On Reducing Response Time for VPN Traffic March 2002

A successful migration from traditional private networks (PNs) to virtual private networks (VPNs) depends on the ability of VPNs to offer similar quality of service (QoS) guarantees. To approach this goal, we propose to use priority scheduling based on flow size differentiation at the edge devices of the VPN. We evaluate the shortest elapsed time (SET) scheduling policy and compare it to different scheduling policies for both, high and low variance flow size distributions. The results show that SET is a more suitable policy than the traditional FIFO policy in terms of minimizing the response time. In particular, SET offers significantly lower mean response time for flow sizes with a high variance than flow sizes with a low variance. We also propose an implementation of SET for VPN customer edge (CE) devices.