next up previous contents
Next: Building up queue with Up: Buffering and queueing analysis Previous: Buffering and queueing analysis   Contents


No significant queueing

Figure 5.1 shows an example of a connection that delivers flood ping packets without significant queueing. The x-axis reflects the time the ICMP request packets were sent, with the timestamp of the first captured ICMP request packet adjusted to zero. The time is measured in seconds. The y-axis reflects the WRTT and is measured in milliseconds. Each data point is drawn as a circle with the centre indicating the values of the data point. Data points are sorted in ascending order by the send-time of their request packets. Successive data points are connected with dotted lines. Later in this chapter, unless particularly stated, all the graphs will follow the same drawing pattern as depicted above.

Figure 5.1: WRTT plot of connection capable of fast delivery
\includegraphics [angle=-90, width=0.9\textwidth]{eps/058-0357/RTT_vs_SentTime.m2.eps}

The source host in this example was atmmon and the destination host was 209.100.140.47, located somewhere in the United States. A traceroute record for this destination host is available on page [*]. The experiment was done at about 22:10 NZDT on Wednesday 25 February, 1999. A thousand flood ping packets were sent and all of them were echoed back. The plot shows that the WRTT fluctuates between about 195 milliseconds to 255 milliseconds.

Where the measurement traffic plus other cross traffic are of a volume which does not fully utilise the bandwidth of the connection, we can identify the signatures as: (1) no or little packet loss; (2) the WRTT fluctuates between a relatively small range above the minimum WRTT; and (3) the WRTT, although it sometimes increases a little, returns quickly to near the minimum WRTT.


next up previous contents
Next: Building up queue with Up: Buffering and queueing analysis Previous: Buffering and queueing analysis   Contents
Xing Deng
July 1999