Cisco 7600 Series Router Software Configuration Guide, Cisco IOS Release 15S

Port Types Supported by PFC QoS

The PFC does not provide QoS for FlexWAN module ports. Refer to this publication for information about FlexWAN module QoS features:

http://www.cisco.com/en/US/docs/routers/7600/install_config/flexwan_config/flexwan-config-guide.html

PFC QoS supports LAN ports. LAN ports are Ethernet ports on Ethernet switching modules, except for the 4-port Gigabit Ethernet WAN (GBIC) modules (OSM-4GE-WAN and OSM-2+4GE-WAN+). Some OSMs have four Ethernet LAN ports in addition to WAN ports.

PFC QoS supports optical services module (OSM) ports. OSM ports are the WAN ports on OSMs.

Overview

Typically, networks operate on a best-effort delivery basis, which means that all traffic has equal priority and an equal chance of being delivered in a timely manner. When congestion occurs, all traffic has an equal chance of being dropped.

QoS makes network performance more predictable and bandwidth utilization more effective. QoS selects (classifies) network traffic, uses or assigns QoS labels to indicate priority, makes the packets comply with the configured resource usage limits (polices the traffic and marks the traffic), and provides congestion avoidance where resource contention exists.

PFC QoS classification, policing, marking, and congestion avoidance is implemented in hardware on the PFC, DFCs, and in LAN switching module port Application Specific Integrated Circuits (ASICs).

Note Cisco 7600 series routers do not support all of the MQC features (for example, Committed Access Rate (CAR)) for traffic that is Layer 3 switched or Layer 2 switched in hardware. Because queuing is implemented in the port ASICs, Cisco 7600 series routers do not support MQC-configured queuing.

Figure 48-1 shows an overview of QoS processing in a Cisco 7600 series router.

Figure 48-1 PFC QoS Feature Processing Overview

The PFC QoS features are applied in this order:

1. Ingress port PFC QoS features:

– Port trust state—In PFC QoS, trust means to accept as valid and use as the basis of the initial internal DSCP value. Ports are untrusted by default, which sets the initial internal DSCP value to zero. You can configure ports to trust received CoS, IP precedence, or DSCP.

– Layer 2 CoS remarking—PFC QoS applies Layer 2 CoS remarking, which marks the incoming frame with the port CoS value, in these situations:

—If a port is configured as untrusted.

—If a port is configured as trusted, but the traffic is not in an ISL, 802.1Q, or 802.1p frame.

On OSM ATM and POS ports, PFC QoS always sets CoS equal to zero.

– Congestion avoidance —If you configure an Ethernet LAN port to trust CoS, QoS classifies the traffic on the basis of its Layer 2 CoS value and assigns it to an ingress queue to provide congestion avoidance. Layer 3 DSCP-based queue mapping is available only on WS-X6708-10GE ports.

2. PFC and DFC QoS features:

– Internal DSCP —On the PFC and DFCs, QoS associates an internal DSCP value with all traffic to classify it for processing through the system. There is an initial internal DSCP based on the traffic trust state and a final internal DSCP. The final internal DSCP can be the same as the initial value or an MQC policy map can set it to a different value.

– MQC policy maps—MQC policy maps can do one or more of these operations:

—Change the trust state of the traffic (bases the internal DSCP value on a different QoS label)

—Set the initial internal DSCP value (only for traffic from untrusted ports)

—Mark the traffic

—Police the traffic

3. Egress Ethernet LAN port QoS features:

– Layer 3 DSCP marking with the final internal DSCP (optionally with PFC)

– Layer 2 CoS marking mapped from the final internal DSCP

– Layer 2 CoS-based congestion avoidance. (Layer 3 DSCP-based queue mapping is available only on WS-X6708-10GE ports.)

These figures provide more detail about the relationship between QoS and the router components:

• Figure 48-2, Traffic Flow and PFC QoS Features with the PFC
• Figure 48-3, PFC QoS Features and Component Overview

Figure 48-2 Traffic Flow and PFC QoS Features with the PFC

Figure 48-2 shows how traffic flows through the QoS features with a PFC:

• Traffic can enter on any type of port and exit on any type of port.
• DFCs implement PFC QoS locally on switching modules.
• For FlexWAN module traffic:

– Ingress FlexWAN QoS features can be applied to FlexWAN ingress traffic.

– Ingress FlexWAN traffic can be Layer 3-switched by the PFC or routed in software by the MSFC.

– Egress PFC QoS is not applied to FlexWAN ingress traffic.

– Egress FlexWAN QoS can be applied to FlexWAN egress traffic.

• For LAN-port traffic:

– Ingress LAN-port QoS features can be applied to LAN-port ingress traffic.

– Ingress PFC QoS can be applied to LAN-port ingress traffic.

– Ingress LAN-port traffic can be Layer-2 or Layer-3 switched by the PFC or routed in software by the MSFC.

– Egress PFC QoS and egress LAN-port QoS can be applied to LAN-port egress traffic.

• For OSM traffic:

– Ingress OSM-port QoS features can be applied to OSM-port ingress traffic.

– Ingress PFC QoS can be applied to OSM-port ingress traffic.

– Ingress OSM-port traffic can be Layer-3 switched by the PFC or routed in software by the MSFC.

– Egress PFC QoS and egress OSM-port QoS can be applied to OSM-port egress traffic.

Figure 48-3 PFC QoS Features and Component Overview

Component Overview

These sections provide more detail about the role of the following components in PFC QoS decisions and processes:

• Ingress LAN Port PFC QoS Features
• PFC and DFC QoS Features
• Egress Port QoS Features

Ingress LAN Port PFC QoS Features

These sections provide an overview of the ingress port QoS features:

• Flowchart of Ingress LAN Port PFC QoS Features
• Port Trust
• Congestion Avoidance

Flowchart of Ingress LAN Port PFC QoS Features

Figure 48-4 shows how traffic flows through the ingress LAN port PFC QoS features.

Figure 48-4 Ingress Port QoS Features

Note Ingress CoS mutation is supported only on 802.1Q tunnel ports. DSCP-based queue mapping is supported only on WS-X6708-10GE ports.

Port Trust

In PFC QoS, trust means to accept as valid and use as the basis of the initial internal DSCP value. You can configure ports as untrusted or you can configure them to trust these QoS values:

• Layer 2 CoS

– A port configured to trust CoS is called a trust CoS port.

– Traffic received through a trust CoS port or configured by a policy map to trust CoS is called trust CoS traffic.

Note Not all traffic carries a CoS value. Only ISL, 802.1Q, and 802.1P traffic carries a CoS value. PFC QoS applies the port CoS value to any traffic that does not carry a CoS value. On untrusted ports, PFC QoS applies the port CoS value to all traffic, overwriting any received CoS value. Received CoS values are preserved only on ports configured to trust CoS.

