Per-Priority Shaping on the Trio MPC/MIC Interfaces Overview

Per-priority shaping enables you to configure a separate shaping rate for each of the five priority levels supported by the Trio MPC/MIC interfaces family. The main use of per-priority shaping rates is to ensure that higher priority services such as voice and video do not starve lower priority services such as data.

There are five scheduler priorities supported by the Trio family:

Guaranteed high (GH)
Guaranteed medium (GM)
Guaranteed low (GL)
Excess high (EH)
Excess low (EL)

The five scheduler priorities support a shaping rate for each priority supported by the Trio family:

Shaping rate priority high (GH)
Shaping rate priority medium (GM)
Shaping rate priority low (GL)
Shaping rate excess high (EH)
Shaping rate excess low (EL)

If each service is represented by a forwarding class queued at a separate priority, then assigning a per-priority shaping rate to higher priority services accomplishes the goal of preventing the starvation of lower priority services.

To configure per-priority shaping rates, include the shaping-rate-excess-high rate <burst-size burst>, shaping-rate-excess-low rate <burst-size burst>, shaping-rate-priority-high rate <burst-size burst>, shaping-rate-priority-low rate <burst-size burst>, or shaping-rate-priority-medium rate <burst-size burst> at the [edit class-of-service traffic-control-profiles tcp-name] hierarchy level and apply the traffic control profile at the [edit interfaces] hierarchy level. You can specify the rate in absolute values, or by using k (kilo-), m (mega-) or g (giga-) units.

You can include one or more of the per-priority shaping statements in a traffic control profile:

[edit class-of-service]traffic-control-profiles {tcp-ge-port {shaping-rate-excess-high rate <burst-size bytes>;shaping-rate-excess-low rate <burst-size bytes>;shaping-rate-priority-high rate <burst-size bytes>;shaping-rate-priority-low rate <burst-size bytes>;shaping-rate-priority-medium rate <burst-size bytes>;}}

Note: We do not recommend configuring the burst size because this can have far-reaching and unintended effects.

There are several important points about per-priority shaping rates:

Per-priority shaping rates are only supported on Trio MPC/MIC interface family interfaces (but not the Trio DPC interfaces).
Per-priority shaping is only available for level 1 and level 2 scheduler nodes. (For more information on hierarchical schedulers, see Configuring Hierarchical Schedulers for CoS.
Per-priority shaping rates are supported when level 1 or level 2 scheduler nodes have static or dynamic interfaces above them. There is no support for dynamic level 2 interface sets.
Per-priority shaping rates are supported on aggregated Ethernet (AE) interfaces.
Per-priority shaping rates are only supported in traffic control profiles.

Per-priority shaping rates can be helpful when the MX Series Ethernet Services Router is in a position between subscriber traffic on an access network and the carrier network, playing the role of a broadband services router. In that case, the MX Series router provides quality-of-service parameters on the subscriber access network so that each subscriber receives a minimum bandwidth (determined by the guaranteed rate) and a maximum bandwidth (determined by the shaping rate). This allows the devices closer to the carrier network to operate more efficiently and more simply and reduces operational network expenses because it allows more centralized network management.

One architecture for using per-priority shaping on the MX Series router is shown in Figure 24. In the figure, subscribers use residential gateways with various traffic classes to support voice, video, and data services. The MX Series router sends this traffic from the carrier network to the digital subscriber line access multiplexer (DSLAM) and from the DSLAM on to the residential gateway devices.

Figure 24: Architecture for Trio DPC and MPC/MIC interface Per-priority Shaping

Image g017442.gif

One way that the MX Series router can provide service classes for this physical network topology is shown in Figure 25. In the figure, services such as voice and video are placed in separate forwarding classes and the services at different priority levels. For example:

All expedited-forwarding queues are services at a priority level of guaranteed high.
All assured-forwarding queues are services at a priority level of guaranteed medium.
All better-than-best-effort queues are services at a priority level of excess high.
All best-effort queues are services at a priority level of excess low.

Note: This list covers only one possible configuration. Others are possible and reasonable, depending on the service provider’s goals. For example, best-effort and better-than-best-effort traffic can have the same priority level, with the better-then-best-effort forwarding class having a higher scheduler weight than the best-effort forwarding class. For more information on forwarding classes, see Configuring Forwarding Classes.

Figure 25: Scheduling Hierarchy for Trio Per-Priority Shaping

Image g017443.gif

Aggregated voice traffic in this topology is shaped by applying a high-priority shaper to the port. Aggregated video traffic is shaped in the same way by applying a medium-priority shaper to the port. As long as the sum of the high- and medium-priority shapers is less than the port speed, some bandwidth is reserved for best-effort and better-then-best-effort traffic. So assured-forwarding and expedited-forwarding voice and video cannot starve best-effort and better-then-best-effort data services. One possible set of values for high-priority (guaranteed high) and medium-priority (guaranteed medium) traffic is shown in Figure 25.

Note: We recommend that you do not shape delay-sensitive traffic such as voice traffic because it adds delay (latency). Service providers often use connection admission control (CAC) techniques to limit aggregated voice traffic. However, establishing a shaping rate for other traffic guards against CAC failures and can be useful in pacing extreme traffic bursts.

You must still apply the traffic control profile to the port

Per-priority shaping statements:

[edit class-of-service]traffic-control-profile {tcp-for-ge-port {shaping-rate-priority-high 500k;shaping-rate-priority-medium 100m;}}

Apply (attach) the traffic control profile to the physical interface (port) at the [edit class-of-services interfaces] hierarchy level:

[edit class-of-service]interfaces {ge-1/0/0 {output-traffic-control-profile tcp-for-ge-port;}}

Traffic control profiles with per-priority shaping rates can only be attached to interfaces that support per-priority shaping.

You can apply per-priority shaping to levels other than the level 1 physical interface (port) of the scheduler hierarchy. Per-priority shaping can also be applied at level 2, the interface set level, which would typically represent the digital subscriber link access multiplexer (DSLAM). At this level you could use per-priority shaping to limit to total amount of video traffic reaching a DSLAM, for example.

You apply (attach) the traffic control profile to an interface set at the [edit class-of-services interfaces] hierarchy level:

[edit class-of-service]interfaces {interface-set svlan-1 {output-traffic-control-profile tcp-for-ge-port;}}

Note: Although you can configure both input and output traffic control profiles, only output traffic control profiles are supported for per-priority shaping.

You can also apply per-priority shaping in a dynamic interface configuration, not in a static configuration as shown so far. To apply per-priority shaping to a dynamic configuration, configure the statements at the [edit dynamic-profiles] hierarchy level:

You must still apply the traffic control profile to the port

Per-priority shaping for dynamic profiles statements:

[edit dynamic-profiles class-of-service]class-of-service {traffic-control-profile {tcp-for-ge-port {shaping-rate-priority-high 500k;shaping-rate-priority-medium 100m;}}}

You can configure per-priority shaping for the traffic remaining with the output-traffic-control-profile-remaining statement on a physical port (a level 2 node) but not for an interface set (a level 3 node).