BCP for multisite multihoming

[David Conrad]

Kevin,

On May 21, 2007, at 1:26 PM, Kevin Day wrote:
> Having an entire /32 per POP is tremendous overkill.

In terms of?

> And, it's still using just as many routes as if this company could  
> just deaggregate the prefix it received.

But it is more likely those routes will actually be seen.  Given  
existing policies, if your intent is to have a multi-site multihoming  
solution and expect each site to be seen by the largest proportion of  
ASes, your options are somewhat limited.

> I'm also assuming that additional prefixes aren't going to be free,  
> which is going to further add to the operating expenses of v6 over v4.

Yep.  A tradeoff.  Obtaining a single /32 and announcing the  
deaggregates will cost you less, but you run the risk of running into  
filters.  I do not know how to quantify that risk.

> It also means the company isn't able to tailor its space usage to  
> each POP.

Not sure why not.  The fact that you have a /32 doesn't mean you have  
to use all 79,228,162,514,264,337,593,543,950,336 addresses in each  
POP.  Yes, you won't be using an absolutely insane amount of address  
space, but under existing policies, that seems to be acceptable.

> Is one allocation/assignment per site really scalable?

Nope, but that doesn't appear to be a concern any longer.

Rgds,
-drc