From: Anthony Pace (anthonypace@fastmail.fm)
Date: Fri Mar 14 2003 - 17:14:37 GMT-3
My question is much simpler. Can I advertise the same route to 2
providers and let the clients closer to provider 1 come in via provider 1
and let the customers closer to provider 2 come in via 2.
Can I get the whole BGP routing table from both providers and choose the
best EGRESS path for any given network between the 2 providers?
In the event of a failure everything will now INGRESS/EGRESS via the only
provider getting my advertisement and sending me the BGP Internet routes?
No prepends, meds or Local preference. Just a filter to make sure I'm not
a transit AS. I can see nothing technically that would prevent this, is
there an administrative policy between NAPS which would prohibit any of
this? There would be asymetrical routing, but I don't see that as a big
issue.
Anthony Pace CCIE #10349
On Fri, 14 Mar 2003 08:46:06 -0500, "OhioHondo"
<ohiohondo@columbus.rr.com> said:
> In addition to controlling load balancing, I prefer to have a
> deterministic
> traffic flow if at all possible. Knowing how the traffic is "supposed to"
> egress the network always helps in troubleshooting.
>
> -----Original Message-----
> From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On Behalf Of
> Voss, David
> Sent: Thursday, March 13, 2003 10:19 PM
> To: 'Anthony Pace'; ccielab@groupstudy.com
> Subject: RE: BGP load balancing (via best path)
>
>
> I have dual circuits at work. I think PREPEND is a very weak method in
> to
> try to dictate traffic mainly because it will not work in every scenario.
> What if you prepend 5 5 5 5 5 to make one circuit less preffered but your
> "customer" is:
>
> CUSTOMER 100 200 300 5 5 5 5 5
> YOUR AS
> CUSTOMER 100 200 300 400 500 600 700 800 9000 1000 1100 1200 1300 1400
> YOUR AS
>
> The customer will still prefer the circuit that you prepend in this case.
>
> I have chosed to advertise one network out a T1 and one network out our
> DS3.
> I then use advertise-map / non-exist-map to advertise the network out the
> other circuit should the router or circuit fail.
>
> The other obvious method is to play with MED's, but I had the luxery of 2
> Class C network to play with.
>
> Regardless, AS-PATH prepend, I believe, is not a 100% reliable way to
> dicate
> traffic.
>
> -----Original Message-----
> From: Anthony Pace [mailto:anthonypace@fastmail.fm]
> Sent: Thursday, March 13, 2003 12:43 PM
> To: ccielab@groupstudy.com
> Subject: BGP load balancing (via best path)
>
>
> All of the implementations I have seen, (as well as the examples in
> Halabi's book) use a Primary/Failover method for both INGRESS and EGRESS
> traffic when BGP is used. Outbound advertisements with AS-PREPEND'S to
> the backup to cause the "world" to come in via the Primary; and perhaps
> better local preference for routes learned on the PRimary to cause our
> outbound traffic to use the primary.
>
> There are many flavors of this,but the principle remains consistant. For
> any given route a primary and a failover.
>
> Here is my question:
>
> What is the detriment of taking in the all BGP routes from 2 providers
> and choosing the best path for outbound traffic; What is the detriment of
> advertising my own address space to 2 providers with no PREPENDS. Just
> let the traffic come in and go out based on the "end systems" proximity
> to either of my 2 providers?
>
> Without AS-PREPENDS to the BACKUP provider, I understand that if the
> PRIMARY provider gave me a block of their address space, I would wind up
> seeing the traffic all come in via the BACKUP link (as the backup
> provider would advertise a more specific, route and the PRIMARY would
> summarize my space into an aggrigate)
>
> Does that sound correct?
>
> If the address space belonged to neither provider, then couldn't I just
> advertise it to both and let the traffic find me through both?
>
> Am I overlooking something very obvious?
>
> Anthony Pace CCIE#10349
>
>
> --
> Anthony Pace
> anthonypace@fastmail.fm
>
> --
> http://www.fastmail.fm - One of many happy users:
> http://www.fastmail.fm/docs/quotes.html
>
>
>
-- Anthony Pace anthonypace@fastmail.fm-- http://www.fastmail.fm - Same, same, but different
This archive was generated by hypermail 2.1.4 : Sat Apr 05 2003 - 08:51:39 GMT-3