 | Copyright © 2009, Juniper Networks, Inc. All rights reserved. Trademark Notice. | |
Some PICs, such as the IQ2 and IQ2E PICs, have eight egress queues enabled by default on platforms that support eight queues.
The IQ2E PICs preserve all of the features of the IQ2 PICs, such as the default support for eight egress queues on platforms that support eight queues. For more information about configuring egress queues on the IQ2E PICs, see Introduction to CoS on Ethernet IQ2 Interfaces.
The IQ2E PICs add features such as the ability to perform hierarchical scheduling. You can mix IQ2 and IQ2E PICs on the same router.
The IQ2E PICs offer:
The IQ2E PICs also offer automatic scheduler allocation across ports, so there is no need to reset the PIC when this changes. Random early detection (RED) keeps statistics on a per-drop-profile basis, improving the ability to perform network capacity planning.
When you configure the per-unit-scheduler option, each logical interface (unit) gets a dedicated scheduler (one scheduler is reserved for overflow). You can configure a per-session-scheduler to shape Layer 2 Tunneling Protocol (L2TP) sessions. The behavior of these two-port scheduler modes is the same as in IQ2 PICs. However, IQ2E PICs use hierarchical schedulers and not shared schedulers. The shared-scheduler statement is not supported on the IQ2E PICs.
For more information about configuring hierarchical schedulers, including examples, see Configuring CoS Hierarchical Schedulers.
You can shape traffic at the physical interface (port), logical interface (unit), or interface set (set of units) levels. Shaping is not supported at the queue level. However, you can configure the rate-limit statement to police the traffic passing through a queue (but only in the egress direction).
At the physical interface (port) level, you can configure only a shaping rate (PIR). At the logical interface (unit) and interface set levels, you can configure both a shaping rate and a guaranteed rate (CIR). Note that the guaranteed rates at any level must be consistent with the parent level’s capacity. In other words, the sum of the guaranteed rates on the logical interface (units) should be less than the guaranteed rate on the interface set, and the sum of the guaranteed rates on the logical interface (units) and interface sets should be less than the guaranteed rate on the physical interface (port).
The weighed RED (WRED) decision on the IQ2E PICs is done at the queue level. Once the accept or drop decision is made and the packet is queued, it is never dropped. Four drop profiles are associated with each queue: low, low-medium, medium-high, and high. WRED statistics are available for each loss priority (this feature is not supported on the IQ2 PICs). Also in contrast to the IQ2 PICs, the IQ2E PICs support WRED scaling profiles, allowing a single drop profile to be reused with a wide range of values. This practice increases the effective number of WRED drop profiles.
The IQ2E PICs provide four levels of strict priorities: strict-high, high, medium-high (medium-low) and low. In contrast to the IQ2 PICs, which supported only one strict-high queue, the IQ2E PICs do not restrict the number of queues with a given priority. There is priority propagation among three levels: the logical interface, the logical interface set, and the physical port. These features are identical to the Enhanced Queueing Dense Port Concentrators (DPCs) for the MX-series routers. For more information about configuring these features, see Configuring CoS for Enhanced Queuing DPCs.
The IQ2E PIC’s queues are serviced with modified deficit round-robin (MDRR), as are the Enhanced Queueing DPCs. Excess bandwidth (bandwidth available after all guaranteed rates have been satisfied) can be shared equally or in proportion to the guaranteed rates. For more information about excess bandwidth sharing, see Configuring Excess Bandwidth Sharing.
| Copyright © 2009, Juniper Networks, Inc. All rights reserved. Trademark Notice. | |