Millions of Microsoft users were left unable to access some online services overnight because of a major service failure.
Hotmail, Office 365 and Skydrive were among the services affected.
Microsoft was still analysing the cause of the problem on Friday morning, but said it appeared to be related to the internet’s DNS address system.
Such a major problem is likely to raise questions about the reliability of cloud computing versus local storage.
Especially embarrassing is the temporary loss of Office 365, the company’s alternative to Google’s suite of online apps.
Its service also went offline briefly in mid-August, less than two months after it launched.
The latest disruption is believed to have lasted for around two-and-a-half hours, between 0300 GMT and 0530 GMT.
In a blog, posted at 0649 GMT, Microsoft said: “We have completed propagating our DNS configuration changes around the world, and have restored service for most customers.”
The Domain Name System (DNS) is responsible for translating URL web addresses , such as bbc.co.uk into the internet’s native system of IP addresses, e.g. 184.108.40.206.
Microsoft is not alone in suffering problems with its cloud-based applications. Google Docs was unavailable for a period on Wednesday.
However, the fact that Microsoft’s Office 365 is a paid-for service, with users charged £4 per month, may raise expectations of a more robust setup.
Moving applications from installed software on individual computers, to web-based “software as a service” has been a major trend in computing in recent years.
Such systems are seen as easier to manage, simpler to scale-up and down, and potentially offering more robust security.
But a number of high profile failures have dented confidence in cloud computing.
Among them have been several failures of Amazon’s EC2 – the company’s remote computing service, which allows businesses to hire additional processing power and storage on demand.
The system failed in April 2011, impacting several large websites, including Foursquare and Reddit.
Another period of down time in August affected many of the same websites.
Spread the risk
“There will be an element of confidence shaken,” said Ken Moody, data centre services manager at the Cloud Computing Centre.
Avoiding major cloud problems in future would depend on IT companies’ ability to spread the risk, according to Mr Moody.
“People should look at smaller data centres which are divided up where resilience could be guaranteed,” he told BBC News.
“Our service level agreements are 99.99% because we don’t put everything into one large data centre.”
Mr Moody said that no-one, including its users, knows exactly how Microsoft’s cloud computing systems are structured.
Building future confidence in the platform may depend on sharing more information.
“There’s a requirement for transparency and communication to prospective clients,” he said. (BBC News)