cisco C7200: Average one-way delay and average ipdv

Average one-way delay and average instantaneous packet delay variation
CISCO 7200

In this test we measure the average one-way delay and average ipdv of an EF stream against different packet sizes and premium queue sizes. The purpose is to verify the effect of classification, marking, policing and scheduling on a single EF class.
EF Stream profile: Stream 1: 300 Kbps, variable packet size
Note: PVC bandwith is 1 Mbps.

Comments:

Average one-way delay is a perfect linear function of the packet size if the maximum rate of the EF class is not exceeded (see figure 1), whatever EF buffer size is deployed. This is expected since no best-effort traffic is present and the EF behaviour aggregate does not experience congestion.
Figure 1 can be compared to the equivalent test results obtained with the IBM 2212: for both platforms one-way delay increases linearly.
When arrival rate < departure rate the ipdv varies within a limited range: [ 51.5, 83.2 ] usec. In figure 2 the ipdv does not show a regular patter. Since the average was computed over less that 10 tests, we think that probably by increasing the population the average would probably follow a more regular pattern.
Figure 2 can be compared to the equivalent test results obtained with router IBM 2212: on the IBM delay variation is more than 10 orders of magnitude larger.

Figure 1: average one-way delay for different packet sizes and different queue sizes.

Figure 2: ipdv for different packet sizes and different EF queue limits. ipdv is almost constant, always below 0.1 msec.

Last modified: Nov 24, 1999