A total of 568 hours of downtime at 13 well-known cloud services since 2007 had an economic impact of more than US$71.7 million dollars, said the International Working Group on Cloud Computing Resiliency (IWGCR) on Monday.
The average unavailability of cloud services is 7.5 hours per year, amounting to an availability rate of 99.9 percent, according to the group’s preliminary results. “It is extremely far from the expected reliability of mission critical system (99.999%). As a comparison, the service average unavailability for electricity in a modern capital is less than 15 minutes per year,” the researchers noted in their paper.
It is the first time that the group, formed in March 2012 by Telecom ParisTech and Paris 13 University, published what it calls the Availability Ranking of World Cloud Computing (ARWC). As cloud services appeal more and more to government agencies and global businesses, it becomes more important that the provided services are reliable, especially when the systems are mission critical, the researchers said. The lack of cloud reliability is not commonly known by the industry, they added.
Their research is based on press reports of cloud outages at services like Twitter, Facebook, Amazon, Microsoft, Google, Yahoo and Paypal, among others.
The costs for an hour-long outage can vary from $89,000 at a travel service provider such as Amadeus, to $225,000 an hour for a service like Paypal, according to the research paper. The figures are based on hourly costs accepted by the industry, the researchers said. Outages at companies like Google, Microsoft and Amazon amount to an estimated $200,000 an hour, according to the group.
Besides the economic impact, the downtime that sometimes can last for days or even a week can affect millions of users.
While the researchers noted that their methodology is imperfect because their information-gathering process was far from exhaustive, they said that the preliminary figures are most likely underestimated. Many outages are not published in the press, leaving a lot of room for missed outages, they said.
There are other caveats in the methodology used, including not having the precise value of the economic cost for each failure or an average hourly cost for each cloud service provider, the researchers said. Besides that, the group noted that its data was not based on the number of users a service has, which would be preferable.
To assess cloud availability better, the group announced plans to adopt new methods for future research.
Loek covers all things tech for the IDG News Service. Follow him on Twitter at @loekessers or email tips and comments to loek_essers@idg.com