Internet Bulk Billing Blues

Robin Whittle - First Principles Consulting 22 November 1996
This an extract from an article in the August issue of Australian Communications about Optus Vision's HFC network - but the issues apply to all HFC operators who wish to implement a cable modem Internet service. Notes added in November are in [brackets].

This is

Return to where details of these articles are listed.

Difficulties of deciding how to price Internet HFC traffic - and of actually measuring that traffic and calculating the charges

Telephony billing systems have developed as services have evolved and Optus Vision's HFC telephony service presents no special problems, because all billing information is generated by the standard DMS-100 exchange software. However the options for billing for Internet use are far less developed and the enormous range of usages and packet destinations for cable modem services with speeds 0.5 Mbps and beyond raise unique problems.

Internet Service Providers can make reasonable assumptions about traffic when each customer is connected via a 28.8 kbps modem, and charge according to time of day and the quality of service they provide. But how do you charge for access time and/or traffic on a 10 Mbps cable modem? If the residents of a single street simultaneously decided to download large sound, image or video files from the USA, they alone would swamp Australia's trans-Pacific link. This hardly seems like the sort of communications service a carrier would want to sell for $50 a month, but that is apparently what is happening in some early services overseas - for want of a better billing system.

Quite apart from their potentially great capacity, cable modems offer the prospect of permanent TCP/IP connectivity - without tying up telephone lines or bandwidth on the HFC cable. One customer may use their service simply for email, just a few hours a day, while another could run a popular WWW server from home which was generating a large and wildly fluctuating upstream traffic day-in day-out.

With the majority of WWW traffic coming from large servers connected directly to the Internet backbone, the upstream traffic on a [residential] HFC customer access network will generally be less than the downstream. Since the expensive backbone and international TCP/IP links have equal capacity in both directions, the "upstream" capacity - to the Australian backbone and to other countries - is likely to be under-utilised. If the upstream traffic is not the limiting factor driving the carriers' TCP/IP costs, then they may be able to offer relatively inexpensive upstream rates - or to not charge for upstream traffic at all. [Actually I think Telstra does not charge for upstream traffic as long as it is less then downstream traffic.] This would have the added advantage of encouraging local content creation. [Leading to export and import replacement traffic and commerce.] However, the upstream capacity of an HFC system has finite limits - especially during telephony busy hours - and so may be charged for accordingly.

One administratively simple approach is to define a limit on upstream and downstream bandwidth for each customer and charge them an hourly rate for the service - but this would not suit business and enthusiast customers whose traffic may on average be quite low, but who are prepared to pay for permanent connectivity and high bandwidth on demand.

NetComm is planning for its TCP/IP HFC-to-LAN interface products, including the NetRocket 512, to perform packet level billing functions. With this architecture, in which the HFC network operator completely controls the interface's software, it will be possible to perform billing operations without risk of interference by dishonest customers. This removes the computational load from the exchange, and facilitates finely tuned billing algorithms for each customer. Neither Motorola's CyberSURFR nor its exchange interface, the CableRouter, perform billing functions. The company suggests that this can be accomplished in the router at the exchange.

No matter how is done, billing TCP/IP traffic by packet count and/or bytes transacted is a significant computational burden. Something like this must be developed, since flat rate charges for broadband Internet connectivity make as little sense as selling a monthly ticket for arbitrary telephone usage or airline travel.

The HFC operator will face significantly different costs routing packets to destinations within their own network, to the national backbones, to other local Internet Service Providers or to a variety of overseas locations. If customers are purchasing and selling full quality music via electronic delivery, involving a 300 Megabyte for a 74 minute CD- R's worth of music, the operator would probably want to charge according to the destination. [There is likely to be a significant difference in cost for routing 300 Megabytes to or from a local customer or artist compared to an a customer or artist in Berlin or Osaka.]

However, neither the IP address, nor the domain name associated with it directly indicates the geographic or network location of the remote host - such things may only known by the routers at the borders of the operators network. These routers are busy as it is coping with explosive Internet growth even before the advent of cable modems and the probable new task of executing billing algorithms. Similar billing problems will need to be solved before on-demand digital video services can be effectively tariffed.

Discriminatory pricing

There are also questions about what sort of charging scheme would be legal and acceptable to operators and customers. The 1995 Ministerial Direction on carrier associates - covering Optus Vision and Telstra Multimedia - has two special provisions which enable carrier associates to exercise positive or negative price discrimination. See Legal Line, by Peter Waters in the November 1995 edition.

Firstly, they may offer discounts to community, charitable or educational customers. Secondly, and most interestingly, they may discriminate on the basis of the "commercial value to a person of the services provided". This was intended to permit the carrier associate to "capture some of the end value of the content services provided over its network." Thus an operator of a commercial WWW site (or a telephone information service) might be charged more than a domestic customer. The directive also anticipates the differing "values per megabyte" of information - a video on demand operator - whose two hour movie of 5 Gbytes of traffic sells for about $10 - may be charged less than a musician or music retailer whose 0.5 Gbyte two hours of music sells for $20.

Under this directive, I might ask for a standard or discounted tariff for the permanent TCP/IP link of my "Temple of Trance" home WWW server - on the pretext of freely distributing devotional material. With compression and encryption, how would the HFC operator know if I was actually selling the hottest techno dance music and using electronic funds transfer to rake in the splondooli?

[Two more paragraphs that did not make it into the printed article . . .]

Other problems with the idea of "capturing end value" are that customers don't want carriers involved in their business in any way, and that one cable modem could support a WWW server which carries traffic for dozens of independent individuals, organisations and businesses.

Quality of service is going to be another sensitive issue, since the Internet itself is bound to be choked for years to come by the demands of millions of customers with pipes fatter than the Australian backbone was last year. Analogous problems with harbour access might be anticipated if millions of pleasureboat owners upgraded to vessels the size of the QEII.

Copyright 1996 Robin Whittle - First Principles Consulting