I used to run two full ipv4 + ipv6 tables on a ccr1009. The first pair loaded in about a minute. The second pair took ~6 minutes. The traffic wasn't really an issue at all (I could saturate my 2gbps upstream with three tcp streams) - it was purely the fact that the individual CCR cores aren't really very powerful, and the routing on RouterOS is single-threaded, so losing one BGP upstream meant you were most likely offline for 5-6 minutes while it processed the routing updates to fail over to another upstream. I now run RouterOS in VM's on 2.5GHz L5420's, with two cores allocated per VM (basically one for routing, one for handling packets) and each runs a full feed to it's upstream, plus full iBGP feeds to each of three other VM's. And they will process a full v4+v6 feed fail or recover in about 40 seconds. When routing goes multithreaded, I'll go back to three CCR1009's, but until them I'm sticking with VM's in two separate Citrix clusters. On 21 August 2015 at 12:39, Terry Sweetser (SkyMesh) < terry+mikrotik@skymesh.net.au> wrote:
Hi Mikrotikians,
Any one recently tried BGP with a full table on a CCR 36 core with 1+ Gbps of traffic?
How did go? Or not go?
-- http://about.me/terry.sweetser
_______________________________________________ Public mailing list Public@talk.mikrotik.com.au http://talk.mikrotik.com.au/mailman/listinfo/public_talk.mikrotik.com.au
-- Damien Gardner Jnr VK2TDG. Dip EE. GradIEAust rendrag@rendrag.net - http://www.rendrag.net/ -- We rode on the winds of the rising storm, We ran to the sounds of thunder. We danced among the lightning bolts, and tore the world asunder