Figure 18 Number of PHY-Ms needed for NIR2-(N,k) networks.
Related Figures (14)
Figure 1: Basic scheme of the current access network. Figure 2: Example of asymmetric MDF connecting the subscriber lines to one ILEC and fou CLEC carriers. Figure 4: Recursive construction of a Clos network Figure 3: Three stage Clos network. Table 1: Synoptic view of candidate networks-(N,n) for AMDF. Figure 5: Example of operations to rearrange the circuits through x without any interruption by exploiting duplicated paths through y. Figure 7: NIR1 network obtained with recursive construction. Figure 6: Example of output and input divertability for a 4 x 4 basic module. Figure 8: NIR2 network obtained with recursive construction. Figure 11: Example of evolution of the Paull matrix in Conf-NIR1 when rearranging the paths. Figure 13: Example of evolution of the Paull matrix in Conf-NIR2 when rearranging the paths. Figure 16: NIR2-(N,n,n) network with output grouping. Table 2: Cost in crosspoints for networks supporting output grouping; N = n* for recursivel. factorized networks. which can be used in (2) to get Figure 19: Number of PHY-Ms needed for NIR1-(N,k) networks. Figure 20: Ratio between the cost of NIR1 and NIR2 networks. een evaluated according to (7). The last block of lines refers to asymmetric 1etworks, and the relative cost has been obtained by extending the method. »logies discussed in the Sec. 4 for these cases. Note that the 4096 x 6544 size 1as been chosen to approximate a realistic ratio between user ports and car- ‘ier ports in operational MDFs. In all such cases, exploiting output grouping illows to reduce costs relevantly, even if the group size is relatively small. Foz xxample, in a 32768 x 32768 switch, by distributing the outputs in 32 groups o: [024, the final cost is just 82% of the original cost without grouping. Note that shis group granularity is more than enough to support 32 different DSLAMs Table 3: Numerical examples showing the cost reduction due to output grouping in NIR2 networks.
