• EF: 300 Kbps, variable packet size
• BE: 2.0 Mbps, 200 packs/sec, 1000 bytes per packet

Comments:

• Figure 1 compares the average one-way delay of an EF stream when it is run in parallel with a best-effort stream against the measures obtained when EF is the only stream active.
  In both cases the average one-way delay increases linearly with the size of the packet. The increase is considerable hen best-effort traffic is added, the increase is up to 150 msec. This is due to the tx-ring queue (FIFO), whose default lenght is 40 particles, like in this test. The consequence of this is that for large sizes the transmit ring the WFQ sheduling vanishes: for long queues, an EF packets can have a lot of best-effort packets in front, as a consequence delay increases.
• Figure 2 compares average ipdv values with and without best-effort traffic. In the latter case, delay variation linearly increases, while in the former delay variation changes greatly according to the packet size. The curve has no regular pattern. More observation points have to be added for better interpolation, however these results need further investigation.
  In this test the maximum transmission time of a best-effort packet is approximately 8 msec (1000 bytes at 1 Mbps). The maximum average delay variation is up to 7.5 msec, which is approximately the transmission time of a best-effort packet.