[ppml] Effects of explosive routing table growth on ISP behavior

[Scott Leibrand]

Brian Dickson wrote:
> Scott Leibrand wrote:
>> Small ISPs have several options in addition to multihoming with full 
>> BGP routes or singly homing.  They can multihome with default only, 
>> announcing their route(s) to the world, but doing their outbound TE 
>> by some method other than BGP.  They could accept partial routes plus 
>> default, giving them a lot of the benefit of outbound BGP route 
>> selection without having to take a full table.  Or, they could 
>> actively filter deaggregates, maintaining full BGP reachability 
>> without having to keep up with their competitors' upgrade cycles.
> It's also not a problem limited to the smallest of ISPs.
>
> Having to replace significant portions of one's infrastructure, before 
> that equipment has been amortized per one's business plan, is a very 
> bad thing.
>
> And equipment costs are generally speaking proportional, for any size 
> of ISP.
>
> Which is to say, even medium to large ISPs may find themselves in 
> trouble (in some cases, again) if they have to replace equipment 
> prematurely.
>
> And medium-to-large ISPs are likely not to have the luxury of going 
> single-homed, or of taking partial routes.
>
> (The last round of bankruptcies certainly wasn't limited to small 
> ISPs, for example. Worldcom? PSInet? to name two biggish networks.)

Yes, explosive routing table growth would definitely a problem for 
everyone taking full BGP routes.  However, I think it's a problem that 
can be addressed, if/when necessary, by requiring that everyone announce 
their minimum-allocation-size covering aggregates, so that folks can 
filter out unnecessary deaggregates.  With that as an alternative, I 
don't think routing table growth alone will push anyone into bankruptcy.

-Scott