Frequently used QoS terms

This section provides definitions for frequently used QoS terminology.

The following terms are used in router Hierarchical QoS to describe the operation and maintenance of a virtual scheduler hierarchy and are presented for reference purposes.

Above-CIR distribution is the second phase of bandwidth allocation between a parent scheduler and its child queues and child schedulers. The bandwidth that is available to the parent scheduler after the within-CIR distribution is distributed among the child members using each child’s level (to define strict priority for the above-CIR distribution), weight (the ratio at a specific level with several children), and the child’s rate value. A rate value equal to the child’s CIR value results in a child not receiving any bandwidth during the above-CIR distribution phase.

Available bandwidth is the bandwidth usable by a parent scheduler to distribute to its child queues and schedulers. The available bandwidth is limited by the parent’s schedulers association with its parent scheduler. If the parent scheduler has a parent of its own and the parent schedulers defined rate value, then available bandwidth is distributed to the child queues and schedulers using a within-CIR distribution phase and an above-CIR distribution phase. Distribution in each phase is based on a combination of the strict priority of each child and the relative weight of the child at that priority level. Separate priority and weight controls are supported per child for each phase.

The Committed Burst Size (CBS) specifies the relative amount of reserved buffers for a specific ingress network XMA or MDA forwarding class queue or egress network port forwarding class queue. The value is entered as a percentage.

The Committed Information Rate (CIR) defines the amount of bandwidth committed to the scheduler or queue.

• For schedulers, the CIR value can be explicitly defined or derived from summing the child member CIR values.

• On a queue, the CIR value is explicitly defined.

  The CIR rate for ingress queues controls the in-profile and out-of-profile policing and ultimately egress in-profile and out-of-profile marking. Queue CIR rates also define the hardware fairness threshold at which the queue is no longer prioritized over other queues.

A child’s (queue or scheduler) CIR is used with the CIR level parameter to determine the child’s committed bandwidth from the parent scheduler. When multiple children are at the same strict CIR level, the CIR weight further determines the bandwidth distribution at that level.

The CIR level parameter defines the strict level at which bandwidth is allocated to the child queue or scheduler during the within-CIR distribution phase of bandwidth allocation. All committed bandwidth (determined by the CIR defined for the child) is allocated before any child receives non-committed bandwidth. Bandwidth is allocated to children at the higher CIR levels before children at a lower level. A child CIR value of zero or an undefined CIR level results in bandwidth allocation to the child only after all other children receive their provisioned CIR bandwidth. When multiple children share a CIR level, the CIR weight parameter further defines bandwidth allocation according to the child’s weight ratio.

The CIR weight parameter defines the weight within the CIR level given to a child queue or scheduler. When multiple children share the same CIR level on a parent scheduler, the ratio of bandwidth given to an individual child is dependent on the ratio of the weights of the active children. A child is considered active when a portion of the offered load is within the child’s defined CIR rate. The ratio is calculated by first adding the CIR weights of all active children, then dividing each child’s CIR weight by the sum. If a child’s CIR level parameter is not defined, that child is not included in the within-CIR distribution and the CIR weight parameter is ignored. A CIR weight of zero forces the child to receive bandwidth only after all other children at that level have received their within-CIR bandwidth. When several children share a CIR weight of zero, all are treated equally.

Child is a logical state of a queue or scheduler that has been configured with a valid parent scheduler association. The child/parent association is used to build the hierarchy among the queues and schedulers.

The level parameter defines the strict priority level for a child queue or scheduler with regards to bandwidth allocation during the above-CIR distribution phase on the child’s parent scheduler. This allocation of bandwidth is done after the within-CIR distribution is finished. All child queues and schedulers receive the remaining bandwidth according to the strict priority level in which they are defined with higher levels receiving bandwidth first and lower levels receiving bandwidth if available.

The Maximum Burst Size (MBS) command specifies the relative amount of the buffer pool space for the maximum buffers for a specific ingress network XMA or MDA forwarding class queue or egress network port forwarding class queue. The value is entered as a percentage.

Offered load is evaluated per child in the scheduler hierarchy. The offered load is the amount of bandwidth a child queue or scheduler can use to accommodate the data passing through the child. It is separated into two portions; within-CIR and above-CIR. Within-CIR offered load is the portion of bandwidth required to meet the child’s CIR value. It can be less than the CIR value but never greater. If the forwarding requirement for the child is greater than the CIR value, the remaining is considered to be the above-CIR offered load. The sum of the within-CIR and above-CIR offered load cannot be greater than the maximum rate defined for the child.

