Your Dad's SLAs And The Cloud

I chatted recently with a number of vendors about service level agreements (SLAs) and cloud computing. Perhaps the quote from these conversations that resonated with me most was, "your dad's SLAs won't cover the use cases for cloud." And, it's true. I joke around that cloud computing is just renamed Internet computing or network computing, but when you start adding in the factors surrounding cloud, it's pretty obvious that Dad's role in cloud SLAs is as an advisor, not implementor.Think about it. Serious applications in a cloud environment have to deal with both limitations and expectations. One example of a limitation is that today's broadband capabilities, at least in the US, aren't scaled to handle the realities of plentiful storage. You can buy a terrabyte for $100, for crying out loud! A terrabyte doesn't move so well, even on a T3. There are a lot of tricks to throw at that, including compression, WAN optimization, and so on, but a quick look at Amazon's AWS Import/Export service reveals the realities of how you transfer this stuff. You FedEx a hard drive unless you want to wait 3-13 days.

But of course, it's not all about bandwidth, either. I had a discussion with Matt Stevens, CTO of Apparent Networks, a cloud performance management vendor. We talked about end to end performance regarding geography and latency. He said, "it goes back to the speed of light: distance, latency, and data loss become very, very important when you don't have your own data center. You can throw infinite bandwidth at it, but the speed of light kicks in, and you're going to hit a physical speed of light limitation that's far lower than you think you've provisioned into your network."

Imad Mouline, CTO of Gomez, another web performance management vendor, points out that expectations are also important. "If what you're interested in is the elasticity of cloud, shouldn't your SLA cover that?" Absolutely. The ability to scale up without massive capex can be a key driver in the decision to use cloud computing facilities. If you're not getting it, and that's why you adopted cloud, we call that #massivefail.

I have written elsewhere of my dislike for one-size-fits-all metrics that don't represent the goals of the project or service. That dislike goes double for cloud.

So the quest, then, is to take an approach that links the desired outcomes and limiting factors of cloud with successful IT service strategies. The core principles of IT service success -- stewardship, customer service, partnership, fiscal responsibility, risk management -- remain the same. How to bring these principles to bear and still stay relevant to the emerging technology is the subject of our February 15 report on cloud SLAs. Stay tuned.