next up previous
Next: Network Size Up: Experimental Evaluation Previous: OTN and DCN Size

Alternate Path Length

Figure 8: Restoration time v/s alternate path length (OTN size = 100 nodes, DCN size = 480 nodes; OTN edge probability = 0.13 and that of DCN = 0.19).
\begin{figure}\begin{center}
\center \mbox{\psfig{figure=graphs/AltPathLen.ps,width=3.25in,angle=-90}}
\end{center}\end{figure}
In Experiment 2, the OTN and DCN sizes were held constant. The restoration time for all alternate path-lengths was measured over a set of ten random DCN/OTN pairs. The pure-random model was used for generating edges with edge probabilities 0.13 and 0.19 and sizes fixed at 100 and 480 nodes for the OTN and the DCN respectively. Figure 8  shows the performance of each of the four schemes on varying path lengths, and the distribution of path lengths (scaled down ten-fold) in the OTN graphs on which the measurements were performed. As seen from the results, Parallel OBS activation scheme scales better than all the other schemes as path length increases. Similarly, parallel activation also scales well with path length but performs worse than its OBS cousin. Parallel schemes thus benefit from the structure of the DCN whereas serial schemes do not.

Figure 9: Restoration time v/s network size (edge probability of OTN = 0.4 and that of DCN = 0.18).
\begin{figure}\begin{center}
\center \mbox{\psfig{figure=graphs/sizeEQ.ps,width=3.25in,angle=-90}}
\end{center}\end{figure}

next up previous
Next: Network Size Up: Experimental Evaluation Previous: OTN and DCN Size
Swapnil Bhatia 2002-08-02