On 16/12/2015 14:08, Donald Sharp wrote:
> We tested this code with 300 peers and the cpu for bgp was under 5%.  We
> don't have access to low end devices, although we do run on some slower
> power-pc's.  Have either of you actually tested this and we have actual
> data to back this up?

haven't measured it, but the problems we previously ran into (i.e. timeouts
on 60/180 in some situations) were sufficiently bad to make me believe that
dropping the default by a factor of 20 is probably not a good idea in the
general case, at least as long as the bgp rde is running on the same o/s
thread as the keepalive system.

If you're committed to this change, it might be better to phase this in by
implementing a config file "version" token and using that to maintain
previous defaults as explicit config commands.  Otherwise, as Martin said,
"apt-get upgrade; service quagga restart" is likely to cause problems for
the vast majority of people who don't read release notes.

Nick


_______________________________________________
Quagga-dev mailing list
[email protected]
https://lists.quagga.net/mailman/listinfo/quagga-dev

Reply via email to