> That you configure your BGP poorly
To configure exit path selection such that you take optimal exit to destination X vs destination Y you need mechanism to either passively or actively measure it. The entire point I am trying to make here is that sweet spot for such measurements is the edge of the network and not on each host, VM, container etc ....
> proper IGP metric may be given by proper policy based on such knowledge
If you are proposing to redistribute BGP into IGP and apply proper metric to each dst route then I think we already differ a little bit in a perspective how to construct a network.
Of course even if I would carry selective destinations in IGP with properly mapped metrics how would host learn about it when it has to pick the src address from N available in end to end multi homing principle ? Is the assumption again that hosts, VMs, LXCs participate now in the IGP ?
And even if it would be a passive IGP listener how do you encode in any IGP today which src address should be selected such that packets will not be dropped due to uRPF check by the upstream provider ?
Many thx,
RR.
RR.
On Wed, Aug 21, 2019 at 10:59 PM Masataka Ohta <mohta@xxxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
Robert Raszuk wrote:
>> IGP metric is used as route preference,
>
> And that's the issue. Neither IGP metric nor BGP path attributes today
> reflect reality of true network paths.
I'm afraid you are not familiar with BGP policy, where you don't have
to blindly rely on path attributes.
> When I look at my BGP table in NY I get path to Europe via Seattle just
> because such path has 1 AS less in the AS-PATH and neglect the fact that
> there is alternative path just next to it with +1 AS but shorter RTT of 150
> ms.
That you configure your BGP poorly does not mean others can't have
better policy to favor the alternative path.
As is written in the draft:
Note that end to end multihoming works with the separation between
inter domain BGP and intra domain routing protocols, if BGP routers,
based on domain policy, assign external routes preference values
(metric) of intra domain routing protocols.
proper IGP metric may be given by proper policy based on such knowledge
(not carried by BGP) as "with +1 AS but shorter RTT of 150ms".
Masataka Ohta
