A power outage brought down Amazon Web Services’ (AWS) Australian cloud operations on Sunday, affecting major streaming sites, news outlets and food delivery services.
AWS listed on its website a number of connectivity issues for its Sydney services. The AWS platform replaces the need for companies to build and maintain their own data centers by providing cloud computing infrastructure – virtual computers for rent and cloud storage space.
Companies affected by outages on Sunday included streaming providers Stan and Foxtel, food delivery service Menulog, Channel Nine television network and various websites belonging to Fairfax Media.
The unconfirmed cause of the outage, which was highlighted at 3:47pm (AEST), is suspected to be the result of a major storm that battered the south eastern Australian coast on Sunday.
The weather caused floods and power outages across the New South Wales region.
Zoning out
Among the affected AWS operations were ElastiCache, Redshift, CloudFormation, CloudHSM and Elastic Beanstalk.
The company’s 12 global geographic regions, each containing data centers, are connected together to minimize the impact of outages. In the Australian region all three of the “availability zones”, or data centers, are sited in Sydney.
After the outage knocked out operations in one availability zone AWS launched new instances of the affected services in other zones, as reported by the Sydney Morning Herald newspaper.
The company said power had been successfully restored to the affected Sydney data center by 5:31pm local time, with work continuing for several hours to re-establish client data connection.
A number of services were still being restored on Monday morning, according to IT News.
Bank customers of Westpac, Commonwealth Bank and ME Bank experienced card failures that prevented them from completing purchases in stores and online, according to the Sydney Morning Herald, but an Amazon spokesperson said these were unrelated to AWS’ difficulties.
In another instance of adverse weather conditions interrupting services in Australia, in January a heatwave caused an outage at the iiNet data center in Perth.
The case for upgrading Australian data centers to full redundancy was made in a DatacenterDynamics article by James Braunegg, managing director of facility operator Micron21, which completed a refit of its Melbourne facility. The upgrade made it the first Uptime Institute certified Tier IV center in the country.
Power outages are however an increasingly unusual source of service downtime. A study published by the Ponemon Institute in January 2016 found that UPS failure is the most common cause, followed by cybercrime, human error and cooling system failure.
The average total cost per minute of unplanned downtime rose from about $8,000 in 2013 to about $9,000 in 2015, the report said. It surveyed operators of 63 data centers in the United States.