JUNOS CoS Components

JUNOS Software supports CoS on J Series and SRX Series devices as indicated in the following topics:

• Code-Point Aliases
• Classifiers
• Forwarding Classes
• Loss Priorities
• Forwarding Policy Options
• Transmission Queues
• Schedulers
• Virtual Channels
• Policers for Traffic Classes
• Rewrite Rules

Code-Point Aliases

A code-point alias assigns a name to a pattern of code-point bits. You can use this name, instead of the bit pattern, when you configure other CoS components such as classifiers, drop-profile maps, and rewrite rules.

Classifiers

Packet classification refers to the examination of an incoming packet. This function associates the packet with a particular CoS servicing level. Two general types of classifiers are supported—behavior aggregate (BA) classifiers and multifield (MF) classifiers. When both BA and and MF classifications are performed on a packet, the MF classification has higher precedence.

In JUNOS Software, classifiers associate incoming packets with a forwarding class (FC) and packet loss priority (PLP) and, based on the associated forwarding class, assign packets to output queues. FC and PLP associated with a packet specify the behavior of a hop, within the system, to process the packet. The per hop behavior (PHB) comprises packet forwarding, policing, scheduling, shaping, and marking. For example, a hop can put a packet in one of the priority queues according to its FC and then manage the queues by checking a packet's PLP. JUNOS Software supports up to eight FCs and four PLPs.

Behavior Aggregate Classifiers

A behavior aggregate (BA) classifier operates on a packet as it enters the device. Using behavior aggregate classifiers, the device aggregates different types of traffic into a single forwarding class to receive the same forwarding treatment. The CoS value in the packet header is the single field that determines the CoS settings applied to the packet. Behavior aggregate classifiers allow you to set the forwarding class and loss priority of a packet based on the Differentiated Services (DiffServ) code point (DSCP) value, DSCP IPv4 value, DSCP IPv6 value, IP precedence value, MPLS EXP bits, or IEEE 802.1p value. The default classifier is based on the IP precedence value. For more information, see Default Behavior Aggregate Classifiers.

JUNOS Software performs BA classification for a packet by examining its layer 2, layer 3, and CoS-related parameters as shown in Table 4.

Note: A BA classifier evaluates Layer 2 and Layer 3 parameters independently; the results that generate from Layer 2 parameters override the results that generate from the Layer 3 parameters.

Default IP Precedence Classifier

With JUNOS Software, all logical interface are automatically assigned a default IP precedence classifier when the logical interface is configured. This default traffic classifier maps IP precedence values to a forwarding class and packet loss priority as shown in Table 5. These mapping results take effect for an ingress packet until it is further processed by another classification method.

Multifield Classifiers

A multifield (MF) classifier is a second method for classifying traffic flows. Unlike the behavior aggregate classifier, a multifield classifier can examine multiple fields in the packet—for example, the source and destination address of the packet or the source and destination port numbers of the packet. With multifield classifiers, you set the forwarding class and loss priority of a packet based on firewall filter rules.

Note: For a specified interface, you can configure both an MF classifier and a BA classifier without conflicts. Because the classifiers are always applied in sequential order, the BA classifier followed by the MF classifier, any BA classification result is overridden by an MF classifier, if they conflict.

JUNOS Software performs MF traffic classification by directly scrutinizing multiple fields of a packet to classify a packet without having to rely upon the output of the previous BA traffic classification. JUNOS Software can simultaneously check a packet's data ranging from layer 2 to layer 7 as shown in Table 6

Using JUNOS Software, you configure an MF classifier with a firewall filter and its associated match conditions. This enables you to use any filter match criteria to locate packets that require classification. For more information on firewall filters and policies, see the JUNOS Software Security Configuration Guide.

Forwarding Classes

Forwarding classes allow you to group packets for transmission. Based on forwarding classes, you assign packets to output queues. The forwarding class plus the loss priority define the per-hop behavior (PHB in DiffServ) of a packet. J Series Services Routers and SRX Series Services Gateways support eight queues (0 through 7). For a classifier to assign an output queue (default queues 0 through 3) to each packet, it must associate the packet with one of the following forwarding classes:

• Expedited forwarding (EF)—Provides a low loss, low latency, low jitter, assured bandwidth, end-to-end service.
• Assured forwarding (AF)—Provides a group of values you can define and includes four subclasses: AF1, AF2, AF3, and AF4, each with three drop probabilities: low, medium, and high.
• Best effort (BE)—Provides no service profile. For the BE forwarding class, loss priority is typically not carried in a class-of-service (CoS) value, and random early detection (RED) drop profiles are more aggressive.
• Network Control (NC)—This class is typically high priority because it supports protocol control.

The SRX Series devices support eight queues.

Note: Queues 4 through 7 are not mapped to forwarding classes. To use queues 4 through 7, you must create custom forwarding class names and map them to the queues. For more information, see Forwarding Class Queue Assignments.

