Australian businesses have been warned they need to spread the risk in their cloud computing operations across different regions after the Sydney storms on Sunday knocked out the operations of numerous Amazon Web Services customers.
The ferocious storms that hit NSW, left AWS clients including Domino’s Pizza, Foxtel, The Iconic, Stan and Domain without websites or key systems for hours.
It served as a warning that sending systems to the cloud, rather than hosting them on-premise, did not remove the risk of costly failures.
The failure represents a major embarrassment for the company, which generated $US2.57 billion revenue in the latest quarter, based largely on the fact that it is perceived as being hugely reliable.
It meant that customers that had committed all of their systems to its care were unable to trade from mid-afternoon until as late as Monday morning.
On Monday AWS refused to discuss the reasons behind the outage, instead referring inquiries to an online site, which showed the status of its data centres.
A spokesman also declined to comment on whether affected businesses would be entitled to compensation, and what damage the outages had caused for its own business locally.
It is understood that power was lost in a number of systems in its data centre, before being restored about 90 minutes later. It then took hours for some systems to be rebooted and brought back online.
Experts said that customers who insisted all their systems and data remained in Australia had been impacted, as other customers simply had their systems clicked over to Singapore and continued to trade as normal.
IBRS analyst Joe Sweeney said many would use the outage as an example of why businesses should not adopt cloud computing.
However it was more of a wake-up call for businesses to structure their technology in such a way that one data centre failing would not be fatal.
“In short, I think the outage is damaging for the entire market – not just AWS,” Mr Sweeney said.
“I think this event is a great opportunity for organisations to rethink what they were trying to get from moving to the cloud. It is also a cautionary tail – ensure your critical applications are built to survive failures of any data centre.”
The outages came as more than 226,000 homes and businesses lost power during the weekend storms. Roads, bridges and public transport were also affected.
The bodies of three men were also discovered in cars caught in floods in separate incidents in the ACT, the NSW Southern Highlands and Sydney’s south-west.
So far, $30 million worth of insurance claims have been made.
Co-founder of event software company Proxima, Dan Nolan, said his company was an AWS client, but was saved from the outage because it had decided to take a multi-availability zone failover approach.
“When we were setting up our infrastructure a bit over a year ago, we took a punt on multi-availability zone failover,” he said.
“It was literally for if this situation ever occurred … the cost otherwise though would have been quite substantial to the business. Having multi-availability zone is a fraction of the cost to having a site go down.”
Mr Nolan said the company had taken heed of lessons learnt in the United States where businesses such as Instagram had suffered major outages in the past.