dropdown, select the CoS value you want to assign to the egressing MPLS packet.

  4. Click Update to save the ingress rule.

  5. Repeat this step for any additional egress QoS policy rules.

Step 7

Click OK to complete the creation of the MPLS custom QoS Policy.

Creating SR-MPLS Custom QoS Policy Using CLI

SR-MPLS Custom QoS policy defines the priority of the packets coming from an SR-MPLS network while they are inside the ACI fabric based on the incoming MPLS EXP values defined in the MPLS QoS ingress policy. It also marks the CoS and MPLS EXP values of the packets leaving the ACI fabric through an MPLS interface based on IPv4 DSCP values defined in MPLS QoS egress policy.

If no custom ingress policy is defined, the default QoS Level (Level3) is assigned to packets inside the fabric. If no custom egress policy is defined, the default EXP value of 0 will be marked on packets leaving the fabric.

Procedure

Step 1

Create SR-MPLS QoS policy.

  1. Create the policy.

    In the following command, replace mplsqos1 with the name of the SR-MPLS QoS policy you want to create. 

    apic1(config-tenant)# policy-map type mplsqos mplsqos1

  2. Create ingress policy rule.

    In the following command:

    • Replace <exp-range-start> and <exp-range-end> with the DSCP range you want the policy to match, for example 10 20.

    • Replace <cos-value> with the CoS value you want to set on the packet when it's matched, for example 3.

    • Replace <dscp-value> with the DSCP value you want to set on the packet when it's matched, for example 15

    • Replace <aci-qos-level> with the ACI QoS Level for the packet while it's inside the ACI fabric, for example level2. 

    apic1(config-tenant-pmap-mplsqos)# match exp <exp-range-start> <exp-range-end> set-cos <cos-value> set-dscp <dscp-value> set-class <aci-qos-level>

  3. Create egress policy rule.

    In the following command:

    • Replace <dscp-range-start> and <dscp-range-end> with the DSCP range you want the policy to match, for example 10 20.

    • Replace <cos-value> with the CoS value you want to set on the packet when it's leaving the fabric, for example 2.

    • Replace <exp-value> with the MPLS EXP value you want to set on the packet when it's leaving the fabric, for example 3. 

    apic1(config-tenant-pmap-mplsqos)# match dscp <dscp-range-start> <dscp-range-end> set-cos <cos-value> set-exp <exp-value>

  4. Exit policy configuration.

    apic1(config-tenant-pmap-mplsqos)# exit

Step 2

Apply SR-MPLS QoS policy.

In the following commands:

  • Replace 101 with the border leaf switch.

  • Replace overlay-1 with the VRF used by the SR-MPLS L3Out.

  • Replace mplsqos1 with the name of the SR-MPLS QoS policy you created in the previous step. 

apic1# configure
apic1(config)# leaf 101
apic1(config-leaf)# vrf context tenant infra vrf overlay-1
apic1(config-leaf-vrf)# exit
apic1(config-leaf)# service-policy type mplsqos mplsqos1
apic1(config-leaf)# exit

Creating SR-MPLS Custom QoS Policy Using REST API

SR-MPLS Custom QoS policy defines the priority of the packets coming from an SR-MPLS network while they are inside the ACI fabric based on the incoming MPLS EXP values defined in the MPLS QoS ingress policy. It also marks the CoS and MPLS EXP values of the packets leaving the ACI fabric through an MPLS interface based on IPv4 DSCP values defined in MPLS QoS egress policy.

If no custom ingress policy is defined, the default QoS Level (Level3) is assigned to packets inside the fabric. If no custom egress policy is defined, the default EXP value of 0 will be marked on packets leaving the fabric.

Procedure

Step 1

Create SR-MPLS QoS policy.

In the following POST:

  • Replace customqos1 with the name of the SR-MPLS QoS policy you want to create.

  • For the qosMplsIngressRule:

    • Replace from="2" to="3" with the EXP range you want the policy to match.

    • Replace prio="level5" with the ACI QoS Level for the packet while it's inside the ACI fabric.

    • Replace target="CS5" with the DSCP value you want to set on the packet when it's matched.

    • Replace targetCos="4" with the CoS value you want to set on the packet when it's matched.

  • For the qosMplsEgressRule:

    • Replace from="CS2" to="CS4" with the DSCP range you want the policy to match.

    • Replace targetExp="5" with the EXP value you want to set on the packet when it's leaving the fabric.

    • Replace targetCos="3" with the CoS value you want to set on the packet when it's leaving the fabric. 

<polUni>
  <fvTenant name="infra">
    <qosMplsCustomPol descr="" dn="uni/tn-infra/qosmplscustom-customqos1" name="customqos1"  status="" >
        <qosMplsIngressRule from="2" to="3" prio="level5" target="CS5" targetCos="4" status="" />
        <qosMplsEgressRule from="CS2" to="CS4" targetExp="5" targetCos="3" status=""/>
    </qosMplsCustomPol>
  </fvTenant>
</polUni>

Step 2

Applying SR-MPLS QoS policy.

In the following POST, replace customqos1 with the name of the SR-MPLS QoS policy you created in the previous step. 

