John Linton
One of the things I never expected to have to deal with, and I'm sure that demonstrates my inexperience and naivety, is being a commercial provider of services to end users that is dependent on third parties for the quality of those services. I realise how stupid that makes me sound/read. For most of the past 24 months we have noticed an increasing number of inexplicable (to us) degraded performances across the country that 'ebb and flow' and can never be tied down to a specific scenario.
Over the last three months of drenching rains in much of Eastern Australia these instances of slow speeds have been somewhat masked by the significantly increased number of "drop outs" and long periods of "service unavailability". Over the last twelve months there is no doubt that Exetel's moves to use P2P filtering and, over the past 3 months, P2P caching may have contributed, slightly, to some strange instances but not really that much - in fact those initiatives have given much more positive than negative results to the overwhelming majority of Exetel broadband users.
What is becoming very clear, both to Exetel and without any doubt all other ISPs and carriers is that, as ADSL2 and higher speed ADSL1 services account for a larger percentage of total network use, the overall demand for bandwidth increases in every section of the network path is exceeding even the most generous allowances. For example - when Exetel connected its first broadband customer in February 2004 we provisioned the bandwidth between the customer and Exetel and between Exetel and the 'outside world' at a ratio of 25 kbps per customer - which had been over the preceding 3 years of providing ADSL1 services via other companies on the generous side. Over the past 4 years that provisioning ratio has risen to around 50 kbps for ADSL1 customers which reflects usage of streaming video and much wider use of P2P.
OK - surprising but understandable and with the fall in pricing of IP bandwidth (no fall in customer connectivity bandwidth pricing) sustainable with only relatively minor end user price increases - the other mitigating factor over that time being the reduction in ADSL1 port rental costs from Telstra Wholesale.
What is clearly not sustainable, at least by the carriers in terms of their back hauls, is the dramatic impact that ADSL2 services have had on those parts of the networks (from the end user via the carrier's network to the ISP). Exetel originally provisioned the parts of its network used for ADSL2 traffic at 50% greater than the parts used for ADSL1 traffic (per user). However over the two years we have been providing ADSL2 services we have increased that to 100% greater than ADSL1. We would be prepared to increase it to, say 200% of that provided to ADSL1 customers if we now believed it would eliminate the constantly shifting (geographically) end user complaints of periods of slow speeds can't play games etc, etc., but we have reached the point where we see no congestion/saturation at any time but there are still issues with packet loss and speed degradation.
While it may well be that there are still issues we need to address that we haven't yet identified I, non-technical person that I am, have reached the end of my ability to believe that this is the case. I wouldn't be surprised if we now found something that we had overlooked but I no longer think it's remotely likely.
The problem is almost certainly summed up in one of the carrier's contracts with a clause that states:
"Carrier" will use its best efforts to maintain bandwidth equivalent to 30 kbps except if other ........:
The other two carriers make no such statement other than to say they don't guarantee anything.
So there is the problem:
We see a need to provision an ADSL2 service at 100 kbps per user and at least one carrier sees it being acceptable to provision it at 30 kbps by contract although they do say they actually "over-provision" it at the moment - by how much they don't say..
I'm not sure what can be done in the immediate term.
Stop selling ADSL2 services?
Seems to be the only way forward.