Configuring CoS on Link Services Interfaces

For link services IQ (lsq-) interfaces, Junos class of service (CoS) is fully supported and functions as described in the Junos Class of Service Configuration Guide. For more information and detailed configuration examples, see Layer 2 Service Package Capabilities and Interfaces.

For link services (ls-) interfaces, CoS functions differently depending on whether the interface is configured on a Link Services PIM in a J Series router or on a Link Services PIC in an M Series or T Series router. For more information, see the following section or the Junos OS Interfaces and Routing Configuration Guide:

CoS for Link Services Interfaces on M Series and T Series Routers

For Link Services PIC interfaces (ls) on M Series and T Series routers, queue 0 is the only queue that you should configure to receive fragmented packets. Configure all other queues to be higher-priority queues.

Table 19 summarizes how CoS queues work on link services (ls) interfaces.

Table 19: Link Services CoS Queues

Supported Bundling Type	Queue 0	Higher-Priority Queues
Hash-based load balancing	No	Yes
MLFR FRF.15	Yes	No
MLFR FRF.16	Yes	No
MLPPP	Yes	No

For M Series and T Series routers, CoS on link services (ls) interfaces works as follows:

On all platforms, the Link Services PIC currently supports up to four queues: 0, 1, 2, and 3.
Queue 0 uses MLFR FRF.15, MLFR FRF.16, or MLPPP to bundle packets.
Higher-priority queues (1, 2, and 3) use hash-based load balancing to bundle packets. IP and MPLS header information is included in the hash.
MLPPP packets traversing link services interfaces using queue 0 are fragmented and distributed across the constituent links. Queue 0 packets are sent on the least utilized link, proportional to its bandwidth. The queue 0 load balancer attempts to maintain even distribution of all traffic across all constituent links. In situations with a small number of high-priority traffic flows (queues 1, 2, and 3), queue 0 traffic might be unevenly distributed.
For the MLFR FRF.16 protocol, only queue 0 works. If you configure a bundled interface to use MLFR FRF.16 with queue 0, then you must ensure the classifier does not send any traffic to queues 1, 2, and 3 on that interface.
To carry high-priority traffic correctly on MLFR FRF.16 interfaces, you must configure an output firewall filter that forces all traffic into queue 0 on the ls-fpc/pic/port.channel interface.
MLFR FRF.15 and MLPPP interfaces support CoS through packet interleaving. The MLFR FRF.16 standard does not support packet interleaving, so all packets destined for an FRF.16 PVC interface must egress from the same queue.
For constituent link interfaces of Link Services PICs, you can configure standard scheduler maps.
For input packets and fragments received from constituent links, you can use regular input firewall filters and standard CoS classifiers on the link services interface.
For packets that pass through a link services interface and are destined for a constituent link interface, all traffic using queue 0 is fragmented. Traffic using higher-priority queues (1, 2, and 3) is not fragmented.
For MLFR FRF.15 and MLPPP, routing protocol packets smaller than 128 bytes are sent to queue 3; routing protocol packets that exceed 128 bytes are sent to queue 0 and fragmented accordingly. For MLFR FRF.16, queue 0 is used for all packet sizes.
You must configure output firewall classification for egress traffic on the link services interface, not directly on the constituent link interface directly.
Inverse multiplexing for ATM (IMA) is not supported on link services interfaces.

For more information, see Configuring Delay-Sensitive Packet Interleaving on Link Services Logical Interfaces and the Junos Policy Framework Configuration Guide.

Example: Configuring CoS on Link Services Interfaces

Configure CoS on a link services interface and its constituent link interfaces.

Note: This example applies to M Series and T Series routers. For examples that apply to a Link Services PIC installed on a J Series router, see the Junos OS Interfaces and Routing Configuration Guide.

Packets that do not match the firewall filters are sent to a queue that performs load balancing by sending fragments to all constituent links.

Packets that match the firewall filters are sent to a queue that does not support packet fragmentation and reassembly; instead, this traffic is load-balanced by sending each packet flow to a different constituent link. Each packet that matches a firewall filter is subjected to a hash on the IP source address and the IP destination address to determine the packet flow to which each packet belongs.

When you configure the MLPPP encapsulation type or the multilink FRF.15 Frame Relay end-to-end encapsulation type, routing protocol packets smaller than 128 bytes are sent to the network-control queue on the constituent link interface. This keeps routing protocols operating normally, even when low-speed links are congested by regular packets.

[edit interfaces]ls-7/0/0 {unit 0 {encapsulation multilink-ppp;interleave-fragments;family inet {filter {output lfi_ls_filter;}address 10.54.0.2/32 {destination 10.54.0.1;}}}}ge-7/2/0 {unit 0 {family inet {address 192.168.1.1/24;}}}ce1-7/3/6 {no-partition interface-type e1;}e1-7/3/6 {encapsulation ppp;unit 0 {family mlppp {bundle ls-7/0/0.0;}}}ce1-7/3/7 {no-partition interface-type e1;}e1-7/3/7 {encapsulation ppp;unit 0 {family mlppp {bundle ls-7/0/0.0;}}}[edit class-of-service]classifiers {dscp dscp_default {import default;}inet-precedence inet-precedence_default {import default;}}code-point-aliases {dscp {af11 001010;af12 001100;af13 001110;af21 010010;af22 010100;af23 010110;af31 011010;af32 011100;af33 011110;af41 100010;af42 100100;af43 100110;be 000000;cs1 001000;cs2 010000;cs3 011000;cs4 100000;cs5 101000;cs6 110000;cs7 111000;ef 101110;}inet-precedence {af11 001;af21 010;af31 011;af41 100;be 000;cs6 110;cs7 111;ef 101;nc1 110;nc2 111;}}forwarding-classes {queue 0 be;queue 1 ef;queue 2 af;queue 3 nc;}interfaces {ge-7/2/0 {scheduler-map sched-map;unit 0 {classifiers {dscp dscp_default;}}}e1-7/3/6 {scheduler-map sched-map;}e1-7/3/7 {scheduler-map sched-map;}ls-7/0/0 {scheduler-map sched-map;unit 0 {classifiers {inet-precedence inet-precedence_default;}}}}scheduler-maps {sched-map {forwarding-class af scheduler af-scheduler;forwarding-class be scheduler be-scheduler;forwarding-class ef scheduler ef-scheduler;forwarding-class nc scheduler nc-scheduler;}}schedulers {af-scheduler {transmit-rate percent 25;buffer-size percent 25;}be-scheduler {transmit-rate percent 25;buffer-size percent 25;}ef-scheduler {transmit-rate percent 25;buffer-size percent 25;}nc-scheduler {transmit-rate percent 25;buffer-size percent 25;}}[edit firewall]filter lfi_ls_filter {term term0 {from {destination-address {192.168.1.3/32;}precedence 5;}then {count count-192-168-1-3;forwarding-class af;accept;}}term default {then {log;forwarding-class best effort;accept;}}}