CLI Book 2: Cisco ASA Series Firewall CLI Configuration Guide, 9.2

Supported QoS Features

The ASA supports the following QoS features:

• Policing—To prevent individual flows from hogging the network bandwidth, you can limit the maximum bandwidth used per flow. See Information About Policing for more information.
• Priority queuing—For critical traffic that cannot tolerate latency, such as Voice over IP (VoIP), you can identify traffic for Low Latency Queuing (LLQ) so that it is always transmitted ahead of other traffic. See Information About Priority Queuing for more information.
• Traffic shaping—If you have a device that transmits packets at a high speed, such as a ASA with Fast Ethernet, and it is connected to a low speed device such as a cable modem, then the cable modem is a bottleneck at which packets are frequently dropped. To manage networks with differing line speeds, you can configure the ASA to transmit packets at a fixed slower rate. See Information About Traffic Shaping for more information.

What is a Token Bucket?

A token bucket is used to manage a device that regulates the data in a flow. For example, the regulator might be a traffic policer or a traffic shaper. A token bucket itself has no discard or priority policy. Rather, a token bucket discards tokens and leaves to the flow the problem of managing its transmission queue if the flow overdrives the regulator.

A token bucket is a formal definition of a rate of transfer. It has three components: a burst size, an average rate, and a time interval. Although the average rate is generally represented as bits per second, any two values may be derived from the third by the relation shown as follows:

average rate = burst size / time interval

Here are some definitions of these terms:

• Average rate—Also called the committed information rate (CIR), it specifies how much data can be sent or forwarded per unit time on average.
• Burst size—Also called the Committed Burst (Bc) size, it specifies in bits or bytes per burst how much traffic can be sent within a given unit of time to not create scheduling concerns. (For traffic shaping, it specifies bits per burst; for policing, it specifies bytes per burst.)
• Time interval—Also called the measurement interval, it specifies the time quantum in seconds per burst.

In the token bucket metaphor, tokens are put into the bucket at a certain rate. The bucket itself has a specified capacity. If the bucket fills to capacity, newly arriving tokens are discarded. Each token is permission for the source to send a certain number of bits into the network. To send a packet, the regulator must remove from the bucket a number of tokens equal in representation to the packet size.

If not enough tokens are in the bucket to send a packet, the packet either waits until the bucket has enough tokens (in the case of traffic shaping) or the packet is discarded or marked down (in the case of policing). If the bucket is already full of tokens, incoming tokens overflow and are not available to future packets. Thus, at any time, the largest burst a source can send into the network is roughly proportional to the size of the bucket.

Note that the token bucket mechanism used for traffic shaping has both a token bucket and a data buffer, or queue; if it did not have a data buffer, it would be a policer. For traffic shaping, packets that arrive that cannot be sent immediately are delayed in the data buffer.

For traffic shaping, a token bucket permits burstiness but bounds it. It guarantees that the burstiness is bounded so that the flow will never send faster than the token bucket capacity, divided by the time interval, plus the established rate at which tokens are placed in the token bucket. See the following formula:

(token bucket capacity in bits / time interval in seconds) + established rate in bps = maximum flow speed in bps

This method of bounding burstiness also guarantees that the long-term transmission rate will not exceed the established rate at which tokens are placed in the bucket.

Information About Policing

Policing is a way of ensuring that no traffic exceeds the maximum rate (in bits/second) that you configure, thus ensuring that no one traffic flow or class can take over the entire resource. When traffic exceeds the maximum rate, the ASA drops the excess traffic. Policing also sets the largest single burst of traffic allowed.

Information About Priority Queuing

LLQ priority queuing lets you prioritize certain traffic flows (such as latency-sensitive traffic like voice and video) ahead of other traffic.