• IP precedence

– A port configured to trust IP precedence is called a trust IP precedence port.

– Traffic received through a trust IP precedence port or configured by a policy map to trust IP precedence is called trust IP precedence traffic.

• DSCP

– A port configured to trust DSCP is called a trust DSCP port.

– Traffic received through a trust DSCP port or configured by a policy map to trust DSCP is called trust DSCP traffic.

Traffic received through an untrusted port is called untrusted traffic.

Congestion Avoidance

PFC QoS implements congestion avoidance on trust CoS ports. On a trust CoS port, QoS classifies the traffic on the basis of its Layer 2 CoS value and assigns it to an ingress queue to provide congestion avoidance. In Release 12.2(33)SRC and later releases, you can configure WS-X6708-10GE trust DSCP ports to use received DSCP values for congestion avoidance. See the “Classification and Marking at Trust CoS Ingress LAN Ports” section for more information about ingress congestion avoidance.

PFC and DFC QoS Features

These sections describe PFCs and DFCs as they relate to QoS:

• Supported Policy Feature Cards
• Supported Distributed Forwarding Cards
• PFC and DFC QoS Feature List and Flowchart
• Internal DSCP Values

Supported Policy Feature Cards

The policy feature card (PFC) is a daughter card that resides on the supervisor engine. The PFC provides QoS in addition to other functionality. The following PFCs are supported on Cisco 7600 series routers:

• PFC3B on the Supervisor Engine 720 and Supervisor Engine 32
• PFC3BXL on the Supervisor Engine 720
• The PFC3C and PFC3CXL on the Route Switch Processor 720 (RSP720)

Supported Distributed Forwarding Cards

The PFC sends a copy of the QoS policies to the distributed forwarding card (DFC) to provide local support for the QoS policies, which enables the DFCs to support the same QoS features that the PFC supports.

The following DFCs are supported on the Cisco 7600 series routers:

• WS-F6K-DFC3B, WS-F6K-DFC3BXL, for use on dCEF256 and CEF256 modules with a Supervisor Engine 720.
• WS-F6700-DFC3B, WS-F6700-DFC3BXL, for use on CEF720 modules with a Supervisor Engine 720.

PFC and DFC QoS Feature List and Flowchart

Table 48-1 lists the QoS features supported on the different versions of PFCs and DFCs.

Figure 48-5 shows how traffic flows through the QoS features on the PFC and DFCs.

Figure 48-5 QoS Features on the PFC and DFCs

Note The DSCP transparency feature makes writing the egress DSCP value into the Layer 3 ToS byte optional.

Internal DSCP Values

During processing, PFC QoS represents the priority of all traffic (including non-IP traffic) with an internal DSCP value. On the PFC, before any marking or policing takes place, PFC QoS derives the initial internal DSCP value as follows:

• From received CoS values or from port CoS values for trust CoS traffic.

Note Traffic from an untrusted ingress port has the port CoS value. If traffic from an untrusted port matches a trust CoS policer, PFC QoS derives the internal DSCP value from the ingress port CoS value.

• Mapped from received IP precedence values for trust IP precedence traffic.
• From received DSCP values for trust DSCP traffic.
• Mapped from port CoS or DSCP values configured in policy maps for untrusted traffic.

For trust CoS traffic and trust IP precedence traffic, PFC QoS uses configurable maps to derive the internal 6-bit DSCP value from CoS or IP precedence, which are 3-bit values.

Marking and policing on the PFC can change the initial internal DSCP value to a final internal DSCP value, which is then used for all subsequently applied QoS features.

Port-Based PFC QoS and VLAN-Based PFC QoS

You can configure each ingress LAN port for either physical port-based PFC QoS (default) or VLAN-based PFC QoS and attach a policy map to the selected interface.

On ports configured for port-based PFC QoS, you can attach a policy map to the ingress LAN port as follows:

• On a nontrunk ingress LAN port configured for port-based PFC QoS, all traffic received through the port is subject to the policy map attached to the port.
• On a trunking ingress LAN port configured for port-based PFC QoS, traffic in all VLANs received through the port is subject to the policy map attached to the port.

On a nontrunk ingress LAN port configured for VLAN-based PFC QoS, traffic received through the port is subject to the policy map attached to the port’s VLAN.

On a trunking ingress LAN port configured for VLAN-based PFC QoS, traffic received through the port is subject to the policy map attached to the traffic’s VLAN.

Egress Port QoS Features

These sections describe egress port QoS features:

• Flowchart of Egress LAN Port Features
• Egress CoS Values
• Egress DSCP Mutation
• Egress ToS Byte
• Egress PFC QoS Interfaces
• Egress ACL Support for Remarked DSCP
• Marking on Egress OSM Ports

Flowchart of Egress LAN Port Features

Figure 48-6 shows how traffic flows through the QoS features on egress LAN ports.

Figure 48-6 Egress LAN Port Scheduling, Congestion Avoidance, and Marking

Egress CoS Values

For all egress traffic, PFC QoS uses a configurable map to derive a CoS value from the final internal DSCP value associated with the traffic. PFC QoS sends the derived CoS value to the egress LAN ports for use in scheduling and to be written into ISL and 802.1Q frames.

Note With Release 12.2(33)SRC and later releases, you can configure WS-X6708-10GE ports to use the final internal DSCP value for egress LAN port classification and congestion avoidance. See Configuring DSCP-Based Queue Mapping.

Egress DSCP Mutation

You can configure 15 egress DSCP mutation maps to mutate the internal DSCP value before it is written in the egress ToS byte. You can attach egress DSCP mutation maps to any interface that PFC QoS supports.

Note ● If you configure egress DSCP mutation, PFC QoS does not derive the egress CoS value from the mutated DSCP value.

Egress ToS Byte

Except when DSCP transparency is enabled, PFC QoS creates a ToS byte for egress IP traffic from the final internal or mutated DSCP value and sends it to the egress port to be written into IP packets. For trust DSCP and untrusted IP traffic, the ToS byte includes the original two least-significant bits from the received ToS byte.

The internal or mutated DSCP value can mimic an IP precedence value (see the “IP Precedence and DSCP Values” section).

Egress PFC QoS Interfaces

You can attach an output policy map to a Layer 3 interface (either a LAN port configured as a Layer 3 interface or a VLAN interface) to apply a policy map to egress traffic.

Note ● Output policies do not support microflow policing.

• You cannot apply microflow policing to ARP traffic.
• You cannot set a trust state in an output policy.

Egress ACL Support for Remarked DSCP

Note Egress ACL support for remarked DSCP is also known as packet recirculation.

The PFC supports egress ACL support for remarked DSCP, which enables IP precedence-based or DSCP-based egress QoS filtering to use any IP precedence or DSCP policing or marking changes made by ingress PFC QoS.

