Table of Contents
Advertisement
Raisecom
ISCOM2600G-HI (A) Series Configuration Guide
Figure 7-8 DRR scheduling
7.1.7 Congestion avoidance
By monitoring utilization of network resources (queues/memory buffer), congestion
avoidance can discard packets actively when congestion occurs or network traffic increases. It
is a traffic control mechanism that is used to resolve network overload by adjusting network
traffic.
The traditional packet loss policy uses the Tail-Drop mode to process all packets equally
without differentiating class of services. When congestion occurs, packets at the end of a
queue are discarded until congestion is resolved.
This Tail-Drop policy may cause TCP global synchronization, making network traffic change
between heavy and low and affecting link utilization.
RED
Random Early Detection (RED) discards packets randomly and prevents multiple TCP
connection from reducing transmission rate simultaneously to avoid TCP global
synchronization.
The RED algorithm configures a minimum threshold and maximum threshold for length of
each queue. In addition:
SP+WRR: a scheduling mode combining the SP scheduling and WRR scheduling. In this
mode, queues on an interface are divided into 2 groups. You can specify the queues
where SP scheduling/WRR scheduling is performed.
SP+DRR: a scheduling mode combining the SP scheduling and DRR scheduling. In this
mode, queues on an interface are divided into 2 groups. You can specify the queues
where SP scheduling/DRR scheduling is performed.
Packets are not discarded when the queue length is smaller than the minimum threshold.
All received packets are discarded when the queue length is greater than the maximum
threshold.
Packets to be received are discarded randomly when the queue length is between the
minimum and maximum thresholds. The greater the queue size is, the higher the packet
drop probability is.
Raisecom Proprietary and Confidential
Copyright © Raisecom Technology Co., Ltd.
7 QoS
293
Advertisement
Table of Contents
