Microsoft is blaming a routine OS upgrade for an outage that hit its Windows Azure cloud-computing infrastructure over the weekend.
In a post on its Windows Azure blog, Microsoft said after the upgrade "the deployment service within Windows Azure began to slow down due to networking issues" on Friday. "This caused a large number of servers to time out and fail," which brought some applications down, according to the post.
Microsoft said the Fabric Controller, a feature built into Azure that manages network resources and performs functions such as load balancing, automatically began taking steps to recover applications that were affected and move them to different servers. Only applications that were running in a single instance on the network went down during the outage, the company said.
"Very few applications running multiple instances went down, although some were degraded due to one instance being down," according to the post.
The ability to perform management tasks from Azure's Web portal was also unavailable for many applications during the outage, which lasted between about 10:30 p.m. Pacific Time Friday and 8:30 p.m. Pacific Time Saturday.
Microsoft said it is refining and tuning Azure's recovery algorithm so when malfunctions occur in the future they will be handled "quickly and gracefully," it said in the post.
The company also recommends that people running applications in Azure deploy them in multiple instances, and will make two the "default in our project templates and samples," according to the post.
"We will not count the second instance against quota limits, so [users] can feel comfortable running two instances of each application role," Microsoft said.
Microsoft confirmed late Monday that Azure users suffered an overnight outage over the weekend during which their applications weren't available.
Currently only a test release of Azure is available, and some early adopters are running applications on it. Users can't expect an early test release of a product to run smoothly without hiccups. However, Azure is a proving ground for how well Microsoft will be able to support the development and deployment of hosted enterprise applications, for which even a short amount of downtime can pose a big problem.
Moreover, last week both Google and Microsoft had outages on their Gmail and Hotmail e-mail services. Outages raise questions about the ability of these companies and other online service providers to maintain a consistent quality of service for end users over the long term.
Microsoft unveiled Azure at its Professional Developers Conference (PDC) in Los Angeles in October and according to public comments made by CEO Steve Ballmer last month, plans to make the infrastructure generally available by November at this year's PDC.