Configuring CoS Scheduling Queues on Logical LSQ Interfaces
For link services IQ (lsq-) interfaces, you can specify a scheduler map for each logical unit. A logical unit represents either an MLPPP bundle or a DLCI configured on a FRF.16 bundle. The scheduler is applied to the traffic sent to an AS or Multiservices PIC running the Layer 2 link services package.
If you configure a scheduler map on a bundle, you must include the per-unit-scheduler statement at the [edit interfaces lsq-fpc/pic/port] hierarchy level. If you configure a scheduler map on an FRF.16 DLCI, you must include the per-unit-scheduler statement at the [edit interfaces lsq-fpc/pic/port:channel] hierarchy level. For more information, see the Junos OS Class of Service Configuration Guide.
If you need latency guarantees for multiclass or LFI traffic, you must use channelized IQ PICs for the constituent links. With non-IQ PICs, because queueing is not done at the channelized interface level on the constituent links, latency-sensitive traffic might not receive the type of service that it should. Constituent links from the following PICs support latency guarantees:
- Channelized E1 IQ PIC
- Channelized OC3 IQ PIC
- Channelized OC12 IQ PIC
- Channelized STM1 IQ PIC
- Channelized T3 IQ PIC
For scheduling queues on a logical interface, you can configure the following scheduler map properties at the [edit class-of-service schedulers] hierarchy level:
- buffer-size—The queue size; for more information, see Configuring Scheduler Buffer Size.
- priority—The transmit priority (low, high, strict-high); for more information, see Configuring Scheduler Priority.
- shaping-rate—The subscribed transmit rate; for more information, see Configuring Scheduler Shaping Rate.
- drop-profile-map—The random early detection (RED) drop profile; for more information, see Configuring Drop Profiles.
When you configure MLPPP and FRF.12 on M Series and T Series routers, you should configure a single scheduler with non-zero percent transmission rates and buffer sizes for queues 0 through 3, and assign this scheduler to the link services IQ interface (lsq) and to each constituent link.
When you configure FRF.16 on M Series and T Series routers, you can assign a single scheduler map to the link services IQ interface (lsq) and to each link services IQ DLCI, or you can assign different scheduler maps to the various DLCIs of the bundle, as shown in Example: Configuring an LSQ Interface as an NxT1 Bundle Using FRF.16. For the constituent links of an FRF.16 bundle, you do not need to configure a custom scheduler. Because LFI and multiclass are not supported for FRF.16, the traffic from each constituent link is transmitted from queue 0. This means you should allow most of the bandwidth to be used by queue 0. The default scheduler transmission rate and buffer size percentages for queues 0 through 3 are 95, 0, 0, and 5 percent, respectively. This default scheduler sends all user traffic to queue 0 and all network-control traffic to queue 3, and therefore it is well suited to the behavior of FRF.16. You can configure a custom scheduler that explicitly replicates the 95, 0, 0, and 5 percent queuing behaviors, and apply it to the constituent links.
 | Note: On T Series and M320 routers, the default scheduler transmission rate and buffer size percentages for queues 0 through 7 are 95, 0, 0, 5, 0, 0, 0, and 0 percent. |
For link services IQ interfaces (lsq), these scheduling properties work as they do in other PICs, except as noted in the following sections.
| Note: On T Series and M320 routers, lsq interfaces do not support DiffServ code point (DSCP) and DSCP-IPv6 rewrite markers. |
Configuring Scheduler Buffer Size
You can configure the scheduler buffer size in three ways: as a temporal value, as a percentage, and as a remainder. On a single logical interface (MLPPP or a FRF.16 DLCI), each queue can have a different buffer size.
If you specify a temporal value, the queuing algorithm starts dropping packets when it queues more than a computed number of bytes. This number is computed by multiplying logical interface speed by the temporal value. For MLPPP bundles, logical interface speed is equal to the bundle bandwidth, which is the sum of constituent link speeds minus link-layer overhead. For MLFR FRF.16 DLCIs, logical interface speed is equal to bundle bandwidth multiplied by the DLCI shaping rate. In all cases, the maximum temporal value is limited to 200 milliseconds.
Buffer size percentages are implicitly converted into temporal values by multiplying the percentage by 200 milliseconds. For example, buffer size specified as buffer-size percent 20 is the same as a 40-millisecond temporal delay. The link services IQ implementation guarantees 200 milliseconds of buffer delay for all interfaces with T1 and higher speeds. For slower interfaces, it guarantees one second of buffer delay.
The queueing algorithm evenly distributes leftover bandwidth among all queues that are configured with the buffer-size remainder statement. The queuing algorithm guarantees enough space in the transmit buffer for two MTU-sized packets.
Configuring Scheduler Priority
The transmit priority of each queue is determined by the scheduler and the forwarding class. Each queue receives a guaranteed amount of bandwidth specified with the scheduler transmit-rate statement.
Configuring Scheduler Shaping Rate
You use the shaping rate to set the percentage of total bundle bandwidth that is dedicated to a DLCI. For link services IQ DLCIs, only percentages are accepted, which allows adjustments in response to dynamic changes in bundle bandwidth—for example, when a link goes up or down. This means that absolute shaping rates are not supported on FRF.16 bundles. Absolute shaping rates are allowed for MLPPP and MLFR bundles only.
For scheduling between DLCIs in a MLFR FRF.16 bundle, you can configure a shaping rate for each DLCI. A shaping rate is expressed as a percentage of the aggregate bundle bandwidth. Shaping rate percentages for all DLCIs within a bundle can add up to 100 percent or less. Leftover bandwidth is distributed equally to DLCIs that do not have the shaping-rate statement included at the [edit class-of-service interfaces lsq-fpc/pic/port:channel unit logical-unit-number] hierarchy level. If none of the DLCIs in an MLFR FRF.16 bundle specify a DLCI scheduler, the total bandwidth is evenly divided across all DLCIs.
| Note: For FRF.16 bundles on link services IQ interfaces, only shaping rates based on percentage are supported. |
Configuring Drop Profiles
You can configure random early detection (RED) on LSQ interfaces as in other CoS scenarios. To configure RED, include one or more drop profiles and attach them to a scheduler for a particular forwarding class. For more information about RED profiles, see the Junos OS Class of Service Configuration Guide.
The LSQ implementation performs tail RED. It supports a maximum of 256 drop profiles per PIC. Drop profiles are configurable on a per-queue, per-loss-priority, and per-TCP-bit basis.
You can attach scheduler maps with configured RED drop profiles to any LSQ logical interface: an MLPPP bundle, an FRF.15 bundle, or an FRF.16 DLCI. Different queues (forwarding classes) on the same logical interface can have different associated drop profiles.
The following example shows how to configure a RED profile on an LSQ interface:
| Note: The RED profiles should be applied only on the LSQ bundles and not on the egress links that constitute the bundle. |
