- Silicon Valley's 19 Coolest Places to Work
- Is Windows 8 Development Worth the Trouble?
- 8 Books Every IT Leader Should Read This Year
- 10 Hot Hadoop Startups to Watch
Network World - Knowing exactly how much and what type of cloud service a company needs is one of the more challenging aspects of deploying a cloud strategy, and most enterprises are getting it wrong, according to experts.
"The whole premise of the cloud should be to drive the IT director to think small, but people aren't doing that," says Sharon Wagner, president of Cloudyn, an Israeli company whose SaaS application helps businesses monitor their cloud usage and provides recommendations on how to right-size it. "Many customers are over provisioning, which leads directly to over paying."
Experts agree. "I think it's a problem just about everyone could be facing," says Paul Burns, an independent cloud analyst at Neovise. Burns believes there is somewhat of a misnomer about paying for cloud services. While many consumers think they are paying for resources they use in the cloud, instead users are actually paying for a certain capacity of compute or storage. Whether they use it or not is up to them.
"Just about every customer likely has an issue with this," says James Staten, a vice president and principal analyst at Forrester. "Some have had that shock bill from their service provider so they've gotten pretty good at dealing with it, but almost everyone could benefit from taking a closer look at their actual usage."
So what can be done about overprovisioning of resources? Companies can fairly easy install measures to prevent it, Staten says. The best way is by having load balancers written directly into application code that automatically scale up and down resources based on need. Just as important as scaling up though is scaling back down when those resources are no longer needed. Staten warns, however, that enterprises can run into some complications. A bug in a software or a distributed denial-of-service attack can, for example, create false scaling requests for additional resources.
Overprovisioning occurs mostly in infrastructure-as-a-service and platform-as-a-service instances because those are metered based on an amount of compute or storage, Wagner says. Software-as-a-service offerings traditionally don't have as much of a problem with the issue because they are generally charged on a per-user basis.
It can happen for a number of reasons and Staten says one of the most common is that IT professionals just aren't thinking about the cloud in the right way. Developers and other end users usually ask for more resources than they need so that they don't have to "go back to the well asking for more," Staten says. It requires a "mind-shift" to think small and scale up, he says.
The issue is complicated by the numerous offerings and prices for cloud services in the market, Wagner says. Cloud vendors offer customers a range of options at different prices when purchasing cloud services.
For example, industry-leading Amazon Web Service's popular Elastic Cloud Compute (EC2) has four versions of on-demand instances to choose from. They range in price from $0.15 per hour for a small instance that provides 1.7G of memory, a single EC2 compute unit, and 160G of instant storage, all the way up to the extra-large version, which is $1.20 per hour and provides 15G of memory, eight EC2 compute units and 1,600G of instant storage. There are midsize and large offerings falling in between. Beyond that, there are offerings for high-memory on demand and high-CPU offerings, each which have different sizes of compute power as well. Or, customers can choose to purchase reserved instances, which are paid for on an annual basis instead of being metered on an hourly rate. Meanwhile, AWS recently announced its 19th price decrease in the past six years, and Windows Azure and Google dropped their prices as well.