John Linton
Some months ago we reached a stage where our original network design/topology which had operated virtually faultlessly for over 4 years began to show its limitations as the number of data and VoIP users continued to increase on a month by month basis and, in the case of VoIP users had begun to grow extremely rapidly. Our network had been built around using gigabit Cisco routers (7300s) and Foundry 24 port and 48 port switches and almost 40 of those devices were in operation in the network which was then shortly to comprise 2 PoPs in Sydney and PoPs in all other mainland capital cities and the ACT (Tasmania was, even then, beginning to slip due to hold ups in getting access to the BassLink connectivity between Victoria and Tasmania.
We therefore made a decision to replace the multiple 7300 LNS, core and border routers with the next level up of Cisco equipment - 10000s for the LNS services and 6500s for the core and border routers. We also made the decision to provide direct IP connections in the inter-State PoPs over the first part of the 2009 financial year. These changes together with the 'simplification' of our external feeds and changes to the integration of the Pipe, WAIX, Akamai and PeerApp caching and sourcing links was, and is, a major set of changes to a live network that supports 3 gigabits of traffic for almost 80,000 users. Naturally any planned changes had to be carefully considered and even more carefully implemented.
It's far beyond my level of knowledge to understand, let alone implement, these sorts of engineering issues - except for the financial issues and the impact on our customers. It's clear to me after two years of using the NetEnforcer P2P management system that there will always be issues with that sort of bandwidth enhancement process. Similarly after almost a year of using the PeerApp P2P caching system it's equally obvious that it is a valuable resource to provide faster downloads for popular downloads and contributes financially as long as IP bandwidth charges remain in excess of $A100 per mbps - which is very true today but may not be true by this time next year.
Our challenge is to provide double today's average bandwidth per user at equal to or hopefully less than today's costs - in other words to achieve a 50% reduction in our overall bandwidth delivery costs. Irrespective of what anyone else in larger ISPs may be able to do, I think this is a very, very difficult objective to achieve despite all of the planning and 'pain' Exetel has gone through to put in place the mechanisms, skills and knowledge to be able to sensibly address these issues. What makes this objective probably more daunting is the very high likelihood that even doubling the average bandwidth per user may not be enough to satisfy the growth in bandwidth demand over the next 12 to 18 months.
This means that as well as doing everything possible to reduce our costs of delivering traffic we have to look for other 'efficiencies'. Providing less traffic per user paid dollar is the reverse side of the reduce the cost of traffic 'coin'. Part of our interest in HSPA is that it will have a totally different demographic of user usage than our current ADSL1 and ADSL2 user base - much lower usage on average per customer. We could add 50,000 HSPA customers at less than one fifth of the infrastructure costs required by another 50,000 ADSL customers and, strangely, make more net profit. Similarly VoIP traffic generates twice to three times the profit that 'data' traffic does. Both of these aspects of service delivery are very important to Exetel in the balance of the current financial year, and in subsequent years, in terms of meeting both our revenue and our profit targets.
We have never made very much money out of providing ADSL services; and we don't have any plans to ever do that as Exetel's raison d'etre was and remains the provision of services at lower costs than any other Australian service provider (if you sell at the lowest cost and if you don't buy at the lowest price, because of your small size, you're never going to make much money). We did figure that after five years (assuming we stayed in business) we would have built a sustainable ability to do that but we did recognize that we needed more than ADSL services (which by definition when we began business we realized would have to be sourced either directly or indirectly from Telstra) and therefore we would have to have in place a plan that would mean that we would buy no services from Telstra over the second five years of Exetel's theoretical existence.
So VoiP, HSPA, Mobile, Free Mobile, SMS via email, fax via email and three other currently planned services are the future 'back bone' of Exetel's business with ADSL still important but not a basis for continuing growth and development. Finding a way to keep Exetel's current 'model' operating at a small profit is very important but the ability to build Exetel's business already no longer relies on services derived directly or indirectly bought from Telstra.
Re-engineering the Exetel network is going to be a difficult set of tasks but it is the 'corner stone' of everything else we have planned.
Making it all work is something no other service provider has ever done - at least as far as I can see.