Accounting in peered CDNs: Initial thoughts ....

[Oliver Spatscheck <spatsch@research.att.com>]

Abhi,

	great summary. One thing we always hear from content provider is the
request of stats for demographics (at least location if possible
more). If we therefore, envision a world in which the entire content of
a site would be delivered using CDN peering being able to aggregate
client demographics would be one requirement. However, I recognize
the complexity of that problem and I am not sure if we want to address
it in the first version.

Oliver

Abhi Deshmukh writes:
 > A few initial thoughts on accounting in peered CDN systems.
 > 
 > 1. Accounting information in CDN Peering would be useful for Monitoring
 > Applications, Billing applications, Reporting and Auditing apps and other
 > apps (like the ones not mentioned in the above list). Accounting information
 > includes usage related information and details of other work done. Based on
 > the discussions at the BoF, the settlement and payment issues are outside
 > the scope of issues being considered.
 > 
 > 
 > 2. Based on the (draft-green-cdnp-gen-arch-02.txt) CDN Peering Architectural
 > Overview draft, it is necessary for the Accounting CPGs to have interfaces
 > with 
 > 	(A) one or more peered Accounting-CPG, 
 > 	(B) one or more origin accounting system, 
 > 	(C) one or more external billing app, and 
 > 	(D) it's own internal accounting system. 
 > 
 > 3. It is possible that the internal accounting system collects information
 > from different CDN elements like Surrogates, Request Routing system, etc and
 > optionally aggregates them. Should we be talking about intra-CDN accounting
 > issues ? I think, we may mention them as needed.
 > 
 > 4. Specifically, inter-CDN accounting issues should be looked at in detail.
 > Do Intra-CDN accounting practices influence the accounting CPG ? This needs
 > to be investigated. 
 > 
 > 5. Certainly, transport of accounting records between accounting CPGs is an
 > important issue. It should be possible to transport different kinds of
 > accounting records using the same transport mechanism. There are issues of
 > pull, push and latency (near-real-time and batch) while transporting
 > accounting records. A candidate transport mechanism must be simple,
 > efficient, and be capable of handling multiple details record formats, etc. 
 > 
 > 6. Work being done in the AAA group including the Diameter spec should be
 > studied to see if it is applicable.
 >                     
 > Your responses, concerns and questions are greatly appreciated.
 > 
 > - Abhi
 >