Understanding Your Cloud: Availability, Performance, Volatility

If you’re anything like me, when you finished reading the VAR Guide to Selecting a Cloud Computing Platform, you became unnaturally attuned to all of the conversations about “the cloud” going on around you. Your IT manager may have presented a plan to “go into the cloud.” You might have heard that 35 CPU-years of cloud computing were used to find “God’s Number” when it came to solving the Rubik’s Cube. Maybe you just channel surfed to a nephology panel discussion on The Weather Channel. With the exception of that last possibility, you’ve noticed that “the cloud” is everywhere. It’s time to pin down what the cloud really entails.

When it is time for you to choose your cloud platform, you will need to know the six key characteristics of the cloud and how reality differs from the commonly held misconceptions. Let’s look at each characteristic, its associated misconception, why it is important and what you should ask as you evaluate cloud solutions to meet your business needs.

1. Availability – You Want Your Cloud to “Be There”

Misconception: All cloud platforms are highly available and globally distributed.

Reality: In a cloud environment, you can’t point at a single server or a drive with your data. That means you don’t have to worry about a drive or processor failing and taking down your site, which is a good thing. That fact doesn’t make your data “highly available,” though. The cloud as a whole is only as reliable as the network it’s on and its data center environment. When you consider the geographic distribution of the cloud infrastructure, do you really want it spread across the country? If you don’t have control of where your data is located, what would keep one line of code from being pulled from a server in San Francisco while the next line is pulled from a server in New York City?

Questions to ask: Does your application need to be highly available? What is acceptable downtime? What is the process and recovery time for failures? How will you monitor availability? How does the provider handle maintenance? Do you know where your data will be geographically? Are multiple locations available for geographic diversity?

2. Performance – You Want Your Cloud to Deliver Results

Misconception: You can swap your dedicated server infrastructure for a cloud infrastructure, one for one.

Reality: If you aren’t tied down to a single piece of hardware and you’re able to scale processing power as high as you need it, you might assume a cloud instance will perform as well as a similarly provisioned dedicated server. The problem is that while you can match specs on paper, your cloud infrastructure is distributed across several pieces of hardware. While many cloud platforms have improved their relative performance, dedicated alternatives still have an edge in certain cases. Understand your workload’s current and potential bottlenecks as you compare the on-paper computing power.

Questions to ask: What are the current bottlenecks in your workload? What are the demands of the workload on each major element of the computing infrastructure? What elements of your workload can you scale-out? What elements of my workload must I scale-up? What services can I offload to better tune my workload?

3. Demand Volatility – You Want Your Cloud to be Dynamic

Misconception: All clouds scale automatically, infinitely and seamlessly.

Reality: The cloud architecture is designed to scale. If you aren’t confined to a single server, you’d theoretically be able to leverage an infinite number of processors and an infinite amount of disk space. The reality is that the cloud may be limited by the software used to run them or by the amount of available resources. Many cloud platforms rely on you to manually select the processing power and disk space you need … This is a good thing when you’re not ready to pay for a huge spike in usage, but if you’re anticipating a volatile workload, it’s important to understand how easy it is to scale up and down the stack to meet changing demand. While utility-based billing allows for you to pay for exactly what you use, it can also be more expensive when you compare it against the same usage on server with dedicated resources. Track your usage and note the peaks in your demand. If you find your workload relatively constant or predictable, a dedicated alternative might actually cost less.

Questions to ask: Is this a project or production-oriented workload? What is the process and level of effort to add additional scale to the workload? What is the frequency of change required to support demand optimally? How predictable is the demand? Does frequency of change make utility billing more or less economical? Does the business highly value predictability costs?

To Be Continued

We’re only halfway through our six-step decision-making framework, and I hope you’re still with me. You may have walked into this series with one question only to walk away with a dozen, but each one of the questions you walk away with will help you understand how the cloud can meet your specific business needs.

Drew Jenkins is manager of channel sales for SoftLayer. Monthly guest blogs such as this one are part of The VAR Guys’ annual sponsorship.