Average instantaneous packet delay variation

Comments:
  1. Without best-effort traffic delay variation increases rapidly with packet sizes in the range [64, 512] bytes, then it reaches a steady state (figure 4).
    The PB size does not impact this metric since in this scenario the premium source is the only one active.
    The maximum delay variation is of 1.1 msec for 1024 bytes packets.
  2. When best-effort traffic is added, then the metric varies a lot without a regular pattern in the range [0.92, 2.16] msec (figure 5) for any PB size with the exception of PB = 1500 bytes. In such cases the average istantanous delay is limited (see figure 6) for a comparison of the two cases when PB = 1500 bytes).
    For PB = 1500 bytes the average delay variation is well above the other curves, in particular it reaches peaks of up to 4.50 msec. This must be connected to the high latency values experienced when PB = 1500 bytes; this performance problem is under study.
  3. Even in presence of best-effort traffic for each PB size (apart from 1500 bytes) delay variation is almost constant. This is a consequence of the weight assigned to the premium queue which determines the departure rate of premium packets. We expect that premium buffer size can affect one-way delay variation when the weight decreases.
    In this test the premium weight is 90% (default value).
Figure 4: average instantaneous one-way delay variation versus packet size for different premium buffer sizes, without best-effort traffic.
Figure 5: average instantaneous one-way delay variation versus packet size for different premium buffer sizes, with best-effort traffic.
Figure 6: comparison of average instantaneous one-way delay versus packet size for different premium buffer sizes, with and without best-effort traffic.

Last modified: Nov 22, 1999