Re: [Lsr] Enhancement related to Area Leader Sub-TLV

[Tony Li <[e-mail]>]

Hi Huaimo,

>     It seems that it is OK for only the leader advertises Area Leader sub-TLV.


I disagree.  Again, without other advertisements, the entire domain is at risk of losing the Area Leader and having an uncomfortable transition.


> After the old/current leader is down, a new leader will be elected and the new leader will advertises Area Leader sub-TLV.


If only one node is advertising the Area Leader sub-TLV, then when the Area Leader fails, some other nodes will be forced to advertise that sub-TLV.  This will have to propagate (through flooding) throughout the domain before a new Area Leader is elected.  Any candidates that lose the election would likely compute and advertise a FT of their own, which would then be subsequently obsoleted. This is a non-trivial waste of computation and LSDB space and bandwidth.  There would also be a considerable churn as intermediate nodes see various different FTs at different times. This would definitely be a stress test of the FT migration mechanisms.

>     After moving priority (one byte) from Area Leader sub-TLV to Dynamic Flooding Sub-TLV, the saving on space when every node advertises a priority is (N-1)*S1 – N, where N is the number of nodes in the area, S1 is the size of Area Leader Sub-TLV. S1 is 4 in IS-IS; S1 is 8 in OSPF.  For N = 1000 and IS-IS, the saving is (1000-1)*4 - 1000 = 2996 (bytes). For N = 1000 and OSPF, the saving is (1000-1)*8 - 1000 = 6992 (bytes).
>     When only one node advertises a priority, (N – 1) bytes more is used.


As Les said, it is undesirable for only one node to advertise area leader candidacy.  In fact, it is preferred that as many capable nodes advertise it as possible.  Thus, there is no savings.

Tony