The ASA supports two types of priority queuing:

• Standard priority queuing—Standard priority queuing uses an LLQ priority queue on an interface (see Configuring the Standard Priority Queue for an Interface), while all other traffic goes into the “best effort” queue. Because queues are not of infinite size, they can fill and overflow. When a queue is full, any additional packets cannot get into the queue and are dropped. This is called tail drop . To avoid having the queue fill up, you can increase the queue buffer size. You can also fine-tune the maximum number of packets allowed into the transmit queue. These options let you control the latency and robustness of the priority queuing. Packets in the LLQ queue are always transmitted before packets in the best effort queue.
• Hierarchical priority queuing—Hierarchical priority queuing is used on interfaces on which you enable a traffic shaping queue. A subset of the shaped traffic can be prioritized. The standard priority queue is not used. See the following guidelines about hierarchical priority queuing:

– Priority packets are always queued at the head of the shape queue so they are always transmitted ahead of other non-priority queued packets.

– Priority packets are never dropped from the shape queue unless the sustained rate of priority traffic exceeds the shape rate.

– For IPsec-encrypted packets, you can only match traffic based on the DSCP or precedence setting.

– IPsec-over-TCP is not supported for priority traffic classification.

Information About Traffic Shaping

Traffic shaping is used to match device and link speeds, thereby controlling packet loss, variable delay, and link saturation, which can cause jitter and delay.

Note Traffic shaping is only supported on the ASA 5505.

• Traffic shaping must be applied to all outgoing traffic on a physical interface or in the case of the ASA 5505, on a VLAN. You cannot configure traffic shaping for specific types of traffic.
• Traffic shaping is implemented when packets are ready to be transmitted on an interface, so the rate calculation is performed based on the actual size of a packet to be transmitted, including all the possible overhead such as the IPsec header and L2 header.
• The shaped traffic includes both through-the-box and from-the-box traffic.
• The shape rate calculation is based on the standard token bucket algorithm. The token bucket size is twice the Burst Size value. See What is a Token Bucket?.
• When bursty traffic exceeds the specified shape rate, packets are queued and transmitted later. Following are some characteristics regarding the shape queue (for information about hierarchical priority queuing, see Information About Priority Queuing):

– The queue size is calculated based on the shape rate. The queue can hold the equivalent of 200-milliseconds worth of shape rate traffic, assuming a 1500-byte packet. The minimum queue size is 64.

– When the queue limit is reached, packets are tail-dropped.

– Certain critical keep-alive packets such as OSPF Hello packets are never dropped.

– The time interval is derived by time_interval = burst_size / average_rate. The larger the time interval is, the burstier the shaped traffic might be, and the longer the link might be idle. The effect can be best understood using the following exaggerated example:

Average Rate = 1000000

Burst Size = 1000000

In the above example, the time interval is 1 second, which means, 1 Mbps of traffic can be bursted out within the first 10 milliseconds of the 1-second interval on a 100 Mbps FE link and leave the remaining 990 milliseconds idle without being able to send any packets until the next time interval. So if there is delay-sensitive traffic such as voice traffic, the Burst Size should be reduced compared to the average rate so the time interval is reduced.

How QoS Features Interact

You can configure each of the QoS features alone if desired for the ASA. Often, though, you configure multiple QoS features on the ASA so you can prioritize some traffic, for example, and prevent other traffic from causing bandwidth problems.

See the following supported feature combinations per interface:

• Standard priority queuing (for specific traffic) + Policing (for the rest of the traffic).

You cannot configure priority queuing and policing for the same set of traffic.

• Traffic shaping (for all traffic on an interface) + Hierarchical priority queuing (for a subset of traffic).

You cannot configure traffic shaping and standard priority queuing for the same interface; only hierarchical priority queuing is allowed. For example, if you configure standard priority queuing for the global policy, and then configure traffic shaping for a specific interface, the feature you configured last is rejected because the global policy overlaps the interface policy.

Typically, if you enable traffic shaping, you do not also enable policing for the same traffic, although the ASA does not restrict you from configuring this.

DSCP and DiffServ Preservation

• DSCP markings are preserved on all traffic passing through the ASA.
• The ASA does not locally mark/remark any classified traffic, but it honors the Expedited Forwarding (EF) DSCP bits of every packet to determine if it requires “priority” handling and will direct those packets to the LLQ.
• DiffServ marking is preserved on packets when they traverse the service provider backbone so that QoS can be applied in transit (QoS tunnel pre-classification).