<polUni>
    <fvTenant name="infra">
        <l3extOut name="mplsOut" status="" descr="bl">
            <l3extLNodeP name="mplsLNP" status="">
                <l3extRsLNodePMplsCustQosPol  tDn="uni/tn-infra/qosmplscustom-customqos1"/>
            </l3extLNodeP>
        </l3extOut>
    </fvTenant>
</polUni>

RoCEv2 and the Required APIC QoS Settings

Remote Direct Memory Access (RDMA) over Converged Ethernet (RoCE) technology allows data to be transferred between servers or from storage to server without having to pass through the CPU and main memory path of TCP/IP. The network adapters transfers data directly to and from the application memory bypassing the operating system and the CPU. This zero copy and CPU offloading approach ensures greater CPU availability for other tasks while providing low latency and reduced jitter. You can use a single fabric for both storage and compute. RoCEv2 provides more functionality by allowing RDMA to be used with both Layer 2 and Layer 3 (UDP/IP) packets, enabling Layer 3 routing over multiple subnets.

Starting with Cisco Application Policy Infrastructure Controller (APIC) release 4.0(1), you can enable RoCEv2 functionality in your fabric by configuring specific QoS options for Layer 3 traffic in Cisco APIC, such as Weighted Random Early Detection (WRED) congestion algorithm and Explicit Congestion Notification (ECN).

The following sections describe how to configure the required QoS options using three different methods: the Cisco APIC GUI, the NX-OS style CLI, and the REST API. Regardless of which method you choose, you must configure the following:

  • Weighted Random Early Detection (WRED) congestion algorithm, which manages congestion on spine switches using the following configuration options:

    • WRED Min Threshold: If the average queue size is below the minimum threshold value, the arriving packets are queued immediately.

    • WRED Max Threshold: If the average queue size is greater than the maximum threshold value, the arriving packets are dropped.

    • WRED Probability: If the average queue size is between the Min and Max threshold, the Probability value determines whether the packet is dropped or queued.

    • WRED Weight: Weight has a range of 0 to 7 and is used to calculate average queue length. Lower weight prioritizes current queue length, while higher weight prioritizes older queue lengths.

  • Explicit Congestion Notification (ECN), which is used for congestion notification. If there is congestion, ECN gets the transmitting device to the reduce transmission rate until the congestion clears, allowing traffic to continue without pause. ECN along with WRED enables end-to-end congestion notification between two endpoints on the network.

  • Priority flow control (PFC), which is used to achieve Layer 2 flow control. PFC provides the capability to pause traffic if there is congestion.

Starting with release 5.2(5), you can use RoCEv2 together with Cisco ACI Multi-Pod, but the following guidelines and restrictions apply:

  • Remote leaf switches do not support RoCEv2 with Cisco ACI Multi-Pod.

  • Enable PFC end-to-end through an Inter-Pod Network (IPN).

  • Make sure that the class of service (CoS) is preserved over the IPNs between the spine switches of the different pods.

  • RoCEv2 supports only Cisco ACI QoS Level 1 or 2 inside the fabric.

  • Configure the IPN with regular PFC and WRED or ECN.

  • Enable Cisco ACI Multi-Pod QoS.

ROCEv2 Hardware Support

The following Cisco hardware is supported for ROCEv2 in this release:

  • Cisco Nexus 9300-EX platform switches

  • Cisco Nexus 9300-FX platform switches

  • Cisco Nexus 9300-FX2 platform switches

  • Cisco Nexus 9300-FX3 platform switches

  • Cisco Nexus 9300-GX platform switches

  • N9K-X9700-EX line cards

  • N9K-C9504-FM-E fabric modules

Configuring Priority Flow Control (PFC) On Interfaces

Before you can configure the appropriate QoS settings for ROCEv2, you must enable PFC on each interface that is connected to ROCE devices. PFC setting can be set to one of three values, on, off, and auto. If you set it to auto, the DCBX protocol negotiates the PFC state on the interface.

You can configure PFC on one or more interfaces using any of the following methods:

Configuring PFC On Interfaces Using GUI

You can use the Cisco APIC GUI to configure PFC state on the interfaces connecting to ROCEv2 devices.

Procedure

Step 1

Log in to Cisco APIC.

Step 2

From the top navigation bar, choose Fabric > Inventory.

Step 3

In the left-hand sidebar, navigate to <pod> > <leaf-switch>.

Step 4

In the main pane, select the Interface tab.

Step 5

In the main pane, from the Mode dropdown menu, select Configuration.

Step 6

Choose an L2 port you want to configure.

Step 7

In the bottom pane, select the FCoE/FC tab.

Step 8

Set the PFC State of the port to On.

Configuring PFC On Interfaces Using CLI

You can use the NX-OS style CLI to configure PFC state on the interfaces connecting to ROCEv2 devices.

Procedure

Step 1

Enter APIC configuration mode.

apic1# config

Step 2

Enter switch configuration.

apic1(config)# leaf 101

Step 3

Enable PFC for specific interfaces.

apic1(config-leaf)# interface ethernet 1/7-9
apic1(config-leaf-if)# priority-flow-control mode on

Configuring PFC On Interfaces Using REST API

You can use REST API to configure PFC state on the interfaces connecting to ROCEv2 devices.

Procedure

Step 1

You can configure PFC state on a group of interfaces using a policy group.