Without egress ACL support for remarked DSCP, egress QoS filtering uses received IP precedence or DSCP values; it does not use any IP precedence or DSCP changes made by ingress PFC QoS as the result of policing or marking.

The PFC provides egress PFC QoS only for Layer 3-switched and routed traffic on egress Layer 3 interfaces (either LAN ports configured as Layer 3 interfaces or VLAN interfaces).

You configure egress ACL support for remarked DSCP on ingress Layer 3 interfaces (either LAN ports configured as Layer 3 interfaces or VLAN interfaces).

On interfaces where egress ACL support for remarked DSCP is configured, the PFC processes each QoS-filtered IP packet twice: once to apply ingress PFC QoS and once to apply egress PFC QoS.

After packets have been processed by ingress PFC QoS and any policing or marking changes have been made, the packets are processed again on the ingress interface by any configured Layer 2 features (for example, VACLs) before being processed by egress PFC QoS.

On an interface where egress ACL support for remarked DSCP is configured, if a Layer 2 feature matches the ingress-QoS-modified IP precedence or DSCP value, the Layer 2 feature might redirect or drop the matched packets, which prevents them from being processed by egress QoS.

After packets have been processed by ingress PFC QoS and any policing or marking changes have been made, the packets are processed on the ingress interface by any configured Layer 3 features (for example, ingress Cisco IOS ACLs, policy based routing (PBR), etc.) before being processed by egress PFC QoS.

The Layer 3 features configured on an interface where egress ACL support for remarked DSCP is configured might redirect or drop the packets that have been processed by ingress PFC QoS, which would prevent them from being processed by egress PFC QoS.

Marking on Egress OSM Ports

Ingress PFC QoS sets DSCP values that can be used by the OSM egress QoS features (see Figure 48-7).

Figure 48-7 Egress WAN Port Marking

Understanding Classification and Marking

The following sections describe where and how classification and marking occur on the Cisco 7600 series routers:

• Classification and Marking at Trusted and Untrusted Ingress Ports
• Classification and Marking at Ingress OSM Ports
• Classification and Marking on the PFC Using Service Policies and Policy Maps
• Classification and Marking on the MSFC

Classification and Marking at Trusted and Untrusted Ingress Ports

The trust state of an ingress port determines how the port marks, schedules, and classifies received Layer 2 frames, and whether or not congestion avoidance is implemented. These are the port trust states:

• Untrusted (default)
• Trust IP precedence
• Trust DSCP
• Trust CoS

Ingress LAN port classification, marking, and congestion avoidance use Layer 2 CoS values only and do not use or set Layer 3 IP precedence or DSCP values.

In Release 12.2(33)SRC and later releases, you can configure WS-X6708-10GE ports to use received DSCP values for ingress LAN port classification and congestion avoidance ( See Configuring DSCP-Based Queue Mapping.).

The following sections describe classification and marking at trusted and untrusted ingress ports:

• Classification and Marking at Untrusted Ingress Ports
• Classification and Marking at Trusted Ingress Ports

Classification and Marking at Untrusted Ingress Ports

PFC QoS Layer 2 remarking marks all frames received through untrusted ingress ports with the port CoS value (the default is zero).

To map the port CoS value applied to untrusted traffic to the initial internal DSCP value, configure a trust CoS policy map that matches the ingress traffic.

Classification and Marking at Trusted Ingress Ports

You should configure ports to trust only if they receive traffic that carries valid QoS labels. QoS uses the received QoS labels as the basis of initial internal DSCP value. After the traffic enters the router, you can apply a different trust state to traffic with a policy map. For example, traffic can enter the router through a trust CoS port, and then you can use a policy map to trust IP precedence or DSCP, which uses the trusted value as the basis of the initial internal DSCP value, instead of the QoS label that was trusted at the port.

These sections describe classification and marking at trusted ingress ports:

• Classification and Marking at Trust CoS Ingress LAN Ports
• Classification and Marking at Trust IP precedence Ingress Ports
• Classification and Marking at Trust DSCP Ingress Ports

Classification and Marking at Trust CoS Ingress LAN Ports

You should configure LAN ports to trust CoS only if they receive traffic that carries valid Layer 2 CoS.

When an ISL frame enters the router through a trusted ingress LAN port, PFC QoS accepts the three least significant bits in the User field as a CoS value. When an 802.1Q frame enters the router through a trusted ingress LAN port, PFC QoS accepts the User Priority bits as a CoS value. PFC QoS Layer 2 remarking marks all traffic received in untagged frames with the ingress port CoS value.

On ports configured to trust CoS, PFC QoS does the following:

• PFC QoS maps the received CoS value in tagged trust CoS traffic to the initial internal DSCP value.
• PFC QoS maps the ingress port CoS value applied to untagged trusted traffic to the initial internal DSCP value.

Note A policy map can change the trust state of the traffic after it enters the router and use received IP precedence or DSCP as the basis of the initial internal DSCP value.

• PFC QoS enables the CoS-based ingress queues and thresholds to provide congestion avoidance. See the “Understanding Port-Based Queue Types” section for more information about ingress queues and thresholds.

Classification and Marking at Trust IP precedence Ingress Ports

You should configure ports to trust IP precedence only if they receive traffic that carries valid Layer 3 IP precedence. For traffic from trust IP precedence ports, PFC QoS maps the received IP precedence value to the initial internal DSCP value, unless there is a policy map that changes the trust state of the traffic. Because the ingress port queues and thresholds use Layer 2 CoS, PFC QoS does not implement ingress port congestion avoidance on ports configured to trust IP precedence. PFC does not mark any traffic on ingress ports configured to trust IP precedence.

Classification and Marking at Trust DSCP Ingress Ports

You should configure ports to trust DSCP only if they receive traffic that carries valid Layer 3 DSCP. For traffic from trust DSCP ports, PFC QoS uses the received DSCP value as the initial internal DSCP value, unless there is a policy map that changes the trust state of the traffic. Because the ingress port queues and thresholds use Layer 2 CoS, PFC QoS does not implement ingress port congestion avoidance on ports configured to trust DSCP. PFC does not mark any traffic on ingress ports configured to trust received DSCP.

In Release 12.2(33)SRC and later releases, you can enable DSCP-based ingress queues and thresholds on WS-X6708-10GE ports to provide congestion avoidance (See Configuring DSCP-Based Queue Mapping).

Classification and Marking at Ingress OSM Ports

PFC QoS associates CoS zero with all traffic received through ingress OSM ports. You can configure ingress OSM port trust states that can be used by the PFC to set IP precedence or DSCP values and the CoS value. You can configure the trust state of each ingress OSM port as follows:

• Untrusted (default)
• Trust IP precedence
• Trust DSCP
• Trust CoS (CoS is always zero for POS and ATM OSM ports because the port CoS value is not configurable on POS and ATM OSM ports.)

