Re: Why recalculation from scratch?

[Yasuhiro Ohara <[e-mail]>]

VishwasM> Hi Yasu,

Hi.

VishwasM> The idea is this. When we are under high-congestion, we are
VishwasM> in a state where the topology information on the router is
VishwasM> incomplete/out of date, doing incremental SPF only further
VishwasM> loads the CPU. So instead of doing incremental SPF with
VishwasM> every change, we instead club changes and do the entire SPF
VishwasM> after a longer period of time, which anyway is best
VishwasM> effort(because of databse discrepancies).

OK. I understand that SPFs from scratch in a long interval is more
prefered than incremental SPFs for each changes. But still wondering
if there's any further reason of prefering "SPF from scratch in a long
interval" than "incremental SPF *in a long interval*" ...

VishwasM> Also check ISO10589, it states that the CPU load for two
VishwasM> incremental SPF's can be as much as a single SPF, I dont
VishwasM> remember the section though(probably in the Annex
VishwasM> somewhere). However I  remember conflicting numbers being
VishwasM> presented at NANOG.

Thank you. I will check RFC1195 (it seems the same as ISO10589).

regards.
yasu