<polUni>
  <infraInfra>
    <qosPfcIfPol name="testPfcPol1"  adminSt="on"/>
    <infraFuncP>
      <infraAccPortGrp name="groupName">
        <infraRsQosPfcIfPol tnQosPfcIfPolName="testPfcPol1"/>
      </infraAccPortGrp>
    </infraFuncP>
  </infraInfra>
</polUni>

Step 2

Alternatively, you can configure PFC state on individual interfaces.

<polUni>
  <infraInfra>
    <qosPfcIfPol name="testPfcPol"  adminSt="auto"/>
    <infraFuncP>
      <infraAccPortGrp name="testPortG">
        <infraRsQosPfcIfPol tnQosPfcIfPolName="testPfcPol"/>
      </infraAccPortGrp>
    </infraFuncP>
    <infraHPathS name="port20">
      <infraRsHPathAtt tDn="topology/pod-1/paths-102/pathep-[eth1/20]"/>
        <infraRsPathToAccBaseGrp tDn="uni/infra/funcprof/accportgrp-testPortG">  
      </infraRsPathToAccBaseGrp>
    </infraHPathS>
  </infraInfra>
</polUni>

Configuring QoS for ROCEv2

After you have enabled PFC on each interfaces that is connected to ROCE devices, you can configure the appropriate QoS settings for ROCEv2.

You can configure QoS for ROCE using any of the following methods:

Configuring QoS for ROCEv2 Using the GUI

You can use the Cisco APIC GUI to configure the required QoS options to enable support for RoCEv2 in your fabric.

Procedure

Step 1

Log in to Cisco APIC.

Step 2

Navigate to Fabric > Access Policies > Policies > Global > QOS Class

Step 3

Select the QOS Class Level for which you want to configure ROCEv2

Step 4

For the Congestion Algorithm option, select Weighted random early detection.

Step 5

For the Congestion Notification option, select Enabled.

Enabling Congestion Notification causes the packets that would be dropped to be ECN-marked instead.

Step 6

For the Min Threshold (percentage) option, set the minimum queue threshold as a percentage of the maximum queue length.

If the average queue size is below the minimum threshold value, the arriving packets are queued immediately.

Step 7

For the Max Threshold (percentage) option, set the maximum queue threshold as a percentage of the maximum queue length.

If the average queue size is greater than the maximum threshold value, the arriving packets are dropped or marked if ECN is enabled.

Step 8

For the Probability (percentage) option, set the probability value.

The probability determines whether the packet is dropped or queued when the average queue size is between the minimum and the maximum threshold values.

Step 9

For the Weight option, set the weight value.

Weight has a range of 0 to 7 and is used to calculate average queue length. Lower weight prioritizes current queue length, while higher weight prioritizes older queue lengths.

Step 10

Check the PFC Admin State checkbox and specify a value for the No-Drop-CoS option to be used by PFC.

Step 11

For the Scope option, select Fabric-wide PFC.

Step 12

Optionally, you can choose to enable the Forward Non-ECN Traffic option, so that non-ECN traffic is not dropped even when the queue is congested. Congestion Notification must be enabled for this option to be configurable.

Configuring QoS for RoCEv2 Using CLI

You can use the NX-OS style CLI to configure the required QoS options to enable support for RoCEv2 in your fabric.

Procedure

Step 1

Enter configuration mode.

apic1# config

Step 2

Choose the QoS Level you want to configure.

In the following command, replace level2 with the QoS Level you want to configure: 

apic1(config)# qos parameters level2

Step 3

Configure the congestion algorithm and its parameters.

apic1(config-qos)# algo wred
apic1(config-qos-algo)# ecn enabled
apic1(config-qos-algo)# maxthreshold 60
apic1(config-qos-algo)# minthreshold 40
apic1(config-qos-algo)# probability 0
apic1(config-qos-algo)# weight 1
apic1(config-qos-algo)# exit

Step 4

(Optional) Configure forwarding of the non-ECN traffic.

You can choose to enable forwarding of all non-ECN traffic, even when the queue is congested.

apic1(config-qos-algo)# fwdnonecn enabled

Step 5

Exit congestion algorithm configuration.

apic1(config-qos-algo)# exit

Step 6

Configure the CoS value for the QoS Level you chose.

If you do not provide the fabric parameter, the default value is set to TOR. 

apic1(config-qos)# pause no-drop cos 4 fabric

Configuring QoS for RoCEv2 Using REST API

You can use REST API to configure the required QoS options to enable support for RoCEv2 in your fabric.

Procedure

Step 1

Configure QoS for RoCEv2.

In the following example, replace level2 with the QoS class you want to configure and the WRED parameters with values appropriate for your environment.

POST URL: https://<apic-ip>/api/node/mo/uni.xml

<qosClass admin="enabled" dn="uni/infra/qosinst-default/class-level2" prio="level2">
    <qosCong algo="wred" wredMaxThreshold="60" wredMinThreshold="40" wredProbability="0"
             ecn="enabled"/>
    <qosPfcPol name="default" noDropCos="cos0" adminSt="yes" enableScope="fabric"/>
</qosClass>

Step 2

(Optional) Configure forwarding of the non-ECN traffic.

You can choose to enable forwarding of all non-ECN traffic, even when the queue is congested.