Classification and Marking on the PFC Using Service Policies and Policy Maps

PFC QoS supports classification and marking with service policies that attach one policy map to these interface types to apply ingress PFC QoS:

• Each ingress port (except FlexWAN interfaces)
• Each EtherChannel port-channel interface
• Each VLAN interface

You can attach one policy map to each Layer 3 interface (except FlexWAN interfaces) to apply egress PFC QoS.

Each policy map can contain multiple policy-map classes. You can configure a separate policy-map class for each type of traffic handled by the interface. There are two ways to configure filtering in policy-map classes:

• Access control lists (ACLs)
• Class-map match commands for IP precedence and DSCP values

Policy-map classes specify actions with the following optional commands:

• Policy-map set commands—For untrusted traffic, PFC QoS can use configured IP precedence or DSCP values as the final internal DSCP value. The “IP Precedence and DSCP Values” section shows the bit values for IP precedence and DSCP.
• Policy-map class trust commands—PFC QoS applies the policy-map class trust state to matched ingress traffic, which then uses the trusted value as the basis of its initial internal DSCP value, instead of the QoS label that was trusted at the port (if any). In a policy map, you can trust CoS, IP precedence, or DSCP.

Note A trust CoS policy map cannot restore received CoS in traffic from untrusted ingress LAN ports. Traffic from untrusted ingress LAN ports always has the port CoS value.

• Aggregate and microflow policers—PFC QoS can use policers to either mark or drop both conforming and nonconforming traffic.

Classification and Marking on the MSFC

PFC QoS sends IP traffic to the MSFC with the final internal DSCP values. CoS is equal to IP precedence in all traffic sent from the MSFC to egress ports.

Figure 48-8 Marking with PFC3 and MSFC2A or MSFC3

Note Traffic that is Layer 3 switched on the PFC does not go through the MSFC and retains the CoS value assigned by the PFC.

Policers

These sections describe policers:

• Overview of Policers
• Aggregate Policers
• Microflow Policers

Overview of Policers

Policing allows you to rate limit incoming and outgoing traffic so that it adheres to the traffic forwarding rules defined by the QoS configuration. Sometimes these configured rules for how traffic should be forwarded through the system are referred to as a contract. If the traffic does not adhere to this contract, it is marked down to a lower DSCP value or dropped.

Policing does not buffer out-of-profile packets. As a result, policing does not affect transmission delay. In contrast, traffic shaping works by buffering out-of-profile traffic, which moderates the traffic bursts. (PFC QoS does not support shaping.)

The PFC supports both ingress and egress PFC QoS, which includes ingress and egress policing. Traffic shaping is supported on some WAN modules.

Note Policers can act on ingress traffic per-port or per-VLAN. With a PFC, for egress traffic, the policers can act per-VLAN only.

You can create policers to do the following:

• Mark traffic
• Limit bandwidth utilization and mark traffic

Aggregate Policers

PFC QoS applies the bandwidth limits specified in an aggregate policer cumulatively to all flows in matched traffic. For example, if you configure an aggregate policer to allow 1 Mbps for all TFTP traffic flows on VLAN 1 and VLAN 3, it limits the TFTP traffic for all flows combined on VLAN 1 and VLAN 3 to 1 Mbps.

• You define per-interface aggregate policers in a policy map class with the police command. If you attach a per-interface aggregate policer to multiple ingress ports, it polices the matched traffic on each ingress port separately.
• You create named aggregate policers with the mls qos aggregate-policer command. If you attach a named aggregate policer to multiple ingress ports, it polices the matched traffic from all the ingress ports to which it is attached.
• Aggregate policing works independently on each DFC-equipped switching module and independently on the PFC, which supports any non-DFC-equipped switching modules. Aggregate policing does not combine flow statistics from different DFC-equipped switching modules. You can display aggregate policing statistics for each DFC-equipped switching module and for the PFC and any non-DFC-equipped switching modules supported by the PFC.
• Each PFC or DFC polices independently, which might affect QoS features being applied to traffic that is distributed across the PFC and any DFCs. Examples of these QoS feature are:

– Policers applied to a port channel interface.

– Policers applied to a switched virtual interface.

– Egress policers applied to either a Layer 3 interface or an SVI. Note that PFC QoS performs egress policing decisions at the ingress interface, on the PFC or ingress DFC.

Policers affected by this restriction deliver an aggregate rate that is the sum of all the independent policing rates.

Microflow Policers

PFC QoS applies the bandwidth limit specified in a microflow policer separately to each flow in matched traffic. For example, if you configure a microflow policer to limit the TFTP traffic to 1 Mbps on VLAN 1 and VLAN 3, then 1 Mbps is allowed for each flow in VLAN 1 and 1 Mbps for each flow in VLAN 3. In other words, if there are three flows in VLAN 1 and four flows in VLAN 3, the microflow policer allows each of these flows 1 Mbps.

You can configure PFC QoS to apply the bandwidth limits in a microflow policer as follows:

• You can create microflow policers with up to 63 different rate and burst parameter combinations.
• You create microflow policers in a policy map class with the police flow command.
• You can configure a microflow policer to use only source addresses, which applies the microflow policer to all traffic from a source address regardless of the destination addresses.
• You can configure a microflow policer to use only destination addresses, which applies the microflow policer to all traffic to a destination address regardless of the source addresses.
• For MAC-Layer microflow policing, PFC QoS considers MAC-Layer traffic with the same protocol and the same source and destination MAC-Layer addresses to be part of the same flow, including traffic with different EtherTypes. You can configure MAC ACLs to filter IPX traffic.
• For IPX microflow policing, PFC QoS considers IPX traffic with the same source network, destination network, and destination node to be part of the same flow, including traffic with different source nodes or source sockets.
• By default, microflow policers only affect traffic routed by the MSFC. To enable microflow policing of other traffic, including traffic in bridge groups, enter the mls qos bridged command.
• You cannot apply microflow policing to ARP traffic.

You can include both an aggregate policer and a microflow policer in each policy map class to police a flow based on both its own bandwidth utilization and on its bandwidth utilization combined with that of other flows.

Note If traffic is both aggregate and microflow policed, then the aggregate and microflow policers must both be in the same policy-map class and each must use the same conform-action and exceed-action keyword option: drop, set-dscp-transmit, set-prec-transmit, or transmit.

For example, you could create a microflow policer with a bandwidth limit suitable for individuals in a group, and you could create a named aggregate policer with bandwidth limits suitable for the group as a whole. You could include both policers in policy map classes that match the group’s traffic. The combination would affect individual flows separately and the group aggregately.

