EtherHaul Operation, Administration and Maintenance Manual
Page 75
9
Quality of Service
The EtherHaul incorporates an advanced Quality of Service engine which complements
the hitless adaptive bandwidth, coding and modulation mechanism.
Quality of Service (QoS) mechanism enables service providers to offer different classes
of service for different types of traffic or customers and is especially important in
wireless links with adaptive capabilities, because changing link conditions may require
the system to drop some traffic according to a predetermined priority and sched uling
scheme.
The EtherHaul QoS engine classifies the incoming packets onto streams using any
combination of VID, PCP, DSCP (or alternatively MPLS EXP bit) fields. Each stream may
be assigned a bandwidth profile with CIR (committed rate), CBS (committed bu rst size),
EIR (excess rate), EBS (excess burst size) and color-mode.
The implemented mechanism supports 3 colors and 2 rates. The frames that fit into
CIR/CBS profile marked drop ineligible and colored “green”. The traffic which is within
excess profile but exceeds committed profile is marked drop eligible (“yellow”). Upon
congestion at egress interface the yellow packets are dropped first. The rest, which are
out of profile, are colored “red” and discarded. The “red” frames are dropped using a
head drop algorithm and “green” frames take precedence of “yellow”.
Figure 9-1 Classification and Policing
These packets are then mapped to 1 of the 8 priority queues (per interface). These
queues are accessed by the scheduling mechanism.
The priority queues of the EtherHaul are accessed using the following scheduling
mechanisms:
Strict Priority: lower priority packets are served only if all higher priority queues are
empty.
Weighted Fair Queuing (WFQ): data packet scheduling technique maintaining
fairness by applying weights to the queues. Each queue is serviced in the order of its
weighted proportion to the available resources.
Shaper: used to control traffic flows in order to optimize or guarantee performance
and improve latency by limiting the maximum bandwidth of certain flows to
maintain fairness and to assure SLA.
