It’s Beginning to Look a lot Like Excess
With fall upon us and the leaves turning colors, it’s a good time to think about how seasons affect computing requirements. Many retailers earn the majority of their revenue during only one period of the year. This is never more apparent than what happens each and every holiday season. Stores all across the country run sales and promotions, while increasing inventories to meet holiday demand and excitement.
When a person responds to a promotion, it requires computing resources; more web traffic requires more computing resources. All of these events require matching computing power to meet the demand. In the past you needed to size your environment for this worst case scenario. With cloud computing that’s no longer the case.
I’m often asked what the difference is between cloud solutions and virtualized solutions. First of all, virtualization is a great technology that has done wonders for the information industry over the past decade. Our systems are operating more efficiently, costs have been reduced and the time to market for new solutions is drastically cut. There is however, the one major problem with simple virtualized solutions that I alluded to earlier.
When you virtualize any environment you still need to plan for your worst case scenario. If at one time during the month you need a certain amount of computing power you still have to size your environment for that period of high demand. Yes, you are probably much better off in the virtualized environment but you still are going to have a lot of excess resources on hand. Excess resources translate to electric bills, manpower costs and overall inefficiencies.
The problem above is exactly where cloud computing shines the brightest. You no longer need to worry about that worst case which happens only once a month. Since a cloud environment allows you to utilize resources on demand, your applications can grow as needed. If millions of people log in Christmas morning to spend their newly acquired gift cards, your application can expand to thousands of servers if needed. Once they are no longer needed, they are decommissioned.
In a true cloud environment you can initiate events based on triggers that you define. In an ecommerce system you might look at the number of connections, or overall system response time. If your application starts to perform sluggishly the cloud API can begin to spawn off additional application servers. Not only are the servers spawned, but they are also added into your load balancing pool that you have in place. Once the load settles down, the systems can be removed from the network, decommissioned and you no longer are paying for them to be in action.
With this type of elasticity you can obtain great cost savings by only paying for the resources as you need them. Take a look at your applications; look at your server counts. I’m sure you can find systems in place that are not being fully utilized and could greatly benefit from being migrated to an elastic cloud platform.
In two weeks, I’ll expand on application elasticity and review some specific technical details of exactly how it works on Amazon’s AWS.