Determining the Queue and TX Ring Limits for a Standard Priority Queue

To determine the priority queue and TX ring limits, use the worksheets below.

Table 19-1 shows how to calculate the priority queue size. Because queues are not of infinite size, they can fill and overflow. When a queue is full, any additional packets cannot get into the queue and are dropped (called tail drop). To avoid having the queue fill up, you can adjust the queue buffer size according to the Configuring the Standard Priority Queue for an Interface.

Table 19-2 shows how to calculate the TX ring limit. This limit determines the maximum number of packets allowed into the Ethernet transmit driver before the driver pushes back to the queues on the interface to let them buffer packets until the congestion clears. This setting guarantees that the hardware-based transmit ring imposes a limited amount of extra latency for a high-priority packet.

Configuring the Standard Priority Queue for an Interface

If you enable standard priority queuing for traffic on a physical interface, then you need to also create the priority queue on each interface. Each physical interface uses two queues: one for priority traffic, and the other for all other traffic. For the other traffic, you can optionally configure policing.

Note The standard priority queue is not required for hierarchical priority queuing with traffic shaping; see Information About Priority Queuing for more information.

Restrictions

• (ASASM) The ASASM does not support priority queuing.
• (ASA 5512-X through ASA 5555-X) Priority queuing is not supported on the Management 0/0 interface.

Detailed Steps

Examples

The following example establishes a priority queue on interface “outside” (the GigabitEthernet0/1 interface), with the default queue-limit and tx-ring-limit:

The following example establishes a priority queue on the interface “outside” (the GigabitEthernet0/1 interface), sets the queue-limit to 260 packets, and sets the tx-ring-limit to 3:

Configuring a Service Rule for Standard Priority Queuing and Policing

You can configure standard priority queuing and policing for different class maps within the same policy map. See How QoS Features Interact for information about valid QoS configurations.

To create a policy map, perform the following steps.

Restrictions

• You cannot use the class-default class map for priority traffic.
• You cannot configure traffic shaping and standard priority queuing for the same interface; only hierarchical priority queuing is allowed.
• (ASASM) The ASASM only supports policing.
• For policing, to-the-box traffic is not supported.
• For policing, traffic to and from a VPN tunnel bypass interface is not supported.
• For policing, when you match a tunnel group class map, only outbound policing is supported.

Guidelines

• For priority traffic, identify only latency-sensitive traffic.
• For policing traffic, you can choose to police all other traffic, or you can limit the traffic to certain types.

Detailed Steps

Examples

Example 19-1 Class Map Examples for VPN Traffic

In the following example, the class-map command classifies all non-tunneled TCP traffic, using an ACL named tcp_traffic:

In the following example, other, more specific match criteria are used for classifying traffic for specific, security-related tunnel groups. These specific match criteria stipulate that a match on tunnel-group (in this case, the previously-defined Tunnel-Group-1) is required as the first match characteristic to classify traffic for a specific tunnel, and it allows for an additional match line to classify the traffic (IP differential services code point, expedited forwarding).

In the following example, the class-map command classifies both tunneled and non-tunneled traffic according to the traffic type:

The following example shows a way of policing a flow within a tunnel, provided the classed traffic is not specified as a tunnel, but does go through the tunnel. In this example, 192.168.10.10 is the address of the host machine on the private side of the remote tunnel, and the ACL is named “host-over-l2l”. By creating a class-map (named “host-specific”), you can then police the “host-specific” class before the LAN-to-LAN connection polices the tunnel. In this example, the “host-specific” traffic is rate-limited before the tunnel, then the tunnel is rate-limited:

The following example builds on the configuration developed in the previous section. As in the previous example, there are two named class-maps: tcp_traffic and TG1-voice.

Adding a third class map provides a basis for defining a tunneled and non-tunneled QoS policy, as follows, which creates a simple QoS policy for tunneled and non-tunneled traffic, assigning packets of the class TG1-voice to the low latency queue and setting rate limits on the tcp_traffic and TG1-best-effort traffic flows.