<qosInstPol dn="uni/infra/qosinst-default" FabricFlushInterval=450 FabricFlushSt="yes">
</qosInstPol>

ECN Buffer Tuning in ACI Fabric

You can use the ECN Buffer Tuning feature to adjust congestion management settings on ACI fabric switches. This feature lets you configure threshold multipliers to optimize how buffers are used based on your network's traffic patterns.

In previous ACI releases, Quality of Service (QoS) settings such as No-Drop Priority Group (PG) buffers, Weighted Random Early Detection (WRED) queue thresholds, and Port Pause thresholds used fixed, non-configurable values:

  • ECN queue threshold: 1K cells

  • Port PFC pause threshold: 1K cells

  • No-Drop PG buffer: 23K cells (uniform across all platforms, regardless of available buffer)

Starting with ACI release 6.2(2), you can configure these threshold values to better align with your network's actual traffic requirements. For details on the available settings, see Adjustable Threshold Settings.

Adjustable Threshold Settings

ECN Buffer Tuning provides four settings you can adjust:

  • No-Drop PG Buffer Threshold: Controls how many cells are set aside for the No-Drop Priority Group. You can set this from 50% to 75% of your platform's total buffer (instead of the fixed 23K cells). Higher values allow more temporary traffic buildup before triggering congestion signals.

  • Port Pause Threshold Multiplier: Controls when PFC pause frames are sent to upstream devices. You can set this from 1K to 4K cells (instead of the fixed 1K cells). This determines how much buffer can be used before asking upstream devices to pause.

  • WRED Queue Base Multiplier: Controls the baseline setting for ECN marking thresholds. You can set this from 0.5× to 2× the baseline value (instead of the fixed 1K cells). Higher values delay ECN marking until queues are more full.

  • WRED Queue Threshold Multiplier: Lets you set different ECN marking thresholds for each QoS traffic class. This lets you tune thresholds for different traffic types based on their requirements.

ECN Buffer Tuning addresses this limitation by giving you the option to configure ECN and PFC thresholds instead of using the low fixed defaults. This lets you adjust thresholds to match your actual traffic patterns so congestion handling is not triggered too early.

Policy Scope and Hierarchy

Policies follow a hierarchical model where node-level configurations override global settings:

  • Fabric Level (Global): Apply the same settings to all switches in your fabric. Use this when all your switches are the same model or when you want consistent settings across the entire fabric.

  • Switch level (node-level): Apply different settings to specific switches. Use this when you have different switch models or when different switches handle different types of traffic.

  • Traffic class level: Set different ECN thresholds for each traffic class (QoS class). Use this to be strict with non-critical traffic while giving more buffer room to important traffic.

Policy Application Rules

  • If a node-level policy is defined for a switch, that switch uses the node-level policy.

  • If no node-level policy exists for a switch, that switch uses the global policy.

  • If the global policy is disabled, all switches revert to factory-default thresholds (unless a node-level policy is enabled).

  • Factory-default thresholds are used only when both global and applicable node-level policies are disabled.

Supported Platforms and Parameter Availability

ECN Buffer Tuning support varies by switch platform. Understanding which parameters are available on your platforms is critical for planning your configuration.

Supported Platforms

ECN Buffer Tuning is supported on FX2 and later releases.

Platform Support Matrix

Not every switch model supports every setting. This table shows what you can configure on each platform:

Table 7. ECN Buffer Tuning Parameter Support by Switch Platform

Switch Model

No-Drop PG Buffer

Port Pause Threshold

WRED Queue Base

Per-Class WRED

GX Platform

✓

✓

✓

✓

FX2 Platform

✓

✓

✓

✓

FX3 Platform

✓

✓

✓

✓

GX2 Platform

✓

✗

✓

✓

H1 Platform (except H2R)

✓

✗

✓

✓

H2R Platform

✓

✗

✗

✗

Platform-Specific Limitations

GX2 and H1 (non-H2R): Dynamic AQM-Based PFC

GX2 and H1 platforms use Active Queue Management (AQM) mode for Priority Flow Control. In this architecture, the ASIC dynamically calculates PFC thresholds based on real-time congestion levels, without requiring configuration. For this reason:

  • Port Pause Threshold Multiplier is not available on these platforms

  • PFC behavior is automatically optimized by the ASIC

  • ECN Buffer Tuning still controls No-Drop PG Buffer, WRED Queue Base, and class-level WRED multipliers

H2R Platform: High Bandwidth Memory (HBM) Architecture

The H2R platform uses a unique architecture combining on-chip buffer (AQM) with High Bandwidth Memory (HBM) off-chip buffers (~2GB total). This off-chip buffer capacity is approximately 20× larger than traditional on-chip buffers. For this reason:

  • Only No-Drop PG Buffer Threshold is configurable on H1 platform

  • Port Pause Threshold Multiplier, WRED Queue Base Multiplier, and class-level multipliers are not needed

  • The massive HBM capacity naturally accommodates bursty traffic without premature ECN/PFC triggering

  • If you attempt to configure other parameters on Ararat Einstok, they will not be effective

Handling Mixed Platform Deployments

If your ACI fabric contains multiple platform types (for example, GX platform leaf switches and GX2 platform leaf switches), plan your configuration strategy:

Option 1: Single Global Policy (Recommended for Homogeneous Fabrics)

  • Create a global policy that uses only parameters supported on all platforms in your fabric

  • For example, if you mix GX and GX2, exclude Port Pause Threshold Multiplier

  • Apply this policy globally to all switches

Option 2: Platform-Specific Node-Level Policies

  • Create a global policy with conservative/balanced multiplier values that work across all platforms

  • Create node-level policies for platform groups that require additional tuning

  • For example: GX2 nodes use platform-specific policy without Port Pause settings

Option 3: Full Platform Separation

  • Create separate policies for each platform type

  • Use leaf profiles to apply platform-specific policies to nodes

  • Example: Policy ecn-gx applied to GX nodes, Policy ecn-gx2 applied to GX2 nodes

Create an ECN Buffer Tuning Policy

Create a new ECN Buffer Tuning policy that defines threshold multipliers for buffer allocation, PFC pause behavior, and WRED queue management.

