Overview

Congestion avoidance techniques monitor network traffic loads in an effort to anticipate and avoid congestion at common network bottleneck points. You achieve congestion avoidance through packet dropping, by using more complex techniques than simple tail drop. With the addition of explicit congestion notification (ECN) extensions to IP, routers have an alternative method of indicating congestion to peers.

This lesson describes the concept of ECN and the Cisco IOS commands that are required to configure and monitor ECN.

Objectives

Upon completing this lesson, you will be able to configure ECN to enhance the congestion avoidance features of WRED. This ability includes being able to meet these objectives:

■ Describe the ECN extensions to IP

■ Identify characteristics of the ECN field in IP

■ Explain how ECN interacts with WRED

■ Identify the Cisco IOS commands that are required to configure ECN extensions to WRED

■ Identify the Cisco IOS commands that are required to monitor ECN-enabled WRED

Explicit Congestion Notification

This topic describes the ECN extensions to IP.

This topic describes the ECN extensions to IP.

TCP determines how many unacknowledged packets it can send by gradually increasing the number of packets the session sends until it experiences a dropped packet; this is known as the TCP window size. As a result, TCP tends to cause router queues to build up at network bottleneck points. When queues become full, tail drop begins dropping all incoming packets until there is room in the queue. Tail drop does not provide differential treatment and therefore some of the fragile flow packets, sensitive to latency, may be dropped. In addition, tail drop can lead to global synchronization of packet loss across multiple flows.

Active queue management mechanisms such as random early detection (RED) or weighted random early detection (WRED) detect congestion before queues fill and overflow. Through the use of selective packet discard, these mechanisms provide congestion indication to end nodes. Therefore, active queue management (congestion avoidance) mechanisms can reduce queuing delays for all traffic sharing a specific queue. In addition, active queue management means that it is no longer necessary to rely on buffer overflow as the only means of indicating congestion.

Traditional active queue management mechanisms, such as RED, rely on the use of packet drops to indicate congestion. Packet dropping in these mechanisms is based on the average queue length exceeding a predefined threshold, rather than only when queues overflow. However, because packets are dropped prior to queues actually overflowing, the router dropping the packet is not always constrained by memory limitations and needs to actually drop the packet. With the Addition of Explicit Congestion Notification to IP (RFC 3168), active queue management allows routers to signal that congestion has been experienced by the router, instead of relying on the use of packet drops. Through the use of signaling congestion, aggressive flows can be slowed, thus reducing the impact of congestion and packet loss on latency-sensitive flows.

Was this article helpful?

0 0

Post a comment