When a child queue is configured with a parent scheduler specified but the parent scheduler does not exist on the object the queue is created on, the state is considered orphaned.

An orphaned state is not the same condition as when a queue is not defined with a parent association. Orphan states are cleared when the parent scheduler becomes available on the object. This can occur when a scheduler policy containing the parent scheduler name is applied to the object that the queue exists on or when the scheduler name is added to the scheduler policy already applied to the object that the queue exists on.

A scheduler becomes a parent when a queue or scheduler defines it as its parent. A queue or scheduler can be a child of only one scheduler. When defining a parent association on a child scheduler, the parent scheduler must already exist in the same scheduler policy and on a scheduler tier higher (numerically lower) than the child scheduler. Parent associations for queues are only checked when, when an instance of the queue is created on a SAP.

A queue is where packets that are forwarded are buffered before scheduling. Packets are not actually forwarded through the schedulers; they are forwarded from the queues directly to ingress or egress interfaces. The association between the queue and the virtual schedulers is intended to accomplish bandwidth allocation to the queue. Because the offered load is derived from queue utilization, bandwidth allocation is dependent on the queue distribution among the scheduler hierarchy. Queues can be tied to only one scheduler within the hierarchy.

The rate defines the maximum bandwidth that is made available to the scheduler or queue. The rate is defined in kilobits per second (kb/s).

• On a scheduler, the rate setting is used to limit the total bandwidth allocated to the scheduler’s child members.

• For queues, the rate setting is used to define the Peak Information Rate (PIR) at which the queue can operate.

A scheduler that has no parent scheduler association (is not a child of another scheduler) is considered to be a root scheduler. With no parent scheduler, bandwidth used by a root scheduler is dependent on offered load of child members, the maximum rate defined for the scheduler, and total overall available bandwidth. Any scheduler can be a root scheduler. Because parent associations are not allowed in Tier 1, all schedulers in Tier 1 are considered be a root scheduler.

A scheduler policy represents a particular grouping of virtual schedulers that are defined in specific scheduler tiers. The tiers and internal parent associations between the schedulers establish the hierarchy among the virtual schedulers. A scheduler policy can be applied to either a multiservice site or to a Service Access Point (SAP). When the policy is applied to a site or SAP, the schedulers in the policy are instantiated on the object and are available for use by child queues directly or indirectly associated with the object.

A tier is an organizational configuration used within a scheduler policy to define the place of schedulers created in the policy. Three tiers are supported: Tier 1, Tier 2, and Tier 3. Schedulers defined in Tier 2 can have parental associations with schedulers defined in Tier 1. Schedulers defined in Tier 3 can have parental associations with schedulers defined at Tiers 1 or 2. Queues can have parental associations with schedulers at any tier level.

A virtual scheduler, defined by a name (text string), is a logical configuration used as a parent to a group of child members that are dependent upon a common parent for bandwidth allocation. The virtual scheduler can also be a child member to another parent virtual scheduler and receive bandwidth from that parent to distribute to its child members.

The weight parameter defines the weight within the above-CIR level given to a child queue or scheduler. When several children share the same level on a parent scheduler, the ratio of bandwidth give to an individual child is dependent on the ratio of the weights of the active children. A child is considered active when a portion of the offered load is above the CIR value (also bounded by the child’s maximum bandwidth defined by the child’s rate parameter). The portion of bandwidth given to each child is based on the child’s weight compared to the sum of the weights of all active children at that level. A weight of zero forces the child to receive bandwidth only after all other children at that level have received their above-CIR bandwidth. When several children share a weight of zero, all are treated equally.

Within the CIR distribution process is the initial phase of bandwidth allocation between a parent scheduler and its child queues and child schedulers. The bandwidth that is available to the parent scheduler is distributed first among the child members using each child’s CIR level (to define a strict priority for the CIR distribution), CIR weight (the ratio at a specific CIR level with several children), and the child’s CIR value. A CIR value of zero or an undefined CIR level causes a child to not receive any bandwidth during the CIR distribution phase. If the parent scheduler has any bandwidth remaining after the within-CIR distribution phase, it is distributed using the above-CIR distribution phase.