CoS Feature Differences Between PTX Series Packet Transport Routers and T Series Routers

This topic provides a list of class-of-service features available on PTX Series routers and compares them with class-of-service features on T Series routers.

Classifiers

• T Series routers support VRF table labels for Layer 3 VPNs. On PTX Series routers, this feature is not supported, unless the router is running Junos OS Evolved. PTX Series routers running Junos OS Evolved support Layer 3 VPNs.

• On T Series routers, IEEE 802.1 classifiers cannot coexist with Layer 3 classifiers. On PTX Series routers, these classifiers can coexist.

• On T Series routers, IEEE classifiers are supported on Ethernet IQ, IQ2, and IQ2-E interfaces. These interfaces have the flexibility of classifying traffic based on inner or outer VLAN tags. On PTX Series routers, IEEE classification is always based on outer VLAN tags.

Rewrite

• PTX Series routers do not support rewrite of both exp and inet-precedence fields using:

  • exp protocol mpls-any

  • exp protocol mpls-inet-both

  • exp protocol mpls-inet-both-non-vpn

• On T Series routers, DSCP rewrite and DSCP IPv6 rewrite are not supported for the MPLS protocol. PTX Series routers support rewrite of both DSCP and DSCP IPv6 for protocol MPLS.

• PTX Series routers support Layer 2 rewrite of 802.1p and 802.1ad, to either the outer VLAN tag, or both outer and inner VLAN tags.

Forwarding Class

• On T Series routers, you can override the default fabric priority queuing of egress traffic by including the priority statement at the following hierarchy level:

  content_copy zoom_out_map

   [class-of-service forwarding-classes queue queue-number class-name]
    priority (high | low);
  

  On PTX Series routers, fabric priority queuing is not supported; therefore, the priority statement for forwarding-classes is not supported.

Tricolor Marking

• On T Series routers, the copy-plp-all statement must be configured to support tricolor marking. On PTX Series routers, tricolor marking is enabled by default.

Schedulers

• T Series routers, which use egress queuing architecture, support chassis and fabric schedulers. Alternatively, PTX Series routers support a virtual output queuing (VOQ) architecture and the fabric schedulers utilize the CoS scheduling parameters to configure the fabric schedulers. There is not a separate configuration for the fabric schedulers on PTX Series routers. With the VOQ architecture, packets are queued and dropped on ingress during congestion.

• On T Series routers, high-priority queues have precedence to acquire excess bandwidth and might consume all excess bandwidth. On PTX Series routers, excess bandwidth is shared based on the ratio of the configured transfer rate. Therefore, all priority queues get a share of excess bandwidth.

• On T Series routers, strict-high-priority queues and high-priority queues are assigned the same hardware priority. On PTX Series routers, strict-high-priority queues and high-priority queues are assigned different hardware priorities. Strict-high-priority can starve other queues if a rate limiter is not applied on PTX Series routers.

• On T Series routers, if a strict-high-priority queue is oversubscribed, it can block all other queues except high-priority queues. On PTX Series routers, if a strict-high-priority queue is oversubscribed, it can block all other queues including high-priority queues.

  To restrict the bandwidth of strict-high priority queues on PTX Series routers, use the transmit-rate rate-limit configuration statement.

• On both T Series routers and PTX Series routers, if a strict-high-priority queue is oversubscribed and results in oversubscription of the guaranteed bandwidth, the distribution of bandwidth that is not taken up by strict-high-priority queues is undetermined. T Series routers and PTX Series routers distribute this unused bandwidth differently.

Buffer Size and Latency

• On T Series routers, memory allocation dynamic (MAD) is enabled by default and can be disabled. A comparable feature on PTX Series routers is buffer-size shared mode of schedulers.

• On T Series routers, the maximum delay bandwidth buffering configured per queue is 50 ms. On PTX Series routers, buffer-size percent 100 requests the platform's full buffer latency at the port or shaping rate, that is, a nominal queue limit of (productFullLatency × portRate) bytes. For all PTX routers, the productFullLatency is 100 ms except for the PTX10001, and line cards LC1201 and LC1202, which have a 25 ms buffer.

• On T Series routers, the maximum latency associated with a packet is fairly consistent and independent of the number of sources sending the traffic to an interface. On PTX Series routers, over-provisioning is possible. When traffic is sent from multiple Packet Forwarding Engines (PFEs), the latency can be higher than when traffic is sent from one PFE due to the PTX dynamically adjusting buffers from multiple PFEs for a Virtual Output Queue (VOQ).

• On T Series routers, a high-priority queue has lower latency than a low priority queue with the same configured transfer rate and same offered load. On PTX Series routers, there is no latency difference.

Drop Profile

• PTX Series PFEs not support drop-profile assignments for a queue based on the protocol. As a consequence, the protocol option for the drop-profile-map configuration statement is treated as protocol any.

Interface Queue Statistics

• On T Series routers, transmitted byte counters are computed using Layer 3 packet length. On PTX Series routers, transmitted byte counters are computed using the full Layer 2 overhead (including all L2 encapsulation and CRC) plus 12 for the inter-packet gap plus 8 for the preamble.

• On Junos OS Evolved PTX Series routers , the tail-dropped counters and the RED-dropped counters are displayed separately in the output of the show interfaces queue command. On Junos OS PTX Series routers, tail-dropped counters are always zero. All the packet drops are shown as RED-dropped in the show interfaces voq output.