In contrast to Amazon's infrastructure-as-a-service approach, Microsoft is layering Windows tools and services atop an easy-to-build cloud infrastructure.
But there was a subtext to what Pachikara and other Microsoft spokesmen have been saying lately about cloud computing. I've spoken with Amy Barzdukas, general manager of communications for the Server and Tools Group; Burley Kawasaki, senior director of product management for SQL Azure; and Mark Hindsbo, VP of developer and platform evangelism.
There's a more explicit comparison between Azure and Amazon Web Service's EC2 afoot these days with these Microsoft executives. Amazon is the market leader in cloud infrastructure as a service, and Microsoft freely concedes the point. "Amazon is a fantastic supplier of infrastructure as a service," said Pachikara. But he quickly adds, Microsoft will let you concentrate on your data and application in the cloud and handle the infrastructure issues for you. In other words, Microsoft is focused on platform as a service, where familiar tools are available for building something to run in a pre-defined, Windows environment. No horsing around with the infrastructure needed.
In case the message wasn’t clear in the Outback example, Pachikara added, “You’d have to do more management and maintenance of the infrastructure on your own in Amazon.”
In another vein, Microsoft is quietly saying it’s a master of cloud infrastructure, the same as Amazon, but offering higher-level services. Microsoft will be able to keep Azure charges down because it’s figured out the most efficient way to run Azure data centers. One implication of designing a cloud data center to tolerate hardware failures is that maintenance of the data center is no longer a 24-hour operation, with someone constantly on call.
“If a disk fails, it stays failed until the next regular maintenance hours. We don’t do 24-hour maintenance. We do eight-hour maintenance,” Hindsbo said at the conference. A core principle of cloud operation is that a hardware failure where one copy of a data set is located triggers an automatic replication to another server or disk. Even with a hardware failure, there are always two, and after a few seconds, three copies of the data in existence.
IT Service Management Must EvolveThe idea of technology being delivered as a service appeals to the 409 IT pros responding to our Service-Oriented IT Survey. But cloud providers are competing for that work, and CIOs are being selective.