Before you begin

  • Determine whether you need a single fabric-wide policy or multiple policies for different switch groups.

  • If your fabric is heterogeneous (multiple platform types), refer to Handling Mixed Platform Deployments for configuration strategy guidance.

Procedure

Step 1

Navigate to Fabric > Access Policies > Policies > Switch > ECN Buffer Tuning Policy.

Step 2

Click the Create icon (plus sign) or the Create ECN Buffer Tuning Policy button.

Step 3

In the Name field, type a descriptive name for the policy. For example: ecn-burst-optimized or ecn-conservative-control.

Step 4

Set the Admin State to Enabled. This activates the policy and allows it to be applied to switches.

Step 5

Enter values for the threshold multipliers (all accept values 1–100):

Table 8. ECN Buffer Tuning Policy Threshold Multipliers
Parameter

Description

Recommended Starting Value

No-Drop PG Buffer Threshold

Controls buffer allocation to the No-Drop Priority Group. Higher values increase buffer headroom for bursty traffic.

50

Port Pause Threshold Multiplier

Adjusts PFC pause triggering. Higher values delay pause; lower values enable more aggressive flow control.

50

WRED Queue Base Multiplier

Sets the baseline for WRED thresholds. Higher values reduce WRED sensitivity; lower values enable earlier congestion detection.

50

Note

 

he Port Pause Threshold Multiplier parameter is not available on GX2 and H1 platforms (which use dynamic AQM-based PFC). For platforms that do not support this parameter, leave it at the default.

Step 6

Click Save. The policy is now created and available for application to switches.

What to do next

After creating the policy, you must apply it to switches or switch policy groups to activate the configuration. See the tasks for applying policies globally or to specific switches.

Configure a Global ECN Buffer Tuning Policy for All Leaf and Spine Switches

Configure the default ECN Buffer Tuning policy to apply globally to all leaf and spine switches in your ACI fabric. The default policy under the ECN Buffer Tuning Policy node applies fabric-wide without requiring leaf or spine profile assignments.

Before you begin

Procedure

Step 1

Navigate to Fabric > Access Policies > Policies > Switch > ECN Buffer Tuning Policy.

Step 2

Select the default policy

Step 3

Set the Admin State to Enabled.

Step 4

Enter values for the threshold multipliers (all accept values 1–100).

Table 9. ECN Buffer Tuning Global Policy Threshold Multipliers

Parameter

Description

Recommended Starting Value

No-Drop PG Buffer Threshold

Controls buffer allocation to the No-Drop Priority Group. Higher values increase buffer headroom for bursty traffic.

50

Port Pause Threshold Multiplier

Adjusts PFC pause triggering. Higher values delay pause; lower values enable more aggressive flow control.

50

WRED Queue Base Multiplier

Sets the baseline for WRED thresholds. Higher values reduce WRED sensitivity; lower values enable earlier congestion detection.

50

Note

 

The Port Pause Threshold Multiplier parameter is not available on GX2 and H1 platforms (which use dynamic AQM-based PFC). For platforms that do not support this parameter, leave it at the default.

Step 5

Click Save. The policy is now applied globally to all leaf and spine switches in the fabric.

Step 6

Verification

After applying the policy, verify that switches have received the configuration. Wait 10–15 seconds for policy synchronization, then perform these verification steps:

  1. Log in to a Spine switch via SSH or console.

  2. Run the command: show system internal aclqos qos thresholds

  3. Confirm that:

    • ECN Buffer Config Status shows ENABLED

    • Configured Threshold Multiplier values match those you set in APIC

    • Calculated Threshold (Cells) values are displayed

All leaf and spine switches in the fabric now use the global ECN Buffer Tuning policy. Thresholds are calculated based on each platform's total buffer capacity and the multiplier values you configured. Unless you create a node-level policy override for specific switches, all switches follow this global policy.

Apply an ECN Buffer Tuning Policy to Specific Leaf Switches

Apply a node-level ECN Buffer Tuning policy to override the global policy on specific Leaf switches. This allows different groups of switches to use different buffer and congestion management settings, which is especially useful in heterogeneous fabrics or when different switch groups handle different traffic patterns.

Before you begin

  • An ECN Buffer Tuning policy has been created.

  • You know the node IDs of the switches you want to apply the policy to.

  • You have reviewed Supported Platforms and Parameter Availability if your fabric contains multiple platform types.

Procedure

Step 1

Navigate to Fabric > Access Policies > Switches > Leaf Switches > Profiles.

Step 2

Create a new Leaf Profile or select an existing one. The profile name should reflect the switches it contains (for example, Leaf-edge-pods or Leaf-burst-traffic).