In addition to BA and MF classification, the forwarding class (FC) of a packet can be directly determined by the logical interface that receives the packet. This FC of a packet can be configured using CLI commands, and if configured, this FC overrides the FC from any BA classification that was previously performed on the logical interface.

The following CLI commands can assign a forwarding class directly to packets received at a logical interface:

Loss Priorities

Loss priorities allow you to set the priority of dropping a packet. You can use the loss priority setting to identify packets that have experienced congestion. Typically, you mark packets exceeding some service level with a high loss priority—a greater likelihood of being dropped. You set loss priority by configuring a classifier or a policer. The loss priority is used later in the work flow to select one of the drop profiles used by random early detection (RED).

You can configure the packet loss priority (PLP) bit as part of a congestion control strategy. The PLP bit can be configured on an interface or in a filter. A packet for which the PLP bit is set has an increased probability of being dropped during congestion.

Forwarding Policy Options

CoS-based forwarding (CBF) enables you to control next-hop selection based on a packet’s class of service and, in particular, the value of the IP packet's precedence bits. For example, you can specify a particular interface or next hop to carry high-priority traffic while all best-effort traffic takes some other path. CBF allows path selection based on class. When a routing protocol discovers equal-cost paths, it can pick a path at random or load-balance across the paths through either hash selection or round-robin selection.

Forwarding policy also allows you to create CoS classification overrides. You can override the incoming CoS classification and assign the packets to a forwarding class based on their input interface, input precedence bits, or destination address. When you override the classification of incoming packets, any mappings you configured for associated precedence bits or incoming interfaces to output transmission queues are ignored.

Transmission Queues

After a packet is sent to the outgoing interface on a device, it is queued for transmission on the physical media. The amount of time a packet is queued on the device is determined by the availability of the outgoing physical media as well as the amount of traffic using the interface.

J Series Services Routers and SRX Series Services Gateways support queues 0 through 7. If you configure more than eight queues on a device, the commit operation fails and the device displays a detailed message stating the total number of queues available.

Schedulers

An individual device interface has multiple queues assigned to store packets temporarily before transmission. To determine the order to service the queues, the device uses a round-robin scheduling method based on priority and the queue's weighted round-robin (WRR) credits. JUNOS schedulers allow you to define the priority, bandwidth, delay buffer size, rate control status, and RED drop profiles to be applied to a particular queue for packet transmission. For more information, see Scheduler Settings.

You can configure per-unit scheduling (also called logical interface scheduling). Per-unit scheduling allows you to enable multiple output queues on a logical interface and associate an output scheduler with each queue.

Transmit Rate

The transmission rate determines the traffic transmission bandwidth for each forwarding class you configure. The rate is specified in bits per second (bps). Each queue is allocated some portion of the bandwidth of the outgoing interface.

This bandwidth amount can be a fixed value, such as 1 megabit per second (Mbps), a percentage of the total available bandwidth, or the rest of the available bandwidth. You can limit the transmission bandwidth to the exact value you configure, or allow it to exceed the configured rate if additional bandwidth is available from other queues (SRX3600, SRX3800, SRX5600, and SRX5800 devices do not support an exact value transmit rate). This property helps ensure that each queue receives the amount of bandwidth appropriate to its level of service.

The minimum transmit rate supported on high-speed interfaces is one-ten thousandth of the speed of that interface. For example, on a Gigabit Ethernet interface with a speed of 1000 Mbps, the minimum transmit rate is 100 Kbps (1000 Mbps x 1/10000). You can configure transmit rates in the range 3200 bps through 160,000,000,000 bps. When the configured rate is less than the minimum transmit rate, the minimum transmit rate is used instead.

Note: Interfaces with slower interface speeds, like T1, E1, or channelized T1/E1/ISDN PRI, cannot support minimum transmit rates because the minimum transmit rate supported on a Services Router is 3200 bps.

Transmit rate assigns the weighted round-robin (WRR) priority values within a given priority level and not between priorities. For more information, see Transmission Scheduling.

Delay Buffer Size

You can configure the delay buffer size to control congestion at the output stage. A delay buffer provides packet buffer space to absorb burst traffic up to a specified duration of delay. When the buffer is full, all packets are dropped.

The system calculates the buffer size for a queue based on the buffer allocation method you specify for it in the scheduler. See Delay Buffer Size Allocation Methods for different buffer allocation methods and Specifying Delay Buffer Sizes for Queues for buffer size calculations.

By default, all J Series device interfaces other than channelized T1/E1 interfaces support a delay buffer time of 100,000 microseconds. On channelized T1/E1 interfaces, the default delay buffer time is 500,000 microseconds for clear-channel interfaces, and 1,200,000 microseconds for NxDS0 interfaces.

On J Series devices, you can configure larger delay buffers on channelized T1/E1 interfaces. Larger delay buffers help these slower interfaces to avoid congestion and packet dropping when they receive large bursts of traffic. For more information, see Configuring Large Delay Buffers with a Configuration Editor.

