John Linton Exetel have been doing a great deal of work in changing the 'network model' with which we created the business and operated it for the past 4 years to a much more complex model that has many more longer term considerations and is far more complex to manage.
Apart from having three major ADSL end user connection providers (Telstra, Optus and Powertel/AAPT) we have always had at least two, and for a short while four 'upstream IP' providers (Optus, Verizon, Powertel/AAPT and PIPE) to which, in a 'logical way', we have recently added two more in Akamai and PeerApp.
Because of the dangers of using one data centre in Sydney we have begun the long and demanding processes of splitting the Sydney based functions over two data centres (the advantage of this was shown last week when one of the Telstra GigE links (the Telstra router) failed three times in a week, which resulted in slowing some 30,000 user's services while Telstra fiddled around denying they had a fault before finally finding they did in fact have a major problem, rather than putting all 30,000 off the air with the previous single GigE link.
The addition of the PoPs in Melbourne and Brisbane in the middle of 2007 and the projected activation of a PoP in Perth later in 2008 will also contine to increase the complexity of how servces are delivered to end users.
Further complications may arise if we do in fact go ahead with the building of our own data centre in the new office space we are in the final stages of negotiating to buy in the Sydney CBD. As this will require a further $A3 million investment by some of Exetel's current shareholders that decision will be very, very carefuly thought through.
For those of you who know the Cisco 'line up' our previous, and still current. network model has been to use Cisco 7300s (of which we now have over 20) as single purpose routers for the different aspects of operating a network (LNS, Border and Core). Using high end Foundry switches we have been able to terminate and control multiple GigE interconnects and cross connects effectively using a single box for each GigE (we use 7600s for the few remaining ATM circuits).
To balance this increasing complexity we will now do the following:
We have now decided to replace 8 of the 7300s used as LNS's with two Cisco 10000s and to replace the 4 core routers in the Clarence Street PoP with a Cisco 6500 or Foundry equivalent.
We will also reduce our major IP providers from 2 to 1 before the end of July 2008.
We will reduce our customer connection carriers from 3 to 2.
We will reduce our SHDSL/Ethernet providers from 3 to 2.
Given that we have just completed more changes to our network in the last four months than we did in the whole of the previous 45 months this program continues to present us with significant challenges and wil be far from easy.
However, if it's successful, it will provide us with greater simplicity of operation and control than we currently have, including the new complexities we've been forced to add over the past year, and this should allow more power and added reliability over the coming months.
I haven't seen the final budgets for these projects and we will have to evaluate the situation when the numbers are 'solid' as I don't think this is either the best time of my personal life or the best time in Exetel's 'life' for so much money to be committed to the future - effectively we are looking at possibly investing more, by 50%, over the next few months in making Exetel 'better' than we have done in the whole of the past four years.