Example 19-2 Priority and Policing Example

In this example, the maximum rate for traffic of the tcp_traffic class is 56,000 bits/second and a maximum burst size of 10,500 bytes per second. For the TC1-BestEffort class, the maximum rate is 200,000 bits/second, with a maximum burst of 37,500 bytes/second. Traffic in the TC1-voice class has no policed maximum speed or burst rate because it belongs to a priority class.

Configuring a Service Rule for Traffic Shaping and Hierarchical Priority Queuing

You can configure traffic shaping for all traffic on an interface, and optionally hierarchical priority queuing for a subset of latency-sensitive traffic.

This section includes the following topics:

• (Optional) Configuring the Hierarchical Priority Queuing Policy
• Configuring the Service Rule

(Optional) Configuring the Hierarchical Priority Queuing Policy

You can optionally configure priority queuing for a subset of latency-sensitive traffic.

Guidelines

• One side-effect of priority queuing is packet re-ordering. For IPsec packets, out-of-order packets that are not within the anti-replay window generate warning syslog messages. These warnings are false alarms in the case of priority queuing. You can configure the IPsec anti-replay window size to avoid possible false alarms. See the crypto ipsec security-association replay command in the command reference.
• For hierarchical priority queuing, you do not need to create a priority queue on an interface.

Restrictions

• For hierarchical priority queuing, for encrypted VPN traffic, you can only match traffic based on the DSCP or precedence setting; you cannot match a tunnel group.
• For hierarchical priority queuing, IPsec-over-TCP traffic is not supported.

Detailed Steps

Configuring the Service Rule

To configure traffic shaping and optional hiearchical priority queuing, perform the following steps.

Restrictions

• Traffic shaping is only supported on the ASA 5505. Multi-core models (such as the ASA 5500-X) do not support shaping.
• For traffic shaping, you can only use the class-default class map, which is automatically created by the ASA, and which matches all traffic.
• You cannot configure traffic shaping and standard priority queuing for the same interface; only hierarchical priority queuing is allowed. See How QoS Features Interact for information about valid QoS configurations.
• You cannot configure traffic shaping in the global policy.

Detailed Steps

Examples

The following example enables traffic shaping on the outside interface, and limits traffic to 2 Mbps; priority queuing is enabled for VoIP traffic that is tagged with DSCP EF and AF13 and for IKE traffic:

Viewing QoS Police Statistics

To view the QoS statistics for traffic policing, use the show service-policy command with the police keyword:

The following is sample output for the show service-policy police command:

Viewing QoS Standard Priority Statistics

To view statistics for service policies implementing the priority command, use the show service-policy command with the priority keyword:

The following is sample output for the show service-policy priority command:

Note “Aggregate drop” denotes the aggregated drop in this interface; “aggregate transmit” denotes the aggregated number of transmitted packets in this interface.

Viewing QoS Shaping Statistics

To view statistics for service policies implementing the shape command, use the show service-policy command with the shape keyword:

The following is sample output for the show service-policy shape command:

The following is sample output of the show service policy shape command, which includes service policies that include the shape command and the service-policy command that calls the hierarchical priority policy and the related statistics:

Viewing QoS Standard Priority Queue Statistics

To display the priority-queue statistics for an interface, use the show priority-queue statistics command in privileged EXEC mode. The results show the statistics for both the best-effort (BE) queue and the low-latency queue (LLQ). The following example shows the use of the show priority-queue statistics command for the interface named test, and the command output.

In this statistical report, the meaning of the line items is as follows:

• “Packets Dropped” denotes the overall number of packets that have been dropped in this queue.
• “Packets Transmit” denotes the overall number of packets that have been transmitted in this queue.
• “Packets Enqueued” denotes the overall number of packets that have been queued in this queue.
• “Current Q Length” denotes the current depth of this queue.
• “Max Q Length” denotes the maximum depth that ever occurred in this queue.