Re: [rrg] Arguments in favour of Core-Edge Elimination vs. Separation?

[Patrick Frejborg <[e-mail]>]

Hi Noel,

On Tue, Jan 26, 2010 at 6:34 AM, Noel Chiappa <[e-mail]> wrote:
>   > From: Patrick Frejborg <[e-mail]>
>
> I should first point out that my original response was in response to, and in
> the context of, your comment:
>
>    >>> I believe that the Trojan Horse is called MPTCP
>
> and my observations were specifically questions about MPTCP (and let me
> repeat by reference my previous comments about how MPTCP is great for its
> core designed functionality, and so I hope it is a success, but it does not
> solve all problems).
>

Very true, MPTCP alone doesn't solve all problems - we need a bunch of
tools to solve this issue, including LISP (or similar)

>
>    > I think an enterprise should have PI-addresses always, PA-addresses is
>    > for residential users.
>
> OK, I'll bite. I couldn't quickly find data on the number of companies in the
> world, but there are something like 2.5 million companies with 5 or more
> employees in the US. Since the EU has more people than the US, and China and
> India each have more people than that (although are less economically
> developed), there are probably something like 10 million non-trivial companies
> worldwide.  Are you really propose dumping 10 million PI entries into the DFZ?
>

Ah, nope - not into the DFZ, never.
In the future core-edge split architecture the enterprise can have
PI-addresses but in the current architecture the usage of PI addresses
should be minimized. If the future core-edge split architecture
encourages the usage of PI addresses, that would be a carrot for the
enterprises to migrate, wouldn't it?

>
>    > [is] a CES solution is only aimed for multi-homed solutions only?
>
> Depends on the particular CES solution, I would think.
>
>
>    > the problem is, when is a site becoming a LCP site, it is event driven,
>    > isn't it? Anytime there is something happening that passes the news
>    > threshold some servers are starting to get hits, depending upon the
>    > nature of the news.
>
> Sure, and sometimes a site melts down because it doesn't have enough server
> capacity, or enough bandwidth, or whatever. Sites not coping well with massive
> increases in traffic volumes is, I would imagine, not uncommon, and it
> manifests itself in a number of ways - some of which are not easily
> ameliorated (e.g. going from a single-server system to a server cluster).
>

True, but if we are aware of this issue in the design phase we should
find a way around it, at least try - otherwise we are not doing a
proper job.

-- patte

>
>    > the returning traffic doesn't need be looked up by a mapping solution,
>    > it is populated by the initiating traffic
>
> The simplistic versions of this tactic can lead to DoS attacks and/or traffic
> hijacking. If you want to avoid a lookup in the return direction, you have to
> authenticate 'unsolicited' bindings.
>
>        Noel
>