Specifications

4-65
Cisco Internet Streamer CDS 2.0-2.3 Software Configuration Guide
OL-13493-04
Chapter 4 Configuring Devices
Configuring the Service Engine
Step 3 Click Submit to save the settings.
Explicit Congestion Notification
The TCP Explicit Congestion Notification (ECN) feature allows an intermediate router to notify the end
hosts of impending network congestion. It also provides enhanced support for TCP sessions associated
with applications that are sensitive to delay or packet loss, including Telnet, web browsing, and transfer
of audio and video data. The major issue with ECN is the need to change the operation of both the routers
and the TCP software stacks to accommodate the operation of ECN.
Congestion Windows
The congestion window (cwnd) is a TCP state variable that limits the amount of data that a TCP sender
can transmit onto the network before receiving an acknowledgment (ACK) from the receiving side of the
TCP transmission. The TCP cwnd variable is implemented by the TCP congestion avoidance algorithm.
The goal of the congestion avoidance algorithm is to continually modify the sending rate so that the
sender automatically senses any increase or decrease in available network capacity during the entire data
flow. When congestion occurs (manifested as packet loss), the sending rate is first lowered and then
gradually increased as the sender continues to probe the network for additional capacity.
Retransmit Time Multiplier
The TCP sender uses a timer to measure the time that has elapsed between sending a data segment and
receiving the corresponding ACK from the receiving side of the TCP transmission. When this retransmit
timer expires, the sender (according to the RFC standards for TCP congestion control) must reduce its
sending rate. However, because the sender is not reducing its sending rate in response to network
congestion, the sender is not able to make any valid assumptions about the current state of the network.
Therefore, in order to avoid congesting the network with an inappropriately large burst of data, the
sender implements the slow start algorithm, which reduces the sending rate to one segment per
transmission. (See the “TCP Slow Start” section on page 4-65.)
You can modify the sender’s retransmit timer by using the Retransmit Time Multiplier field in the CDSM
or the tcp increase-xmit-timer-value global configuration command in the CLI. The retransmit time
multiplier modifies the length of the retransmit timer by one to three times the base value, as determined
by the TCP algorithm that is being used for congestion control.
When making adjustments to the retransmit timer, be aware that they affect performance and efficiency.
If the retransmit timer is triggered too early, the sender pushes duplicate data onto the network
unnecessarily; if the timer is triggered too slowly, the sender remains idle for too long, unnecessarily
slowing data flow.
TCP Slow Start
Slow start is one of four congestion-control algorithms used by TCP. The slow start algorithm controls
the amount of data being inserted into the network at the beginning of a TCP session, when the capacity
of the network is not known.
For example, if a TCP session began by inserting a large amount of data into the network, much of the
initial burst of data would probably be lost. Instead, TCP initially transmits a modest amount of data that
has a high probability of successful transmission. TCP then probes the network by sending increasing
amounts of data as long as the network does not show signs of congestion.
The slow start algorithm begins by sending packets at a rate that is determined by the congestion window,
or cwnd variable. (See the “Congestion Windows” section on page 4-65.) The algorithm continues to
increase the sending rate until it reaches the limit set by the slow start threshold (ssthresh) variable.