The perils of patching in the cloud

Microsoft experienced a handful of embarrassing problems with the latest version of Hotmail, as Ted Samson chronicled in InfoWorld’s Tech Watch blog yesterday. The problems persist to this day. When I read Samson’s summary, I couldn’t help but wonder what went wrong — and how cloud services aren’t supposed to have these problems.

Microsoft took almost two months to roll out the new Hotmail. As Mike Schackwitz explained on the Windows Live blog, Microsoft started rolling out the new version on June 15. “Hotmail servers are grouped into clusters, and we have hundreds of clusters. We start by upgrading a single cluster, wait to make sure that it’s running smoothly and that we didn’t miss anything during testing, and then we continue on to the next cluster. If we find a bug, we’ll fix it before continuing further, and give that fix to everybody who has been upgraded so far.”

Apparently the rollout didn’t go very smoothly. According to the Windows Live blog, Microsoft didn’t switch over the final cluster until August 3, seven weeks later.

Why did it take so long?

It couldn’t have been a hardware problem. After all, these are captive servers — Microsoft owns and operates, preens and prods, and patches them — and presumably many of the best and brightest Microsofties worked around the clock to get the job done. Seven weeks?

What about compatibility problems? Hotmail lives in a comparatively well-defined world. The universe of Web browsers is reasonably finite. Testing changes to a cloud application can’t hold a candle to regression testing Windows, for example, or Office — yet Microsoft pumps out fixes to both on a monthly basis.

More distressing: Cloud applications are supposed to be relatively immune to update problems, aren’t they? They’re certainly nowhere near as difficult as patching the Windows kernel — an activity Microsoft undertakes with alarming frequency these days. We already know about security and privacy and performance issues while working in the cloud. Do we have to worry about versioning and patching problems, too?

Versioning isn’t a simple issue. During that span of seven weeks, for example, plenty of support people were left twisting in the wind. Microsoft rolled out the new Hotmail randomly — there was no way to raise your hand and say, “I need the new version, let me in!” If you were trying to help someone running the new Hotmail, you probably had to log on with their Live ID to get a taste of the new product. Imagine how that would play in your organization if you had a mission-critical application dependent on Hotmail.

Is this the future of cloud computing: roving patches that hit and miss and take seven weeks to stabilize? Where the best solution is to ditch Microsoft’s own product and use one from Google?

Many other problems with Hotmail have surfaced: Attached photos get rotated incorrectly in the delivered message. They’re stored for 30 days on SkyDrive whether you want them there or not. Several people report that Word and Excel documents may or may not open in the cloud, and one person says that a router hang in the middle of opening a document resulted in the message being deleted. Meanwhile, one Windows Live Solution Center thread dealing with new Hotmail complaints mushroomed to roughly 600 entries, many of them unprintably angry, before being locked down.

If Microsoft can’t do the cloud right, who can?

This story, “Beware the perils of patching in the cloud,” was originally published at InfoWorld.com. Get the first word on important tech news with the InfoWorld Tech Watch blog.