Abstract

It has been recently observed that the dominant factor on network performance after a failure in a packet switched network is the transient or nonstationary congestion period triggered by the backlog of packets due to retransmission after a failure. Given this observation, we address the survivable network design problem for VP-based ATM networks by introducing the transient time threshold to clear backlog as one of the quality-of-service parameters and present a design framework. Through computational work on test networks, we observe that if this parameter is not incorporated, a network (where services do backlog traffic after a failure) may be under designed resulting in noticeably longer transient period than any acceptable threshold.

To obtain a post-script file of this paper, click here . This PS file is about 238K bytes.