For policy map classes that include both an aggregate and a microflow policer, PFC QoS responds to an out-of-profile status from either policer and, as specified by the policer, applies a new DSCP value or drops the packet. If both policers return an out-of-profile status, then if either policer specifies that the packet is to be dropped, it is dropped; otherwise, PFC QoS applies a marked-down DSCP value.

Note To avoid inconsistent results, ensure that all traffic policed by the same aggregate policer has the same trust state.

Policing uses the Layer 2 frame size. You specify the bandwidth utilization limit as a committed information rate (CIR). You can also specify a higher peak information rate (PIR). Packets that exceed a rate are “out of profile” or “nonconforming.”

In each policer, you specify if out-of-profile packets are to be dropped or to have a new DSCP value applied to them (applying a new DSCP value is called “markdown”). Because out-of-profile packets do not retain their original priority, they are not counted as part of the bandwidth consumed by in-profile packets.

If you configure a PIR, the PIR out-of-profile action cannot be less severe than the CIR out-of-profile action. For example, if the CIR out-of-profile action is to mark down the traffic, then the PIR out-of-profile action cannot be to transmit the traffic.

For all policers, PFC QoS uses a configurable global table that maps the internal DSCP value to a marked-down DSCP value. When markdown occurs, PFC QoS gets the marked-down DSCP value from the table. You cannot specify marked-down DSCP values in individual policers.

Note ● Policing with the conform-action transmit keywords supersedes the ingress LAN port trust state of matched traffic with trust DSCP or with the trust state defined by a trust policy-map class command.

• By default, the markdown table is configured so that no markdown occurs: the marked-down DSCP values are equal to the original DSCP values. To enable markdown, configure the table appropriately for your network.
• When you apply both ingress policing and egress policing to the same traffic, both the input policy and the output policy must either mark down traffic or drop traffic. PFC QoS does not support ingress markdown with egress drop or ingress drop with egress markdown.

Understanding Port-Based Queue Types

Port-based queue types are determined by the ASICs that control the ports. The following sections describe the queue types, drop thresholds, and buffers that are supported on the Cisco 7600 series router LAN modules:

• Ingress and Egress Buffers and Layer 2 CoS-Based Queues
• Ingress Queue Types
• Egress Queue Types
• Module to Queue Type Mappings

Ingress and Egress Buffers and Layer 2 CoS-Based Queues

The Ethernet LAN module port ASICs have buffers that are divided into a fixed number of queues. When congestion avoidance is enabled, PFC QoS uses the traffic’s Layer 2 CoS value to assign traffic to the queues. The buffers and queues store frames temporarily as they transit the switch. PFC QoS allocates the port ASIC memory as buffers for each queue on each port.

The Cisco 7600 series router LAN modules support the following types of queues:

• Standard queues
• Strict-priority queues

The Cisco 7600 series router LAN modules support the following types of scheduling algorithms between queues:

• Weighted Round Robin (WRR)—WRR does not explicitly reserve bandwidth for the queues. Instead, the amount of bandwidth assigned to each queue is user configurable. The percentage allocated to a queue defines the amount of bandwidth allocated to the queue.
• Deficit weighted round robin (DWRR)—In addition to the operation provided by WRR, DWRR keeps track of any low-priority queue under-transmission and compensates in the next round.
• Strict-priority queueing—Strict priority queueing allows delay-sensitive data such as voice to be dequeued and sent before packets in other queues are dequeued, giving delay-sensitive data preferential treatment over other traffic. The router services traffic in the strict-priority transmit queue before servicing the standard queues. After transmitting a packet from a standard queue, the switch checks for traffic in the strict-priority queue. If the switch detects traffic in the strict-priority queue, it suspends its service of the standard queue and completes service of all traffic in the strict-priority queue before returning to the standard queue.

The Cisco 7600 series router LAN modules provides congestion avoidance with these types of thresholds within a queue:

• Weighted Random Early Detection (WRED)—On ports with WRED drop thresholds, frames of a given CoS value are admitted to the queue based on a random probability designed to avoid buffer congestion. The probability of a frame with a given CoS being admitted to the queue or discarded depends on the weight and threshold assigned to that CoS value.

For example, if CoS 2 is assigned to queue 1, threshold 2, and the threshold 2 levels are 40 percent (low) and 80 percent (high), then frames with CoS 2 will not be dropped until queue 1 is at least 40 percent full. As the queue depth approaches 80 percent, frames with CoS 2 have an increasingly higher probability of being discarded rather than being admitted to the queue. Once the queue is over 80 percent full, all CoS 2 frames are dropped until the queue is less than 80 percent full. The frames the switch discards when the queue level is between the low and high thresholds are picked out at random, rather than on a per-flow basis or in a FIFO manner. This method works well with protocols such as TCP that can adjust to periodic packet drops by backing off and adjusting their transmission window size.

• Tail-drop thresholds—On ports with tail-drop thresholds, frames of a given CoS value are admitted to the queue until the drop threshold associated with that CoS value is exceeded; subsequent frames of that CoS value are discarded until the threshold is no longer exceeded. For example, if CoS 1 is assigned to queue 1, threshold 2, and the threshold 2 watermark is 60 percent, then frames with CoS 1 will not be dropped until queue 1 is 60 percent full. All subsequent CoS 1 frames will be dropped until the queue is less than 60 percent full. With some port types, you can configure the standard receive queue to use both a tail-drop and a WRED-drop threshold by mapping a CoS value to the queue or to the queue and a threshold. The switch uses the tail-drop threshold for traffic carrying CoS values mapped only to the queue. The switch uses WRED-drop thresholds for traffic carrying CoS values mapped to the queue and a threshold. All LAN ports of the same type use the same drop-threshold configuration.

Note In Release 12.2(33)SRC and later releases, you can enable DSCP-based queues and thresholds onWS-X6708-10GE ports. See Configuring DSCP-Based Queue Mapping.

The combination of multiple queues and the scheduling algorithms associated with each queue allows the switch to provide congestion avoidance.

Figure 48-9 illustrates the drop thresholds for a 1q4t ingress LAN port. Drop thresholds in other configurations function similarly.

Figure 48-9 Receive Queue Drop Thresholds

Ingress Queue Types

To see the queue structure of a LAN port, enter the show mls qos queuing interface { ethernet | fastethernet | gigabitethernet | tengigabitethernet } slot/port | include type command. The command displays one of the following architectures:

• 1q2t indicates one standard queue with one configurable tail-drop threshold and one nonconfigurable tail-drop threshold.
• 1q4t indicates one standard queue with four configurable tail-drop thresholds.
• 1q8t indicates one standard queue with eight configurable tail-drop thresholds.
• 2q8t indicates two standard queues, each with eight configurable tail-drop thresholds.
• 8q8t indicates eight standard queues, each with eight thresholds, each configurable as either WRED-drop or tail-drop.
• 1p1q4t indicates:

– One strict-priority queue

– One standard queue with four configurable tail-drop thresholds.

