I use Microtik CCR1009 as my core router now.
I currently have 1 ISP upstream link, so the BGP is done inside this core router.
Im getting 2 more ISP upstreams this month end.
Just wanted to know, if the same core router can handle 3 ISP BGP inside it and run smooth or not ?
We have used the CCR1009 for a public feed with some clients and it really doesn’t work well beyond one full IPv4 BGP table. Once we added the second peering, it started to struggle with CPU.
You probably need to look at a CCR1016 or 1036 to add two more feeds.
Contacted Microtik Support and they said this,
“Any CCR series routers can handle 3 full feeds with no issues.”
I would trust anything that IPANetEngineer says - he’s an expert and he works with advanced configurations in real-world deployments.
As long as BGP table scans are still performed on only one CPU core (RouterOS 7 anyone?), I would advise against using more than one BGP full feed on any tilera-based piece of hardware. A CCR1016 barely finishes the BGP table scan of two BGP full feeds before having to start the next one, so personally I wouldn’t go beyond 1 full BGP feed + peerings where applicable. It makes zero difference if you’re using a 1009, 1016, 1036 or even 1072 because they all only add more CPU cores, but don’t make the individual cores perform faster.
Oops!
So you say to use 3 dedicated CCR1009 for each ISP upstream link ?
Then all 3 converges to one core router ??
And then it goes to NAS router which has radius and PPPOE everything happening inside ?
Do you have a need to take full BGP tables from your ISP’s? If all you are looking to do is advertise your networks for upstream reliability, then you could take all three into one CCR 1009 without injected routes (only default) from them.
Or you could write some inbound filters to split the tables across the three. There are multiple work arounds. Though as mentioned above, going beyond 1 full table on the 1009 is really not advised.
Okay Im not very well versed in BGP.
Here is what we do.
We get our Dedicated Leased Enterprise Drop at our office from the upstream provider.
They give us a /28 for WAN and LAN IP.
And I just send them my IP blocks and ask them to advertise.
They advertise it and I simply added them into my CCR1009 and it worked perfect.
May be you will understand if you see my screenshot.
Screen Shot 2016-01-12 at 10.48.30 PM.png
I don’t completely agree with this…we have seen many times where a CCR higher than CCR1009 converges faster on the exact same deployment even though only one core is used for BGP across all of them. It’s possible that the other functions that are handled in multicore (IP forwarding, queuing, etc) process more quickly when more cores are available and speeds up convergence in general across the network. This is based mainly on real world observations since we do work all over the world for service providers on a daily basis.
We recently swapped two CCR1009s as BGP edge peering routers (1 ISP each) with CCR1016s and saw a drastic performance improvement in convergence time across the network. Not sure exactly why since as others have said BGP is tied to one core. Based on that experience which was very repeatable, I won’t recommend a CCR1009 as a peering router for more than one ISP feed + full IPv4 table.
Also, what is often overlooked, is that the speed of handling a full BGP table has as much to do with how fast the ISP will send the table as it does with the speed of your router. Often times if the ISP peer is busy, it may take 5 to 10 mins to take in a full table. Many ISPs can set performance limits on how quickly a table will be sent based on the resources available.
I have one ISP link in my CCR1009 and if a fiber cut happens or anything, my BGP takes like 5 minutes to get back to action.
I hope, when I advertise other 2 ISP links in the same router, its not going to take forever 
And BTW, my core router is only used for BGP and ISP feed.
All the queue, logins, users, DNS, everything happens within the NAS router.
Core router only feeds internet to the NAS router.
Put a default route towards the ISP with a distance of 250 and then as long as you have a default in your table, even if BGP re-converges, you will likely only have a few seconds of disruption while the BGP table is still rebuilding. If you only have one provider, there is usually no need for a full table. Often times, there isn’t much of a need with two ISPs.
Okay!
Did you look at my screenshot ?
Is there any other place, I need to configure something apart from BGP window ?