John Linton
I don't look at bandwidth usage by customer type as often as I did in Exetel's first two to three years. This has come about partly due to the 'pressures' of too many things to do in a day and partly due to the complexity of the multiple sources of bandwidth 'types' now being used to meet different types of customer's needs. In the 'old days' we simply had IP feeds from Optus and Powertel which were combined on to one MRTG chart and a second MRTG chart showing the Telstra connection to and from our NSW customers (we only provided services to NSW in those days) it was a pretty simple thing to look at two MRTG charts to see whether there was a need approaching for more IP bandwidth.
Today we have PoPs in every State and Territory except Tasmania and multiple sources of IP feeds in all of those PoPs as well as very different types of feeds ranging from Pipe, Akamai, PeerApp, inter-State cross connects and direct IP feeds in to four different PoPs. It's much more difficult to get a 'quick' view of exactly what is happening as the different services now (Ethernet, SHDSL, ADSL1, ADSL2, VoIP and now HSPA) have completely different time 'profiles' and usage patterns per different user type.
It was pretty simple in the 'old days' when IP planning consisted of taking an average usage per customer of 25 kbps, multiplying it by the forecast number of customers at the end of each month and adding 15% as a 'safety buffer' and placing IP and customer connectivity upgrade orders once a month based on that simple arithmetic. It's a lot more complicated now and much easier to get wrong which requires more 'safety buffer' and more inefficiency (cost) because those safety buffers are now required for over thirty different connections instead of three.
The good news is that the average usage of an ADSL1 user hasn't really changed much over the past five years - it's increased from around 25 kbps average per user in peak times to around 35 kbps per ADSL1 user in peak times. This is a percentage increase of 40% which is more than offset by the decline in IP pricing of well over 50% and 'clawed back' by the fact that Telstra's customer connectivity bandwidth pricing has remained the same over the same time period.
Similarly our SHDSL and Ethernet business customers have not increased their average usage by much more than 20% over the last five years and while the customer connectivity costs have remained pretty much the same the IP costs have declined by over 50%.
These small gains in cost/efficiency have been 'swamped' by the growth in percentage terms of the number of customers Exetel now has for its ADSL2 services which are now roughly 25% of all data customers. An ADSL2 customer requires an average of 100 kbps in peak times and that includes restricting P2P for NSW users from 7 pm to midnight. As the percentage of ADSL2 to ADSL1 users keeps slowly increasing this situation will continue to mean that more bandwidth is needed on average each month.
There is little/no doubt that the cost of IP bandwidth will continue to fall over the next 4 quarters as more bandwidth comes 'on stream' from SX, Telstra and Pipe. Exetel is already being offered IP bandwidth at 30% below what we are currently paying. There is also little doubt that the ongoing improvements in the PeerApp and similar products will make caching of various types of traffic more efficient and that the growth in the Akamai clusters will produce more 'free' bandwidth.
Whether these 'improvements' keep pace with the relatively steeply growing average user usage of IP remains to be seen - right now it isn't but that is due to a number of 'one off' factors for us and is probably affecting much bigger service providers in similar ways.
We will look at what we think might happen in much more detail over the coming weeks but one thing I think will result is the reduction in the 'off peak period' from the current 12 hours. Our intention for this period, which has worked out very well over the past almost five years, was to reduce the overall cost of IP traffic by making it attractive to the end user to 'queue' their large downloads to times outside the peak usage periods. This has, largely, been the case and the 'discipline' shown by a large percentage of Exetel's heavier downloading customers in doing this has helped all customers by keeping IP charges very low.
As we have grown the 'discipline' of the early customers (who were almost certainly more knowledgeable on average) has dissipated and the downloading results of less knowledgeable customers have caused the savings curve to flatten for the first time and shows every sign of trending down instead of the slight but consistent upward trend we have 'enjoyed' since March 2004.
In hindsight, a wonderful thing, I should have realised that the 12 hour period was too 'generous' and not done it but, I have to admit it was entirely my mistake - as usual as I was enamoured of the 'marketing' advantage of "half of each day free".
As my maamah used to saaayy 'one more thing to worry about". (apologies to Forrest).