• 1p1q0t indicates:

– One strict-priority queue

– One standard queue with no configurable threshold (effectively a tail-drop threshold at 100 percent).

• 1p1q8t indicates the following:

– One strict-priority queue

– One standard queue with these thresholds:

—Eight thresholds, each configurable as either WRED-drop or tail-drop

—One non configurable (100 percent) tail-drop threshold

Egress Queue Types

To see the queue structure of an egress LAN port, enter the show mls qos queuing interface { ethernet | fastethernet | gigabitethernet | tengigabitethernet } slot/port | include type command.

The command displays one of the following architectures:

• 2q2t indicates two standard queues, each with two configurable tail-drop thresholds.
• 1p2q2t indicates the following:

– One strict-priority queue

– Two standard queues, each with two configurable WRED-drop thresholds

• 1p3q1t indicates the following:

– One strict-priority queue

– Three standard queues with these thresholds:

—One threshold configurable as either WRED-drop or tail-drop

—One nonconfigurable (100 percent) tail-drop threshold

• 1p2q1t indicates the following:

– One strict-priority queue

– Two standard queues with these thresholds:

—One WRED-drop threshold

—One non-configurable (100 percent) tail-drop threshold

• 1p3q8t indicates the following:

– One strict-priority queue

– Three standard queues, each with eight thresholds, each threshold configurable as either WRED-drop or tail-drop

• 1p7q8t indicates the following:

– One strict-priority queue

– Seven standard queues, each with eight thresholds, each threshold configurable as either WRED-drop or tail-drop

Module to Queue Type Mappings

The following tables show the module to queue structure mapping:

• Supervisor Engine Module QoS Queue Structures
• Ethernet and Fast Ethernet Module Queue Structures
• Gigabit and 10/100/1000 Ethernet Modules
• 10 Gigabit Ethernet Modules

PFC QoS Global Settings

The following global PFC QoS settings apply:

Default Values With PFC QoS Enabled

These sections list the default values that apply when PFC QoS is enabled:

• Receive-queue Size Percentages
• Transmit-Queue Size Percentages
• Bandwidth Allocation Ratios
• Default Drop-Threshold Percentages and CoS Value Mappings

Note The ingress LAN port trust state defaults to untrusted with QoS enabled.

Receive-queue Size Percentages

Transmit-Queue Size Percentages

Bandwidth Allocation Ratios

Default Drop-Threshold Percentages and CoS Value Mappings

The following tables list the default drop-thresholds values and CoS mappings for different queue types:

• 1q4t Receive Queues
• 1p1q4t Receive Queues
• 1p1q0t Receive Queues
• 1p1q8t Receive Queues
• 1q8t Receive Queues
• 2q8t Receive Queues
• 8q8t Receive Queues
• 2q2t Transmit Queues
• 1p2q2t Transmit Queues
• 1p3q8t Transmit Queues
• 1p7q8t Transmit Queues
• 1p3q1t Transmit Queues
• 1p2q1t Transmit Queues

1q2t Receive Queues

The following table lists the default values when you configure mls qos and mls qos trust cos.

1q4t Receive Queues

The following table lists the default values when you configure mls qos and mls qos trust cos.

1p1q4t Receive Queues

1p1q0t Receive Queues

1p1q8t Receive Queues

1q8t Receive Queues

The following table lists the default values when you configure mls qos and mls qos trust cos.

2q8t Receive Queues

The following table lists the default values when you configure mls qos and mls qos trust cos.

8q8t Receive Queues

The following table lists the default values when you configure mls qos and mls qos trust cos or mls qos trust dscp.

2q2t Transmit Queues

1p2q2t Transmit Queues

1p3q8t Transmit Queues

1p7q8t Transmit Queues

The following table lists the default values when you configure mls qos and mls qos trust cos or mls qos trust dscp.

1p3q1t Transmit Queues

1p2q1t Transmit Queues

Default Values With PFC QoS Disabled

Note If you enable only mls qos and do not configure any trusts, the default values for the queues 1p1q4t, 1q2t, 1q4t, 1q8t, 2q8t,8q8t or 1p7q8t are equal to the values when PFC QoS is disabled.

General Guidelines

• The match ip precedence and match ip dscp commands filter only IPv4 traffic.
• The match precedence and match dscp commands filter IPv4 and IPv6 traffic.
• The set ip dscp and set ip precedence commands are saved in the configuration file as set dscp and set precedence commands.
• PFC QoS supports the set dscp and set precedence policy map class commands for IPv4 and IPv6 traffic.
• The flowmask requirements of QoS, NetFlow, and NetFlow data export (NDE) might conflict, especially if you configure microflow policing.
• With egress ACL support for remarked DSCP and VACL capture both configured on an interface, VACL capture might capture two copies of each packet, and the second copy might be corrupt.
• You cannot configure egress ACL support for remarked DSCP on tunnel interfaces.
• Egress ACL support for remarked DSCP supports IP unicast traffic.
• Egress ACL support for remarked DSCP is not relevant to multicast traffic. PFC QoS applies ingress QoS changes to multicast traffic before applying egress QoS.
• NetFlow and NetFlow data export (NDE) do not support interfaces where egress ACL support for remarked DSCP is configured.
• When egress ACL support for remarked DSCP is configured on any interface, you must configure an interface-specific flowmask to enable NetFlow and NDE support on interfaces where egress ACL support for remarked DSCP is not configured. Enter either the mls flow ip interface-destination-source or the mls flow ip interface-full global configuration mode command.
• Interface counters are not accurate on interfaces where egress ACL support for remarked DSCP is configured.
• You cannot apply microflow policing to traffic that has been permitted by egress ACL support for remarked DSCP.
• Traffic that has been permitted by egress ACL support for remarked DSCP cannot be tagged as MPLS traffic. (The traffic can be tagged as MPLS traffic on another network device.)
• When you apply both ingress policing and egress policing to the same traffic, both the input policy and the output policy must either mark down traffic or drop traffic. PFC QoS does not support ingress markdown with egress drop or ingress drop with egress markdown. (CSCea23571)
• If traffic is both aggregate and microflow policed, then the aggregate and microflow policers must both be in the same policy-map class and each must use the same conform-action and exceed-action keyword option: drop, set-dscp-transmit, set-prec-transmit, or transmit.
• You cannot configure PFC QoS features on tunnel interfaces.
• PFC QoS does not rewrite the payload ToS byte in tunnel traffic.
• PFC QoS filters only by ACLs, dscp values, or IP precedence values.
• For these commands, PFC QoS applies identical configuration to all LAN ports controlled by the same application-specific integrated circuit (ASIC):

