John Linton One of the major costs of operating an internet service is the cost of connectivity from your PoP(s) to the internet. While this cost isn't anything like as large as the cost of the connection between the end user and your PoP(s) it is the second largest, single, cost of operating a small ISP business.
There are relatively few providers of IP connectivity bandwidth and an even smaller number of actual 'cables' connecting Australia to the rest of the world. So when it comes to 'negotiating' with the relatively few providers there are the usual difficulties of the 'cartel' type pricing situations.
Peering Points (such as those set up in WA and Victoria and the commercial ones set up by PIPE in every mainland capital city) have enabled the cost of delivering traffic from Australian locations (particularly those in the same city as the end user) to be delivered very, very inexpensively. Over the past ten years the amount of this traffic, as a proportion of total traffic, has slowly, but significantly, increased which has delivered some significant cost savings to small and medium, or even large, ISPs.
This increase has been driven by many things but three, in my opinion, have made the most difference. These are:
1) The wide use of P2P in Australia where downloads are sourced from other Australian users rather than international web sites
2) The ever increasing number of Australian based games sites (scaled back now by the popularity of games such as WoW)
3) The delivery of major software 'titles' from peer points set up in Australia like the Akamai service in Sydney hosted by PIPE.
Almost 20% of Exetel's total connectivity traffic is now sourced, at peak times, from PIPE with the peak last week reaching over 350 mbps out of a total theoretical bandwidth provisioning of 1.75 mbps. This has increased from less than 3% when we first put in a PIPE connection some two years ago.
The average cost, at the moment, of international connectivity bandwidth is around $240 per mbps compared to the PIPE cost (to Exetel) of around $10 per mbps. A very significant cost reduction contribution. However the very major portion of the PIPE traffic is P2P with legitimate software downloads from Akamai and games traffic being less than 25% of the total derived from PIPE.
Exetel has decided to trial a P2P cache solution which will begin sometime in November with the objective of providing a further 400 mbps of locally sourced data during peak times. This has, of course, yet to be proven and it may well not succeed. If it does the cost of that '400 mbs' will be around $100 per mbps; far more costly than PIPE but very significantly lower cost than the bandwidth we currently buy from Optus and Verizon - and even lower cost than the likely cost of bandwidth purchased directly from Southern Cross - in the event that proves feasible.
It's an interesting set of difficult to conjugate scenarios as we begin the boringly predictable, but nevertheless long drawn out irrespective of the predictability of the result, processes of negotiating for extentions of current IP bandwidth contracts or seek to put in place new IP contracts with different providers.
However the PIPE experience, the experience we now have with the processes of restricting P2P at various times and the as yet unknown results of the P2P caching trials means that it is no longer sufficient to seek the lowest price of IP connectivity (and ensure its quality) as the only method of delivering an internet service. It's far more complicated.
The overall goal remains to reduce the cost of IP connectivity to around $120.00 per mbps which is going to take some doing but is going to be essential for the planned growth of the business and, if you want to look at the most pessimistic downside - simple survival as a commercial entity.