I remember trying to affect BGP routing and never being able to. Carriers either don't support it or change their level of support. Never worked out. Seems the best was just to buy what you could afford, the more the merrier, then whittle it down to your top two % traffic carriers when the contract comes up.
On Fri, Nov 13, 2020 at 7:40 AM Dennis Burgess via AF <[email protected]> wrote: > I would suggest maybe two 2 gig commits with 10gig ports uncapped. This > will allow either provider to go down and with good network design you can > use multiple routers to prevent a outage of your equipment. Should be > cheaper than 4 1gig internet connections and better than one 10gig . š > > > > Dennis Burgess > > Mikrotik : Trainer, Network Associate, Routing Engineer, Wireless > Engineer, Traffic Control Engineer, Inter-Networking Engineer, Security > Engineer, Enterprise Wireless Engineer > Hurricane Electric: IPv6 Sage Level > Cambium: ePMP > > Author of "Learn RouterOS- Second Editionā > Link Technologies, Inc -- Mikrotik & WISP Support Services > Office: 314-735-0270 Website: http://www.linktechs.net > Create Wireless Coverageās with www.towercoverage.com > > -----Original Message----- > From: AF <[email protected]> On Behalf Of Adam Moffett > Sent: Thursday, November 12, 2020 7:38 PM > To: [email protected] > Subject: Re: [AFMUG] Multiple carriers > > I like the curling analogy. No perfect load balancing across the Internet. > > You can advertise the supernet out of all carriers and also advertise > individual /24's out of the preferred carrier. BGP will use the longer > subnet mask first and fallback to the supernet. If you have a bunch of > non-contiguous allocations then obviously that's not an option, but if you > have something bigger than a /24 then you can do it. > > I also would note that if Connection "A" goes down, connection "B", "C", > or "D" might need to carry all the load and you may not know which way it > will come in. You can test and see what happens in a planned outage, and > tweak your settings just so, but I don't think you can guarantee that > topology upstream from you didn't change at some point between your testing > and the eventual unplanned outage. If you need 4Gb, then you might just > need two 4Gb pipes to have meaningful redundancy. Depends what you're > willing to live with. There's a saying that "slow is better than down", > but IMO the phone blows up either which way so best to avoid either "slow" > or "down" conditions. > > I'll also note that every time I've ordered a 1gig circuit they gave me > GigE optics. If you order anything bigger, even if it's 1.1Gig then it'll > be a 10Gig interface by necessity. An upgrade then is just a phone call, > whereas the upgrade from 1Gig is a scheduled outage to swap interface cards. > > > On 11/12/2020 7:15 PM, Matt Hoppes wrote: > > Right. Which is not ideal. > > > >> On Nov 12, 2020, at 7:11 PM, Ken Hohhof <[email protected]> wrote: > >> > >> There are smarter people than me here on the topic of BGP, but I > >> believe load balancing via prepends is an inexact science. It's like > >> the guy with the broom in curling, you can influence but not dictate > the outcome. > >> > >> You'd probably have better control advertising each subnet via just > >> one of the upstream providers, but then you lose the advantage of > redundant feeds. > >> > >> -----Original Message----- > >> From: AF <[email protected]> On Behalf Of Matt Hoppes > >> Sent: Thursday, November 12, 2020 5:45 PM > >> To: [email protected] > >> Subject: [AFMUG] Multiple carriers > >> > >> I have a situation where I can buy several 1 gigabit pipes from > >> several top tier carriers relatively inexpensively. > >> Or I can buy one fat pipe from one carrier. > >> > >> Say I need 4 gigabits of bandwidth and have four 1 gigabit pipes from > >> 4 carriers running BGP, is there a best way to load balance these? > >> Just AS pre-pend subsets on each carrier so certain subnets prefer one > over another? > >> Or is there a better way? > >> -- > >> AF mailing list > >> [email protected] > >> http://af.afmug.com/mailman/listinfo/af_af.afmug.com > >> > >> > >> > >> -- > >> AF mailing list > >> [email protected] > >> http://af.afmug.com/mailman/listinfo/af_af.afmug.com > > -- > AF mailing list > [email protected] > http://af.afmug.com/mailman/listinfo/af_af.afmug.com > -- > AF mailing list > [email protected] > http://af.afmug.com/mailman/listinfo/af_af.afmug.com > -- Lewis Bergman 325-439-0533 Cell
-- AF mailing list [email protected] http://af.afmug.com/mailman/listinfo/af_af.afmug.com
