Here are seven of the most pertinent issues you should address with your cloud computing SLA and master service agreements, according to a recent discussion between Winston Bumpus of DMTF, John Pereira of Intel and ODCA, and Mark Thiele of Switch Inc.
Start Basic
Rather than come at cloud with a specific and developed environment, start small and basic, says Pereira. This will cut down on the complexity of your SLA from the start and enable you to transfer data loads or grow them organically. There’s always more cloud.
Build from Key Drivers
Refine your comparison of different cloud providers and SLAs through five generic master service drivers. The OCDA suggests the umbrella categories of: cost and service, including availability and multitenancy; trust and sustainability, including privacy, portability and IP; data, including ownership and record retention; issue management, including indemnity, monitoring and liability; and baseline contract details such as governing law, divestiture and variations to terms.
Start with an Exit Strategy
With the mindset that data and its requirements and volumes inevitably change, you can head off issues with vendor lock-in if your SLA and provider state a desire to be able to move off this cloud deployment from the start. This is obvious with project-based or planned deployments, but Thiele says that even event-based deployments should spell out an exit. “In being agile, realistically, one of the biggest things you can try to be successful with is to get out of the cloud you’ve just bought in to,” Thiele says.
Consider Lowest Availability
Issues with availability can span a number of technological, networking or architecture problems. With that in mind, Thiele recommends finding the “lowest common denominator” with cloud availability across aspects in the SLA for a realistic view of access to the deployment by your users and customers.
What Shouldn’t You Expect?
For both vendors and buyers, there are certainly areas of the cloud that move beyond your control. To find those issues most vital to your interest and expectations of the cloud on the user side, Pereira suggested finding what dictates your availability, especially when it comes to applications of the cloud that directly reach customers, who can be most turned off by downtime. For sellers, there should be a realistic effort with promises of uptime across independent but associated technologies in the larger cloud offering.
Validation and Indemnification
Validating service is no easy task, though can be better defined from the start by working along the lines of usage models provided by standards organizations. If and when service is impacted, IT and business should measure the importance and true cost effects if a deployment is out of reach or otherwise corrupted. For example, credits may work fine with back up data, though a rebate may not cover the actual losses from dismayed customers.
When to Break it Off
When a deployment first starts up or if use is going well, “no one looks at contracts,” Pereira says. That’s why it’s vital to outline the service expectations and consequences right from the start. This way, if you want to drop a cloud vendor or even just add on elsewhere, the issue has been raised as part of both your SLA as well as your internal data plans.
For more on cloud computing ...
Click here for, Cloud Computing Exchange, our page dedicated to cloud strategies, trends and news. For our weekly newsletter dedicated to the cloud, register here. For a replay of this hour-long discussion, click here. All photos used with permission from Thinkstock.