In addition to simplifying nodal processing and network architecture,
a good virtual path layout/capacity design reduces network system cost
s and improves overall QoS (Quality of Service) manageability and syst
em utilization. A shortest-path based layout scheme and a virtual path
capacity allocation scheme are proposed in this paper. For ease of im
plementation, we adopt a deterministic Virtual Path (VP) capacity assi
gnment scheme. Each VP contains multiple Virtual Channels (VCs). The b
andwidth assigned to each VC affects both the VC's QoS and the remaini
ng capacity in the corresponding VP. We have recently proposed a novel
bandwidth allocation scheme, in which bandwidth assignments are made
based on a bandwidth queue model, to maximize the network utilization
and user satisfaction. In this paper, this bandwidth queue model is us
ed to quantify VP costs and a constrained non-linear optimization prob
lem with VP capacities as decision variables is formulated. Since Swit
ched Virtual Connection (SVC) traffic may be routed through different
VPs, and hence affect the optimal VP capacity design, SVC routing para
meters are also included as decision variables and determined together
with the VP capacities.