In order to assess the quality of our resource allocation scheme, we compared it with two other, comparable protocols. Figures 6 and 5 depict connection acceptance rates as functions of increasing connection arrival rates for varying numbers of destinations per connection. In each figure, we compare the connection acceptance rate of the Three-Pass protocol proposed in this work with the traditional Tenet Suite 2 [13] establishment protocol, and the Dynamic Resource Migration protocol [5] (DRM).
As described earlier, the traditional Tenet Scheme 2 protocol is a two-pass establishment that does not perform resource relaxation. DRM is a one-pass establishment protocol that relies on a dynamic local balancing scheme to adapt the resource allocations dynamically to the changing load in the network. As to be expected, the three-pass protocol performs better than the basic Suite 2 protocol, in particular for situations with high load, either because of high arrival rates or a large number of destinations. This is because Suite 2 does not perform relaxation, which quickly saturates high priority queues, and therefore wastes resources. On the other side, DRM consistently performs slightly better than the three-pass protocol. When a new connection is established, DRM uses load balancing to relax the resource allocations for that connection. In addition, DRM adapts the resource allocations dynamically to reflect the changing load in the system as new connections are established or existing connections are torn down. We believe that the increase in performance of DRM over the three-pass algorithm is due mostly to the dynamic behaviors of DRM. However, the data collected in Figures 6 and 5 indicates that the small increase in performance may not warrant the additional complexity of running a dynamic resource reallocation scheme.