Windows Azure, Xbox Live Experience Problems as Xbox One Launches
November 22nd, 2013 By: Rich Miller
This is a big week for Microsoft. Thursday’s release of the Xbox One gaming system marked the culmination of years of hardware development (including a reported $100 million investment in the controller design) – and months of marketing hype. To boost its focus on the online gaming aspect of Xbox One, Microsoft touted its use 300,000 servers to support the Xbox One network.
But launch day turned out to be a rough ride for the Windows Azure cloud computing service, which helps power Xbox Live. The platform was plagued by problems for much of Thursday, including a storage outage early in the day, followed by performance problems that Microsoft attributed to DNS issues outside of its network.
“We are experiencing an issue with Storage in North Central US,” Microsoft reported on its Azure service dashboard. “Customers may experience intermittent timeouts when accessing storage in North Central US. The number of customers impacted is extremely low.” The problems soon spread to other regions – including the South Central US, North Europe, Southeast Asia, West Europe, East Asia, East US and West US – before being resolved at around 11:30 am UTC time.
The North Central US would suggest the issue may focus on Microsft’s infrastructure in Chicago, where the company built a huge container data center with servers housed inside shipping containers. But don’t blame the containers – the storage in Chicago is housed in a traditional colocation data hall in another part of the Chicago facility.
Shortly afterward, Azure started seeing “an intermittent issue” with networks in the North Europe and West US regions. A Microsoft rep later tweeted that that issue was DNS-related and did not originate on their system.
— Scott Guthrie (@scottgu) November 22, 2013
Nonethelesss, the multi-region issues got the attention of one of a prominent analyst tracking cloud services.
Azure’s cross-region outages means that multi-region deployment isn’t really a useful strategy for higher availability.
— Lydia Leong (@cloudpundit) November 22, 2013