I made the trek to Interop New York yesterday morning, and jumped right into my first breakout session: Measuring Cloud Performance. I know what you’re thinking: More marketing mumbo jumbo on the buzz word that’s lost all meaning. And there were definitely moments of that, especially with a VP of product marketing on the panel, but during the predominantly question-and-answer forum, there were also moments of frustration and frank discussion on the state of the cloud computing industry.
Moderated by Hooman Beheshti, a VP at Strangeloop, the discussion was led by panelists David Link, CEO of ScienceLogic; Randy Bias, CEO of Cloudscaling; Jason Read, founder of CloudHarmony; Russell Rothstein, VP of Product Marketing for OpTier; and Alex Polvi, CEO of CloudKick.
The Hard-Hitting Cloud Questions: What is Cloud Performance?
Beheshti began the conversation with what seems to be the standard jumping off point for any discussion concerning cloud: What the heck is it? The panelists gave a few answers: The difference of app performance when moving to a multi-tenant environment. Paying attention to the performance changes that happen based on shared resources and oversubscription. Understanding the latency that happens between the business and the cloud and the cloud and the customer.
That’s all well and good, but not specific enough for the audience members actively seeking a move to the cloud. The concern that seemed to be irking most of the audience was the lack of visibility into the cloud. One man spoke up: He wants to move his disaster recovery function from a failover site to the cloud, but felt he was getting the runaround from cloud providers. He wanted to know how to get the big picture of service providers’ capabilities to ensure acceptable end-user performance through setting up SLAs.
The general attitude of the panel was that you can’t, really. David Link pointed out that that’s where a lot of service providers fail. Amazon Web Services’ API basically provides a description of their services and configuration data and, for a fee, some performance data. For the most part, however, analytics are widely unavailable.
I caught up with the attendee who asked the original question about failing over his disaster recovery functions, and asked him what’s got him the most frustrated. “It’s frustrating to feel like you’re looking for an infrastructure rather than a solution,” he said, adding that his law firm with about 1500 users can’t meet the overhype of cloud computing.
When I brought this up with Randy Bias, he responded that a lot of people looking into the cloud don’t fully understand all that it entails. “There’s no pixie dust,” Bias said. “There’s no cloud dust that’s going to magically make your application run better.” With all of the hype that cloud receives, it’s become a sort of computing Second Coming, and Bias says to snap out of it. The main challenge is that leadership doesn’t understand the difference between the infrastructure they’re running now and the cloud. They want to move these enterprise applications, complete with their flaws, expecting the cloud to make them run better, not understanding that a move to the cloud might first include migrating to SaaS or building a new app whose architecture can work properly within the cloud.
Dealing with service providers’ opacity
One critique of service providers in the cloud is the lack of visibility into their operations. Bias suggests utilizing third-party offerings from companies such as panelist’s Alex Polvi’s CloudKick to build upon services like Amazon Web Services that Amazon can’t provide itself.
Bias’s suggestion for signing up with a cloud provider that doesn’t offer full visibility? Find someone who can provide services such as reverse mapping to allow you to compare your own performance with other users with similar workloads. He focused on the “immaturity” of the cloud’s ability to respond to performance issues: “Automated correlation and response is the key…the capability exists, maturity is the issue.”
So what does that mean for the rest of us?
Cloud computing has an almost cult following with evangelists preaching the economic benefits of migrating to the cloud. The bottom line at Interop New York? Great reaping follows great sowing. Depending on the type of workload the enterprise wants to deploy in the cloud, there are more factors to consider rather than simply which service provider to use. At this point in the industry’s development, unfortunately, there’s still a lot of heavy lifting left on the shoulders of the user.