
Sorry, that's what comes from pasting straight from excel:)
hopefully this version is more legible.
Mike
I know this is just one datapoint, but here are some results for the
comparison of the two PLSN algorithms for a real world network having a few
asymmetric cost links (14 out of a total of 340).
The numbers give the number of source destination pairs which are subject
to looping for each of the links which generate a nonzero number of loops.
There are 187 nodes in this network, and hence 34782 possible source
destination pairs.
As you can see, the results for the symmetric algorithm are (as expected)
much better than the asymmetric, even though there will be some loops extra
loops caused by the use of the "wrong" algorithm. But note that in no case
is the symmetric algorithm worse than the asymmetric.
Clearly a network with more asymmetric links (or perhaps links having more
asymmetry, or both) would be more likely to give a result in favour of the
asymmetric algorithm. But this network is fairly typical of the real world
collection I have.
asym sym
4920 0
1214 0
1141 1141
1020 0
516 0
472 0
452 352
426 0
390 0
327 0
183 85
180 0
175 174
157 138
157 138
157 138
124 17
120 0
110 0
94 15
94 15
89 15
56 20
54 54
50 0
36 0
30 0
28 0
27 8
14 14
7 0
6 0
0 0
0 0
_______________________________________________
Rtgwg mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/rtgwg

