|
|
|
Jon Crowcroft, jon@cs.ucl.ac.uk |
|
|
|
|
|
Fairness does not mean everyone pays the same
price - it means |
|
If a subscriber pays more, they can relate the
amount more that they get to the amount less other subscribers get |
|
Competition between providers is visible |
|
Fairness requires network enforcement (vis TCP
games…..even with RED queues) |
|
|
|
|
|
|
Assurance about charges is a double edged sword
too: |
|
You get what you pay for |
|
You don’t give away what wasn’t paid for |
|
Notion of “conservation” of QoS across
boundaries of provisioning - I.e. commutative and associative $/bps |
|
|
|
|
There’s subscribers and providers - subscribers
can be organised hierarchically, as can providers, and providers can peer
too. |
|
There’s the classification of the unit of flow
by source, destination, class of service |
|
or for QoS, by flow specification per unit
charge time and time of day. |
|
|
|
|
The complexity of call logging (and subsequent
log processing) is a function of the granularity of the log (notion of
class of user = me, my workstation, my department, my institute, my
national net, or what)? |
|
Hierarchy can be exploited to reduce complexity
at a point, but assurance and competition require global exposure of the I
logs up and down the hierarchy. |
|
|
|
|
|
If you support QoS, you need to do telephony
style bills unless we can re-educate the accountants |
|
Argues that Class of Service may be better since |
|
lower utilisation is more than offset by the |
|
cheaper billing |
|
But then ISPs already did that to Telcos:-) |
|