John Linton
I was filling out the quarterly statistics report imposed on us by the ABS earlier this morning and some of the ratios I was filling in looked very strange. As it's, yet another, bureaucracy imposed and, to me, completely meaningless chore I have never paid much attention to providing this information. However the current figures prompted me to look up the reports from previous quarters which I had kept in a table form but had never bothered to look at the table in any detail before.
It had been obvious over the last 12 months that the demand by DSL users for increased downloads had accelerated and had caused Exetel, as well as many/most other independent ISPs, to either increase their plan prices or to reduce their included data allowances - or in some cases both.
When I re-worked the format of the table a very stark 'picture' emerged. In the quarter ending 30th September 2004 the average bandwidth usage at the peak of the heaviest usage day per customer was 18 mbps; in other words Exetel needed 1 mps of ingress/egress and customer connection bandwidth per 50 customers. I then added to the table the cost of bandwidth per mbps over the same period so as not to distort the cost impact.
This usage increased to 22 mbps in the quarter ending 30th September 2005 and increased again to 25 mbps in the quarter ending 2006. The increases were significant in that they represented around a 20% increase in bandwidth usage per average customer per year. The cost of ingress/egress bandwidth fell by around 18% in the same timeframe so the cost impact was minimal. The customer connection bandwidth didn't fall as much - less than 6% over that period - but the cost per customer of that bandwidth is much less so the financial impact was not that significant.
However in the nine months from 1st October 2006 to 30th June 2007 the average usage per customer at the peak period of the busiest day has increased from 25 mbps to 35 mbps - a 40% leap upward in that period. It's true that the cost per mbps in that period has fallen by around 12% but the cost impact that caused the July 1st 2007 price increases remains only 'line ball' at the current usage levels which will now need to be watched even more closely.
It's likely that ingress/egress, irrespective of what appears in the press recently, will fall by another 10% between now and the end of 2007 which will provide a counter to increased usage and in 2008 the costs of providing will drop sharply if the planned P2P caching is as effective in Australia as it is in other countries.
Irrespective of any positive effects from caching it appears likely that Exetel's current IP bandwidth demands will exceed 2 gigabits per second within the next 6 - 8 months and that in itself will pose some interesting network design issues.
Never a dull day (even when doing the dullest of bureacracy imposed trivial tasks).