Note: For a J Series Services Router, if the buffer size percentage is set to zero for T1 interfaces, traffic does not pass.

Scheduling Priority

Scheduling priority determines the order in which an output interface transmits traffic from the queues, thus ensuring that queues containing important traffic are provided better access to the outgoing interface.

The queues for an interface are divided into sets based on their priority. Each set contains queues of the same priority. The device examines the sets in descending order of priority. If at least one queue in a set has a packet to transmit, the device selects that set. If multiple queues in the set have packets to transmit, the device selects a queue from the set according to the weighted round-robin (WRR) algorithm that operates within the set.

The packets in a queue are transmitted based on the configured scheduling priority, the transmit rate, and the available bandwidth. For more information, see Transmission Scheduling.

Shaping Rate

Shaping rates control the maximum rate of traffic transmitted on an interface. You can configure the shaping rate so that the interface transmits less traffic than it is physically capable of carrying.

You can configure shaping rates on logical interfaces. By default, output scheduling is not enabled on logical interfaces. Logical interface scheduling (also called per-unit scheduling) allows you to enable multiple output queues on a logical interface and associate an output scheduler and shaping rate with the queues.

By default, the logical interface bandwidth is the average of unused bandwidth for the number of logical interfaces that require default bandwidth treatment. You can specify a peak bandwidth rate in bits per second (bps), either as a complete decimal number or as a decimal number followed by the abbreviation k (1000), m (1,000,000), or g (1,000,000,000). The range is from 1000 through 32,000,000,000 bps.

For low-speed interfaces, the queue-limit values might become lower than the interface MTU so that traffic with large packets can no longer pass through some of the queues. If you want larger-sized packets to flow through, set the buffer-size configuration in the scheduler to a larger value. For more accuracy, the 100-ms queue-limit values are calculated based on shaper rates and not on interface rates.

RED Drop Profiles

A drop profile is a feature of the random early detection (RED) process that allows packets to be dropped before queues are full. Drop profiles are composed of two main values—the queue fullness and the drop probability. The queue fullness represents percentage of memory used to store packets in relation to the total amount that has been allocated for that queue. The drop probability is a percentage value that correlates to the likelihood that an individual packet is dropped from the network. These two variables are combined in a graph-like format.

When a packet reaches the head of the queue, a random number between 0 and 100 is calculated by the device. This random number is plotted against the drop profile having the current queue fullness of that particular queue. When the random number falls above the graph line, the packet is transmitted onto the physical media. When the number falls below the graph line, the packet is dropped from the network.

When you configure the RED drop profile on an interface, the queue no longer drops packets from the tail of the queue (the default). Rather, packets are dropped after they reach the head of the queue.

You specify drop probabilities in the drop profile section of the class-of-service (CoS) configuration hierarchy and reference them in each scheduler configuration. For each scheduler, you can configure multiple separate drop profiles, one for each combination of loss priority (low, medium-low, medium-high, or high) and IP transport protocol (TCP or non-TCP or any).

Note: For J Series devices and SRX210, SRX240, and SRX650 devices, tcp and non-tcp values are not supported, only the value “any” is supported.

You can configure a maximum of 32 different drop profiles.

To configure RED drop profiles, include the following statements at the [edit class-of-service] hierarchy level of the configuration:

Default Drop Profiles

By default, if you configure no drop profiles, RED is still in effect and functions as the primary mechanism for managing congestion. In the default RED drop profile, when the fill level is 0 percent, the drop probability is 0 percent. When the fill level is 100 percent, the drop probability is 100 percent.

Virtual Channels

On J Series devices, you can configure virtual channels to limit traffic sent from a corporate headquarters to branch offices. Virtual channels might be required when the headquarters site has an expected aggregate bandwidth higher than that of the individual branch offices. The router at the headquarters site must limit the traffic sent to each branch office router to avoid oversubscribing their links.

You configure virtual channels on a logical interface. Each virtual channel has a set of eight queues with a scheduler and an optional shaper. You can use an output firewall filter to direct traffic to a particular virtual channel. For example, a filter can direct all traffic with a destination address for branch office 1 to virtual channel 1, and all traffic with a destination address for branch office 2 to virtual channel 2.

Although a virtual channel group is assigned to a logical interface, a virtual channel is not the same as a logical interface. The only features supported on a virtual channel are queuing, packet scheduling, and accounting. Rewrite rules and routing protocols apply to the entire logical interface.

Policers for Traffic Classes

Policers allow you to limit traffic of a certain class to a specified bandwidth and burst size. Packets exceeding the policer limits can be discarded, or can be assigned to a different forwarding class, a different loss priority, or both. You define policers with firewall filters that can be associated with input or output interfaces.

Rewrite Rules

A rewrite rule modifies the appropriate CoS bits in an outgoing packet. Modification of CoS bits allows the next downstream device to classify the packet into the appropriate service group. Rewriting or marking outbound packets is useful when the device is at the border of a network and must alter the CoS values to meet the policies of the targeted peer.