BGP peering strategies for smaller routers

Nick Hilliard nick at foobar.org
Tue May 3 22:18:23 UTC 2016


William Herrin wrote:
> I respectfully disagree. Sourcing more ram won't fix the next bit of
> sloppiness with the software. Or the one after that. Once the manager
> of that team starts to accept poor code quality, the only thing with a
> chance of fixing it is strong customer push-back.

You could feasibly argue that the box was shipped with too little RAM
when it was released in 2008/2009.  Alternatively, you could argue that
4G was fine way back then and that it's hardly surprising that it's too
little now because the DFZ has grown from 220k prefixes to 620k.

Anyway, this is all completely academic because the box is end-of-life
and the chances of getting this problem fixed are zero.

So there is simply no point in berating some poor TAC representative
about this problem, end of story.

The replacement box (ASR1001-X) has 8G ram.  That's fine for today.  In
8 years, it probably won't be fine.

> And it is poor code quality. Even slicing and dicing the ram in odd
> ways, there's just no excuse for an order-of-magnitude increase in ram
> required to run the same algorithms on the same data.

If RAM were expensive, your argument would make sense, but RAM is not
expensive.

I don't really think it's any more viable to complain about an 8yo
end-of-life box being ill equipped to handle today's Internet than it is
to complain about the fact that you can't run a desktop operating system
on a computer with 8 megs of ram, like you used to be able to.  Truth
is, 8 megs of RAM then was more expensive than 8 gigs of RAM these days.

Can we move on now?

Nick


More information about the NANOG mailing list