Step 3

Create a Leaf Selector and specify the node range (for example, nodes 101–102).

Step 4

In the selector, link a Leaf Policy Group . You can use an existing policy group or create a new custom one, if needed (for example, LeafPolGrp-custom).

Step 5

In the Leaf Policy Group, set the ECN Buffer Tuning Policy to the policy you want to apply to this node range.

Step 6

Click Save.

Leaf switches in the specified node range now use the assigned ECN Buffer Tuning policy, overriding any global policy. All other Leaf switches continue to use the global policy (if one is applied) or factory defaults (if no global policy is enabled).

Apply an ECN Buffer Tuning Policy to Specific Spine Switches

Apply a node-level ECN Buffer Tuning policy to override the global policy on specific spine switches. This allows different groups of switches to use different buffer and congestion management settings, which is especially useful in heterogeneous fabrics or when different switch groups handle different traffic patterns.

Before you begin

  • An ECN Buffer Tuning policy has been created.

  • You know the node IDs of the switches you want to apply the policy to.

  • You have reviewed Supported Platforms and Parameter Availability if your fabric contains multiple platform types.

Procedure

Step 1

Navigate to Fabric > Access Policies > Switches > Spine Switches > Profiles.

Step 2

Create a new Spine Profile or select an existing one. The profile name should reflect the switches it contains (for example, spine-edge-pods or spine-burst-traffic).

Step 3

Create a Spine Selector and specify the node range (for example, nodes 101–102).

Step 4

In the selector, link a Spine Policy Group . You can use an existing policy group or create a new custom one, if needed (for example, spinePolGrp-custom).

Step 5

In the spine Policy Group, set the ECN Buffer Tuning Policy to the policy you want to apply to this node range.

Step 6

Click Save.

Spine switches in the specified node range now use the assigned ECN Buffer Tuning policy, overriding any global policy. All other spine switches continue to use the global policy (if one is applied) or factory defaults (if no global policy is enabled).

Configure Class-Level WRED Queue Thresholds

Configure Weighted Random Early Detection (WRED) queue thresholds on a per-QoS-class basis. This allows you to apply different ECN marking sensitivities to different traffic classes, prioritizing critical applications while managing best-effort traffic more aggressively.

Before you begin

  • At least one QoS class is already configured in APIC.

  • An ECN Buffer Tuning policy has been created and applied globally or to specific switches.

Procedure

Step 1

Navigate to Fabric > Access Policies > Policies > Global > QoS Class.

Step 2

Select or create a QoS class (for example, Level-1).

Step 3

Expand the Congestion or QoS Threshold section. Look for the WRED Queue Threshold Multiplier field.

Step 4

In WRED Queue Threshold Multiplier, enter a value (1–100):

  • Lower values (20–40): More aggressive congestion management for best-effort traffic.

  • Mid-range values (50): Apply balanced ECN marking

  • Higher values (70–90): Reduced sensitivity for latency-sensitive traffic.

Step 5

Click Save.

Switches now apply different ECN thresholds to each class. During congestion:

  • High-priority classes (higher multipliers) experience less ECN marking and lower packet loss

  • Low-priority classes (lower multipliers) experience more ECN marking, reducing their throughput and freeing buffer for high-priority traffic

  • Overall network fairness and QoS objectives are achieved

Troubleshooting Cisco APIC QoS Policies

The following sections summarize common troubleshooting scenarios for Cisco APIC QoS.

Unable to Update a Configured QoS Policy

  1. Invoke the following API to ensure that qospDscpRule is present on the leaf. 

    GET https://192.0.20.123/api/node/class/qospDscpRule.xml

  2. Ensure that the QoS rules are accurately configured and associated to the EPG ID to which the policy is attached.

    Use the following NX-OS style CLI commands to verify the configuration.

    leaf1# show vlan  
leaf1# show system internal aclqos qos policy detail
 
apic1# show running-config tenant tenant-name policy-map type qos custom-qos-policy-name 
apic1# show running-config tenant tenant-name application application-name epg epg-name

Show QoS Interface Statistics by Using the CLI

CLI displays statistics for eth1/1 for only QoS classes – level1, leve2, level3, level4, level5, level6, and policy-plane – if you don’t use the "detail" option. 

NXOS ibash cli: tor-leaf1# show queuing interface ethernet 1/1 [detail]

If you want to display statistics for control-plane and span classes for an interface, you need to use CLI with the "detail" option.

Example: fabric 107 show queuing interface ethernet 1/1 detail

APIC CLI:
swtb123-ifc1# fabric node_id show queuing interface ethernet 1/1
Given below is the expected output:
swtb95-leaf1# show queuing interface ethernet 1/31
====================================
Queuing stats for ethernet 1/31
====================================
Qos Class level3
====================================
Rx Admit Pkts : 0 Tx Admit Pkts : 0 
Rx Admit Bytes: 0 Tx Admit Bytes: 0 
Rx Drop Pkts : 0 Tx Drop Pkts : 0 
Rx Drop Bytes : 0 Tx Drop Bytes : 0
====================================
Qos Class level2
====================================
Rx Admit Pkts : 0 Tx Admit Pkts : 0 
Rx Admit Bytes: 0 Tx Admit Bytes: 0 
Rx Drop Pkts : 0 Tx Drop Pkts : 0 
Rx Drop Bytes : 0 Tx Drop Bytes : 0
====================================
Qos Class level1
====================================
Rx Admit Pkts : 0 Tx Admit Pkts : 0 
Rx Admit Bytes: 0 Tx Admit Bytes: 0 
Rx Drop Pkts : 0 Tx Drop Pkts : 0 
Rx Drop Bytes : 0 Tx Drop Bytes : 0
====================================
Qos Class level6
====================================
Rx Admit Pkts : 0 Tx Admit Pkts : 401309848 
Rx Admit Bytes: 0 Tx Admit Bytes: 47354562064 
Rx Drop Pkts : 0 Tx Drop Pkts : 2066740320 
Rx Drop Bytes : 0 Tx Drop Bytes : 140538341760

