A telco analyst firm responsible for a scathing report on Internet access in Australia has defended its claims amid a barrage of criticism.
The Epitiro 2008 Australian Internet Performance Index report claimed Australians receive 65.5 percent of advertised Internet speeds over local TCP connections, and only 53.7 percent of touted ADSL2+ speeds.
Overseas Internet connectivity was worse still, with speeds plummeting to a quarter of those advertised, and ADSL2+ dropping as low as 6.8 percent. Ping times for gamers to servers in the US, the UK and Japan were below acceptable levels. The report concluded Australian broadband may struggle with overseas connectivity.
The company which supplies benchmarking data to telcos around the world, installed “automated reporting agents” in Sydney, Melbourne and Brisbane that logged Internet speeds from October to December last year for new standard customer accounts setup without the knowledge of ISPs. The report covered Web browsing, streaming media, Voice over Internet Protocol (VoIP), and online gaming.
The report gave Telstra the top spot for deliverable Internet speeds, followed by TPG, iiNet, Netspace, and AAPT, while Optus came in eighth spot, behind Westnet.
Local telecommunications analyst Paul Budde said the report should be taken “with a grain of salt” due to variables that may skew results, but conceded broadband speeds in Australia are poor.
“There so many uncertainties that influence these kind of surveys, like geography, and the quality of local exchanges and in-house networks,” Budde said.
“Depending on where you measure it, download speeds could vary a lot, even between suburbs. This is so important because two people on the same service could have hugely different Internet speeds.
“The top five providers in the report would rank about the same across most surveys...If HFC (Hybrid Fibre Coaxial) cable was tested, I would suspect Optus would rank as high as Telstra.”
Epitro Australasia managing director Mike Cranna said enough good data exists to produce accurate results -- despite the network uncertainties -- for a decisive results.
“A very clear picture emerges across different regions and different customer-premises equipment. There is a degree of homogeneity in both networks and customer-premises equipment that means it is possible to see the trends despite these variables,” Cranna said.
“You do not need to measure every endpoint in a network to understand its performance; any network engineer knows this, and it is a fact intrinsic to the concept of a network.
“You do not need to measure e-mail to every household to know the e-mail platform has fallen over! Similarly, to test for kidney failure, you test the kidney; it's not necessary to test every cell in the body.”
Cranna said the fibre networks Budde referred to were not tested because of limited coverage. He said it does measure HFC networks but did not publish the results because the technology is not readily available.
The firm tests data network speeds for synchronisation, cached and non-cached HTTP downloads, DNS, ping and packet loss, and has separate procedures for VoIP and video. Tests are run from laboratories in Sydney, Melbourne and Brisbane. Variables are measured from the same rack and exchange.
Some 3000 agents used to measure speeds are distributed across the country which Cranna said provides for a “very clear picture of the overall performance of ISPs”. Statistics on variables such as customer-premises equipment gathered by the agents can be measured against laboratory results.
Epitiro reported last month that broadband speeds had “small but significant” improvments from Q3 to Q4 last year.