Re: Content [Distribution Network] Peering (was Re:Candidatere-charter/new WG)

["Hilarie Orman" <HORMAN@novell.com>]

Perhaps "CDN Access Point" is a useful term for describing what
"vertical peering" provides?  And would "virtual" be more descriptive
than "vertical"?  Or "aggregated"?  This is a vitally important part of
the architecture, so getting the terminology right is worthwhile.

Hilarie


>>> Dan Li <lidan@cisco.com> 11/09/00 12:29PM >>>

>For example, say a cable modem provider (CMP) puts caching proxies in each 
>of several cable head ends. CMP informs their subscribers in each area of 
>their local cache, and all subscribers explicitly configure their 
>browsers. The CMP then creates a peering relationship with several CDNs to 
>have content distributed into these caching proxies, and to provide 
>logging information to those CDNs in return. This means there is 
>distribution and accounting happening between the CMP and the peer CDNs. 
>Therefore, the caching proxies take on a partial role as a "surrogates". 
>So far, CMP's cache deployment has some traits of a CDN. However, there is 
>"null" redirection (the cache in a subscriber's head end is defined as 
>"good" for any content request).  Also, the CMP is not operating a 
>"relatively complex system encompassing multiple points of presence around 
>the world" [draft-day-cdnp-model-02]. Each cable head end is an isolated 
>and tightly-coupled relationship between the subscribers in that area and 
>the cache.

This is the "vertical peering" scenario, meaning a set of caches in one 
administrative domain without full-fledged CDN functionality can 
participate as leaves in a full-fludged CDN of another administrative 
domain. Technically, the protocol for vertical peering may be similar in 
many aspects to that used inside of a CDN. E.g., content preloading, 
invalidation, and accounting.

Dan

At 03:08 PM 11/9/00 -0500, Fred Douglis wrote:
>Phil,
>
>You make a good point: "CDN Peering" may be too specific to cover the 
>scenario
>you've described, unless one is generous with the definition of a CDN.
>
>However, I think "Content Peering" is a poor term, because we're not peering
>"content", we're peering entities that deliver content.  (This is similar to
>the forwarding-versus-routing debate that has hit the cdn mailing list 
>today).
>
>I'm not sure I have a third suggestion that I like more than these two, just
>that I think "Content Peering" is the wrong direction to take this.  For 
>now I
>would be inclined to include your scenario under the CDN framework.
>
>Fred