For Microsoft, which announced its cloud computing platform this week, it seems like service level agreements don't just mean uptime. That thinking could shake up the industry.Though Microsoft isn't giving any firm details on its plans for service level agreements with Windows Azure and the Azure Services Platform, presentations and discussions with Azure's developers here have dropped plenty of hints.
In a presentation at the company's Professional Developers Conference today, Microsoft's Gianpaolo Carraro, director of SAAS architecture in the company's architecture strategy team, said that SLA's shouldn't just mean meeting uptime promises, though that's the common definition of cloud SLA's these days. Companies will typically offer something like a 99.5% uptime promise, and if that isn't met, customers get some sort of refund.
That just doesn't go far enough, Carraro said. Not only should these uptime promises be available, but instead of the cloud vendor measuring the uptime, customers should be able to have some sort of monitoring meter to understand what the uptime actually was for them, and then reconcile that with the cloud vendor's measure. That may be challenging, since it will be difficult to determine whether it was the cloud vendor, the network provider, or the customer who was actually at fault.
But that's just the start. He added that cloud SLA's might also include other optional features, such as promises that certain types of transactions will take a certain length of time, management APIs, programmatic access to the health model of a service, the choice to make applications use firewall friendly protocols, the ability to pause or stop an application or a piece of one from running on the fly, and the ability to do things like trigger back-up of data at certain points in time.
In addition, one of Azure's developers said in an interview, Microsoft is considering SLA's that would enable companies to do things like choose the geographic location where they want their applications to run, though that would likely be expensive.
These shifts could have profound implications for the services industry, since many vendors today resist setting up even uptime-based SLA's for cloud-based applications because many simply can't deliver on their promises. However, offering SLA's that don't hinge only on uptime should actually be more of a win for service providers, since they hedge problems with uptime with more controllable options. In the long run, offering a plethora of cloud SLA's should even entice hesitant companies to move to the cloud.