– rcv-queue cos-map

– wrr-queue cos-map

• Except for WS-X6716-10GE, WS-X6708-10GE, WS-X6704-10GE, WS-X6748-SFP, WS-X6724-SFP, and WS-X6748-GE-TX modules, PFC QoS applies identical configuration to all the LAN ports controlled by the same application-specific integrated circuit (ASIC) for these commands:

– rcv-queue random-detect

– rcv-queue queue-limit

– wrr-queue queue-limit

– wrr-queue bandwidth

– priority-queue cos-map

– wrr-queue threshold

– rcv-queue threshold

– wrr-queue random-detect

– wrr-queue random-detect min-threshold

– wrr-queue random-detect max-threshold

• Configure these commands only on physical ports. Do not configure these commands on logical interfaces:

– priority-queue cos-map

– wrr-queue cos-map

– wrr-queue random-detect

– wrr-queue random-detect max-threshold

– wrr-queue random-detect min-threshold

– wrr-queue threshold

– wrr-queue queue-limit

– wrr-queue bandwidth

– rcv-queue cos-map

– rcv-queue bandwidth

– rcv-queue random-detect

– rcv-queue random-detect max-threshold

– rcv-queue random-detect min-threshold

– rcv-queue queue-limit

– rcv-queue cos-map

– rcv-queue threshold

PFC Guidelines

• All versions of the PFC support QoS for IPv6 unicast and multicast traffic.
• To display information about IPv6 PFC QoS, enter the show mls qos ipv6 command.
• The QoS features implemented in the port ASICs (queue architecture and dequeuing algorithms) support IPv4 and IPv6 traffic.
• The PFC supports IPv6 named extended ACLs and named standard ACLs.
• The PFC supports the match protocol ipv6 command.
• Because of conflicting TCAM lookup flow key bit requirements, you cannot configure IPv6 DSCP-based filtering and IPv6 Layer 4 range-based filtering on the same interface. For example:

– If you configure both a DSCP value and a Layer 4 “greater than” (gt) or “less than” (lt) operator in an IPv6 ACE, you cannot use the ACL for PFC QoS filtering.

– If you configure a DSCP value in one IPv6 ACL and a Layer 4 “greater than” (gt) or “less than” (lt) operator in another IPv6 ACL, you cannot use both ACLs in different class maps on the same interface for PFC QoS filtering.

• You can apply aggregate and microflow policers to IPv6 traffic.
• With egress ACL support for remarked DSCP configured, the PFC does not provide hardware-assistance for these features:

– Cisco IOS reflexive ACLs

– TCP intercept

– Context-Based Access Control (CBAC)

– Network Address Translation (NAT)

• You cannot apply microflow policing to ARP traffic.
• The PFC does not apply egress policing to traffic that is being bridged to the MSFC.
• The PFC does not apply egress policing or egress DSCP mutation to multicast traffic from the MSFC.
• PFC QoS does not rewrite the ToS byte in bridged multicast traffic.

Class Map Command Restrictions

• PFC QoS supports the match any class map command.
• PFC QoS supports class maps that contain a single match command.
• PFC QoS does not support these class map commands:

– match classmap

– match destination-address

– match input-interface

– match qos-group

– match source-address

Policy Map Command Restrictions

PFC QoS does not support these policy map commands:

• class class_name destination-address
• class class_name input-interface
• class class_name protocol
• class class_name qos-group
• class class_name source-address

Policy Map Class Command Restrictions

PFC QoS does not support these policy map class commands:

• bandwidth
• priority
• queue-limit
• random-detect
• set qos-group
• service-policy

Supported Granularity for CIR and PIR Rate Values

PFC QoS has the following hardware granularity for CIR and PIR rate values:

Within each range, PFC QoS programs the PFC with rate values that are multiples of the granularity values.

Supported Granularity for CIR and PIR Token Bucket Sizes

PFC QoS has the following hardware granularity for CIR and PIR token bucket (burst) sizes:

Within each range, PFC QoS programs the PFC with token bucket sizes that are multiples of the granularity values.

IP Precedence and DSCP Values

Enabling PFC QoS Globally

To enable PFC QoS globally, perform this task:

This example shows how to enable PFC QoS globally:

This example shows how to verify the configuration:

Configuring DSCP Transparency

To enable DSCP transparency, which preserves the received Layer 3 ToS byte, perform this task:

When you preserve the received Layer 3 ToS byte, QoS uses the marked or marked-down CoS value for egress queueing and in egress tagged traffic.

This example shows how to preserve the received Layer 3 ToS byte:

Configuring Trust State

To enable or disable the trust state over the internal recirculate path, use the mls qos recirc untrust command in the global configuration mode.

When you preserve the received Layer 3 ToS byte, QoS uses the marked or marked-down CoS value for egress queueing and in egress tagged traffic.

This example shows how to preserve the received Layer 3 ToS byte:

Enabling Queueing-Only Mode

To enable queueing-only mode on the router, perform this task:

When you enable queueing-only mode, the router does the following:

• Disables marking and policing globally
• Configures all ports to trust Layer 2 CoS

Note The router applies the port CoS value to untagged ingress traffic and to traffic that is received through ports that cannot be configured to trust CoS.

This example shows how to enable queueing-only mode:

Enabling Microflow Policing of Bridged Traffic

By default, microflow policers affect only routed traffic. To enable microflow policing of bridged traffic on specified VLANs, perform this task:

This example shows how to enable microflow policing of bridged traffic on VLANs 3 through 5:

This example shows how to verify the configuration:

Enabling VLAN-Based PFC QoS on Layer 2 LAN Ports

Note ● PFC QoS supports VLAN-based QoS with DFC3s installed.

• You can attach policy maps to Layer 3 interfaces for application of PFC QoS to egress traffic. VLAN-based or port-based PFC QoS on Layer 2 ports is not relevant to application of PFC QoS to egress traffic on Layer 3 interfaces.

By default, PFC QoS uses policy maps attached to LAN ports. For ports configured as Layer 2 LAN ports with the switchport keyword, you can configure PFC QoS to use policy maps attached to a VLAN. Ports not configured with the switchport keyword are not associated with a VLAN.

To enable VLAN-based PFC QoS on a Layer 2 LAN port, perform this task:

This example shows how to enable VLAN-based PFC QoS on Fast Ethernet port 5/42:

This example shows how to verify the configuration:

Note Configuring a Layer 2 LAN port for VLAN-based PFC QoS preserves the policy map port configuration. The no mls qos vlan-based port command reenables any previously configured port commands.

Enabling Egress ACL Support for Remarked DSCP

To enable egress ACL support for remarked DSCP on an ingress interface, perform this task:

When configuring egress ACL support for remarked DSCP on an ingress interface, note the following information:

• To enable egress ACL support for remarked DSCP only for the traffic filtered by a specific standard, extended named, or extended numbered IP ACL, enter the IP ACL name or number.
• If you do not enter an IP ACL name or number, egress ACL support for remarked DSCP is enabled for all IP ingress IP traffic on the interface.