Show QoS Interface Statistics by Using the APIC GUI

Use the APIC GUI to view the QoS statistics.

Navigate to Fabric->Inventory>Pod Number > Node Hostname > Physical Interfaces > Interface-> QoS Stats to view the the QoS stats as shown below:

Verifying IPN Ingress Interface Settings

This section describes how to verify the IPN ingress interface settings you have configured in Configuring QoS On IPN Devices. 

IPNPOD2# show policy-map interface ethernet 1/50.4 input

Global statistics status :   enabled

Ethernet1/50.4

  Service-policy (qos) input:   ACI-CLASSIFICATION 
    SNMP Policy Index:  285215377

    Class-map (qos):   CONTROL-TRAFFIC (match-all)

     Slot 1
        1434 packets 
     Aggregate forwarded :
        1434 packets 
      Match: dscp 48,56
      set qos-group 7

    Class-map (qos):   UserLevel1 (match-all)
     Aggregate forwarded :
        0 packets 
      Match: dscp 46
      set qos-group 6

    Class-map (qos):   UserLevel2 (match-all)
     Aggregate forwarded :
        0 packets 
      Match: dscp 24
      set qos-group 3

    Class-map (qos):   UserLevel3 (match-all)

     Slot 1
        25 packets 
     Aggregate forwarded :
        25 packets 
      Match: dscp 0
      set qos-group 0

    Class-map (qos):   SpanTraffic (match-all)
     Aggregate forwarded :
        0 packets 
      Match: dscp 8
      set qos-group 1

    Class-map (qos):   iTraceroute (match-all)
     Aggregate forwarded :
        0 packets 
      Match: dscp 40
      set qos-group 5
 
 
IPNPOD2# show policy-map interface ethernet 1/49.4 input
Global statistics status :   enabled

Ethernet1/49.4

  Global statistics status :   enabled

Ethernet1/49.4

  Service-policy (qos) input:   ACI-CLASSIFICATION 
    SNMP Policy Index:  285215373

    Class-map (qos):   CONTROL-TRAFFIC (match-all)

     Slot 1
        5149 packets 
     Aggregate forwarded :
        5149 packets 
      Match: dscp 48,56
      set qos-group 7

    Class-map (qos):   UserLevel1 (match-all)
     Aggregate forwarded :
        0 packets 
      Match: dscp 46
      set qos-group 6

    Class-map (qos):   UserLevel2 (match-all)
     Aggregate forwarded :
        0 packets 
      Match: dscp 24
      set qos-group 3

    Class-map (qos):   UserLevel3 (match-all)

     Slot 1
        960 packets 
     Aggregate forwarded :
        960 packets 
      Match: dscp 0
      set qos-group 0

    Class-map (qos):   SpanTraffic (match-all)
     Aggregate forwarded :
        0 packets 
      Match: dscp 8
      set qos-group 1

    Class-map (qos):   iTraceroute (match-all)
     Aggregate forwarded :
        0 packets 
      Match: dscp 40
      set qos-group 5

Verifying IPN Egress Interface Settings

This section describes how to verify the IPN egress interface settings you have configured in Configuring QoS On IPN Devices. 

IPNPOD1# show queuing interface e 1/3 | b "GROUP 7"

slot  1
=======


Egress Queuing for Ethernet1/3 [System]
------------------------------------------------------------------------------
QoS-Group# Bandwidth% PrioLevel                Shape                   QLimit
                                   Min          Max        Units   
------------------------------------------------------------------------------
      7             -         1           -            -     -            9(D)
      6             -         2           -            -     -            9(D)
      5             0         -           -            -     -            9(D)
      4             0         -           -            -     -            9(D)
      3            20         -           -            -     -            9(D)
      2             0         -           -            -     -            9(D)
      1             1         -           -            -     -            9(D)
      0            59         -           -            -     -            9(D)
+-------------------------------------------------------------+
|                              QOS GROUP 0                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |          125631|              70|
|                   Tx Byts |        42902871|            8836|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                              QOS GROUP 1                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |               0|               0|
|                   Tx Byts |               0|               0|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                              QOS GROUP 2                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |               0|               0|
|                   Tx Byts |               0|               0|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                              QOS GROUP 3                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |               0|               0|
|                   Tx Byts |               0|               0|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                              QOS GROUP 4                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |               0|               0|
|                   Tx Byts |               0|               0|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                              QOS GROUP 5                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |               0|               0|
|                   Tx Byts |               0|               0|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                              QOS GROUP 6                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |          645609|             217|
|                   Tx Byts |       115551882|           25606|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                              QOS GROUP 7                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |           23428|               9|
|                   Tx Byts |         4132411|            1062|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                      CONTROL QOS GROUP                      |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |            6311|               0|
|                   Tx Byts |          809755|               0|
|            Tail Drop Pkts |               0|               0|
|            Tail Drop Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                         SPAN QOS GROUP                      |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |               0|               0|
|                   Tx Byts |               0|               0|
|            Tail Drop Pkts |               0|               0|
|            Tail Drop Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+


