Dual Pass Queuing - Alcatel-Lucent 7450 Quality Of Service Manual

HSMDA QoS Model

Dual Pass Queuing

In the standard queuing model, the ingress hardware performs both per service ingress SLA
enforcement as well as per switch fabric destination based virtual output queuing. Due to the
requirement that each queue on ingress be mapped to a single switch fabric destination, when
a SAP for a service type that forwards to multiple switch fabric destinations (such as VPLS,
IES and VPRN services), a single ingress service queue is created as multiple hardware
virtual output queues in the hardware. In environments where the SAP or subscriber density
requires more service queues than can be created with the available ingress hardware queues,
two passes through the hardware are performed. This dual pass mechanism allows a single
hardware queue to represent a service queue on the first pass through the hardware while the
second pass allows the hardware to use shared virtual output queues into the switch fabric.
HSMDA removes the need to perform two passes through ingress since the service or
subscriber-based queuing is being performed on the XMA or MDA. This frees the ingress
hardware to perform the virtual output queuing function using shared queues mapped to
switch fabric destinations.
Egress Intermediate Destination Secondary Shapers
The HSMDA supports placement at the apex of an aggregation network servicing multiple
DSLAMs or other subscriber last-mile aggregation devices such as DSLAMs or gigabit
passive optical network optical network terminals (GPON ONT). When operating in this
fashion, multiple subscriber aggregators will be reached over an HSMDA 10GE port. The
egress port scheduler combines all subscriber queues of the same scheduling class and
services the queues in a byte fair round robin fashion. An effect of this behavior results in
more packets being forwarded into the aggregation network towards a DSLAM than the
DSLAM can accept. If the HSMDA egress port is congested, the egress bandwidth
represented by the downstream discarded packets to the DSLAM may have been allocated to
packets destined to other DSLAMs. The HSMDA supports egress secondary shapers that are
used to provide a control mechanism to prevent downstream overruns without affecting the
class based scheduling behavior on the port. When used in this manner, the secondary shaper
is called an intermediate destination shaper. Egress secondary shapers are configured on a per
port basis.
762
• Intermediate destination shaper (egress)
• Per queue-group shaper
• Per queue shaper
Quality of Service Guide