This example shows how to enable egress ACL support for remarked DSCP on Fast Ethernet port 5/36:

Creating Named Aggregate Policers

To create a named aggregate policer, perform this task:

When creating a named aggregate policer, note the following information:

• Aggregate policing works independently on each DFC-equipped switching module and independently on the PFC, which supports any non-DFC-equipped switching modules. Aggregate policing does not combine flow statistics from different DFC-equipped switching modules. You can display aggregate policing statistics for each DFC-equipped switching module and for the PFC and any non-DFC-equipped switching modules supported by the PFC.
• Each PFC or DFC polices independently, which might affect QoS features being applied to traffic that is distributed across the PFC and any DFCs. Examples of these QoS feature are:

– Policers applied to a port channel interface.

– Policers applied to a switched virtual interface.

– Egress policers applied to either a Layer 3 interface or an SVI. Note that PFC QoS performs egress policing decisions at the ingress interface, on the PFC or ingress DFC.

Policers affected by this restriction deliver an aggregate rate that is the sum of all the independent policing rates.

• You can apply aggregate policers to IPv6 traffic.
• Policing uses the Layer 2 frame size.
• See the “PFC QoS Configuration Guidelines and Restrictions” section for information about rate and burst size granularity.
• The valid range of values for the CIR bits_per_second parameter is as follows:

– Minimum—32 kilobits per second, entered as 32000

– Maximum—10 gigabits per second, entered as 10000000000

• The normal_burst_bytes parameter sets the CIR token bucket size.
• The maximum_burst_bytes parameter sets the PIR token bucket size.
• When configuring the size of a token bucket, note the following information:

– The minimum token bucket size is 1 kilobyte, entered as 1000 (the maximum_burst_bytes parameter must be set larger than the normal_burst_bytes parameter).

– The maximum token bucket size is 32 megabytes, entered as 32000000.

– To sustain a specific rate, set the token bucket size to be at least the rate value divided by 4000 because tokens are removed from the bucket every 1/4000th of a second (0.25 ms).

– Because the token bucket must be large enough to hold at least one frame, set the parameter larger than the maximum size of the traffic being policed.

– For TCP traffic, configure the token bucket size as a multiple of the TCP window size, with a minimum value at least twice as large as the maximum size of the traffic being policed.

• The valid range of values for the pir bits_per_second parameter is as follows:

– Minimum—32 kilobits per second, entered as 32000 (the value cannot be smaller than the CIR bits_per_second parameters)

– Maximum—10 gigabits per second, entered as 10000000000

• (Optional) You can specify a conform action for matched in-profile traffic as follows:

– The default conform action is transmit, which sets the policy map class trust state to trust DSCP unless the policy map class contains a trust command.

– To set PFC QoS labels in untrusted traffic, enter the set-dscp-transmit keyword to mark matched untrusted traffic with a new DSCP value or enter the set-prec-transmit keyword to mark matched untrusted traffic with a new IP precedence value. The set-dscp-transmit and set-prec-transmit keywords are only supported for IP traffic. PFC QoS sets egress ToS and CoS from the configured value.

– Enter the drop keyword to drop all matched traffic.

Note When you configure drop as the conform action, PFC QoS configures drop as the exceed action and the violate action.

• (Optional) For traffic that exceeds the CIR, you can specify an exceed action as follows:

– The default exceed action is drop, except with a maximum_burst_bytes parameter ( drop is not supported with a maximum_burst_bytes parameter).

Note When the exceed action is drop, PFC QoS ignores any configured violate action.

– Enter the policed-dscp-transmit keyword to cause all matched out-of-profile traffic to be marked down as specified in the markdown map.

Note When you create a policer that does not use the pir keyword and the maximum_burst_bytes parameter is equal to the normal_burst_bytes parameter (which is the case if you do not enter the maximum_burst_bytes parameter), the exceed-action policed-dscp-transmit keywords cause PFC QoS to mark traffic down as defined by the policed-dscp max-burst markdown map.

• (Optional) For traffic that exceeds the PIR, you can specify a violate action as follows:

– To mark traffic without policing, enter the transmit keyword to transmit all matched out-of-profile traffic.

– The default violate action is equal to the exceed action.

– Enter the policed-dscp-transmit keyword to cause all matched out-of-profile traffic to be marked down as specified in the markdown map.

– For marking without policing, enter the transmit keyword to transmit all matched out-of-profile traffic.

Note When you apply both ingress policing and egress policing to the same traffic, both the input policy and the output policy must either mark down traffic or drop traffic. PFC QoS does not support ingress markdown with egress drop or ingress drop with egress markdown.

This example shows how to create a named aggregate policer with a 1-Mbps rate limit and a 10-MB burst size that transmits conforming traffic and marks down out-of-profile traffic:

This example shows how to verify the configuration:

The output displays the following:

• The AgId parameter displays the hardware policer ID.
• The policy maps that use the policer are listed in the square brackets ([]).

Configuring a PFC QoS Policy

These sections describe PFC QoS policy configuration:

• PFC QoS Policy Configuration Overview
• Configuring MAC ACLs
• Configuring ARP ACLs for QoS Filtering
• Configuring a Class Map
• Verifying Class Map Configuration
• Configuring a Policy Map
• Verifying Policy Map Configuration
• Attaching a Policy Map to an Interface

Note PFC QoS policies process both unicast and multicast traffic.

PFC QoS Policy Configuration Overview

Note To mark traffic without limiting bandwidth utilization, create a policer that uses the transmit keywords for both conforming and nonconforming traffic.

These commands configure traffic classes and the policies to be applied to those traffic classes and attach the policies to ports:

• access-list (Optional for IP traffic. You can filter IP traffic with class-map commands.):

– PFC QoS supports these ACL types:

– The PFC supports IPv6 named extended ACLs and named standard ACLs.

– The PFC supports ARP ACLs.

Note —The PFC does not apply IP ACLs to ARP traffic.

—You cannot apply microflow policing to ARP traffic.

– The PFC does not support IPX ACLs. With a PFC, you can configure MAC ACLs to filter IPX traffic.

– PFC QoS supports time-based Cisco IOS ACLs.

– Except for MAC ACLs and ARP ACLs, refer to the Cisco IOS Security Configuration Guide, Release 12.2, “Traffic Filtering and Firewalls,” at this URL:

http://www.cisco.com/en/US/docs/ios/12_2/security/configuration/guide/scfacls.html

– See “Configuring Network Security,” for additional information about ACLs on the Cisco 7600 series routers.

• class-map (optional)—Enter the class-map command to define one or more traffic classes by specifying the criteria by which traffic is classified.
• policy-map —Enter the policy-map command to define the following: