Re: new to cdn internetworking

[Eric Dean <eric@crystalballinc.com>]

I don't disagree, but arguably 50% of all Internet requests are proxied.
I'm not a proponent just stating the facts.  It is also arguable which
direction service providers are heading.  I agree that backbone networks
are certainly not using proxies; however, there are a variety of
subscription-based services using technologies such as L2TP,IPSec, PPOE,
PPOA that are overlays that offer things like content filtering and other
value-adds.  I do agree that caching proxies are not being deployed inline
with giga-routers though.

On Tue, 13 Nov 2001, Christian Kuhtz wrote:

> On Tue, Nov 13, 2001 at 12:14:30PM -0500, Eric Dean wrote:
> > 
> > When running everyone's through forward proxies (which does every business
> > via a firewall and some major ISPs), the proxy becomes a layer7 network
> > device that can do pretty much anything with content requests.
> 
> I've been pretty quiet in this discussion, although I've been
> lurking on this list since the very beginning.
> 
> It might be worthwhile pointing out that several major providers
> have failry recently turned off, removed or otherwise have disabled
> or are in the process of disabling their forward proxy caching
> infrastructures.  Others find that the only way to make their
> deployments work out operationally is with substantial
> contributions by the vendors who sold the gear.
> 
> If there is an implicit assumption here that forward proxies (and
> associated infrastructures) are common place, or easily financially 
> justified or operationalized, or otherwise readily available and
> implementable, or work generally well, that may be considerably
> out of touch with reality.
> 
> By the same token, placing something else into the path of traffic
> is not something providers typically look forward to, and a
> some of the commonly used architectures do not lend themselves to
> such implementations being straightforward at all;  this
> completely excludes the operations view of the world which is an
> additional issue as already hinted at above.
> 
> Typically, doing anything inband is something most providers view
> with a considerable amount of scepticism to say the least.
> For example, most providers never don't or never have liked present
> day caching infrastructures in their networks.  You might say "well,
> they're there now, so, why wouldn't they adopt this"; well, they're
> also in the process of giving up on considerable financial
> investments made to date for a host of reasons because of the
> considerable pain it has and is causing, and lack of ROI.
> 
> It would seem quite a few people appear to think that they can ride
> the way of the forward proxy caching infrastructures into the
> future -- that may be a quite mistaken assumption..
> 
> By the same token, don't bank on being able to change anything low
> in the stack.  This stuff should ride up high in the stack, without
> munching packets in the core of the network.  Such things as a
> content addressing scheme which is an abstraction of network
> topology etc; rather than having inband widgets in the network which
> translate, redirect, rewrite stuff on the fly.
> 
> Various vendors will tell you that they can do all this on the fly
> in their gear in your backbone.  That ASICs are fast enough to do
> all this.  Or that you could place one of their devices in front
> of every edge box.  Bla bla bla.  Pigs can fly.
> 
> It's not hard to figure out why that idea floats about as well as 
> a brick.
> 
> Just thought it might be helpful pointing out these somewhat common
> sense facts after I see this discussion circling again and again
> around these points.
> 
> Hope this was helpful to somebody out there.
> 
> Cheers,
> Chris
> 

Eric Dean
President, Crystal Ball Inc.
W 703-322-8000
F 703-322-8010 
M 703-597-6921