Ingress Queuing for Ethernet1/3
-----------------------------------------------------
QoS-Group#                 Pause                     
           Buff Size       Pause Th      Resume Th   
-----------------------------------------------------
      7              -            -            - 
      6              -            -            - 
      5              -            -            - 
      4              -            -            - 
      3              -            -            - 
      2              -            -            - 
      1              -            -            - 
      0              -            -            - 


Per Port Ingress Statistics
--------------------------------------------------------
     Hi Priority Drop Pkts                           0
    Low Priority Drop Pkts                           0
Ingress Overflow Drop Pkts                           0



PFC Statistics
------------------------------------------------------------------------------
TxPPP:                    0,   RxPPP:                    0
------------------------------------------------------------------------------
PFC_COS QOS_Group   TxPause             TxCount   RxPause             RxCount
      0         0  Inactive                   0  Inactive                   0
      1         0  Inactive                   0  Inactive                   0
      2         0  Inactive                   0  Inactive                   0
      3         0  Inactive                   0  Inactive                   0
      4         0  Inactive                   0  Inactive                   0
      5         0  Inactive                   0  Inactive                   0
      6         0  Inactive                   0  Inactive                   0
      7         0  Inactive                   0  Inactive                   0
------------------------------------------------------------------------------
 
 
 
IPNPOD2# show queuing interface e 1/4 
 
 slot  1
=======


Egress Queuing for Ethernet1/4 [System]
------------------------------------------------------------------------------
QoS-Group# Bandwidth% PrioLevel                Shape                   QLimit
                                   Min          Max        Units   
------------------------------------------------------------------------------
      7             -         1           -            -     -            9(D)
      6             -         2           -            -     -            9(D)
      5             0         -           -            -     -            9(D)
      4             0         -           -            -     -            9(D)
      3            20         -           -            -     -            9(D)
      2             0         -           -            -     -            9(D)
      1             1         -           -            -     -            9(D)
      0            59         -           -            -     -            9(D)
+-------------------------------------------------------------+
|                              QOS GROUP 0                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |           63049|               0|
|                   Tx Byts |        15968783|               0|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                              QOS GROUP 1                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |               0|               0|
|                   Tx Byts |               0|               0|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                              QOS GROUP 2                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |               0|               0|
|                   Tx Byts |               0|               0|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                              QOS GROUP 3                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |               0|               0|
|                   Tx Byts |               0|               0|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                              QOS GROUP 4                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |               0|               0|
|                   Tx Byts |               0|               0|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                              QOS GROUP 5                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |               0|               0|
|                   Tx Byts |               0|               0|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                              QOS GROUP 6                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |         1141418|               0|
|                   Tx Byts |       237770324|               0|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                              QOS GROUP 7                    |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |           32440|               0|
|                   Tx Byts |         6986806|               0|
| WRED/AFD & Tail Drop Pkts |               0|               0|
| WRED/AFD & Tail Drop Byts |               0|               0|
|              Q Depth Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                      CONTROL QOS GROUP                      |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |            6275|               0|
|                   Tx Byts |          804748|               0|
|            Tail Drop Pkts |               0|               0|
|            Tail Drop Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+
|                         SPAN QOS GROUP                      |
+-------------------------------------------------------------+
|                           |  Unicast       |Multicast       |
+-------------------------------------------------------------+
|                   Tx Pkts |               0|               0|
|                   Tx Byts |               0|               0|
|            Tail Drop Pkts |               0|               0|
|            Tail Drop Byts |               0|               0|
|       WD & Tail Drop Pkts |               0|               0|
+-------------------------------------------------------------+


Ingress Queuing for Ethernet1/4
-----------------------------------------------------
QoS-Group#                 Pause                     
           Buff Size       Pause Th      Resume Th   
-----------------------------------------------------
      7              -            -            - 
      6              -            -            - 
      5              -            -            - 
      4              -            -            - 
      3              -            -            - 
      2              -            -            - 
      1              -            -            - 
      0              -            -            - 


Per Port Ingress Statistics
--------------------------------------------------------
     Hi Priority Drop Pkts                           0
    Low Priority Drop Pkts                           0
Ingress Overflow Drop Pkts                           0



PFC Statistics
------------------------------------------------------------------------------
TxPPP:                    0,   RxPPP:                    0
------------------------------------------------------------------------------
PFC_COS QOS_Group   TxPause             TxCount   RxPause             RxCount
      0         0  Inactive                   0  Inactive                   0
      1         0  Inactive                   0  Inactive                   0
      2         0  Inactive                   0  Inactive                   0
      3         0  Inactive                   0  Inactive                   0
      4         0  Inactive                   0  Inactive                   0
      5         0  Inactive                   0  Inactive                   0
      6         0  Inactive                   0  Inactive                   0
      7         0  Inactive                   0  Inactive                   0
------------------------------------------------------------------------------