When working with a cloud installation, costs can easily skyrocket -- whether you build a private cloud or adopt public cloud computing services. Though enterprises factor in initial purchasing costs, they often overlook costs associated with building and managing clouds, as well as ongoing maintenance they require from IT teams.
There are ways to keep the
The cost of in-house cloud development versus purchasing
To analyze the costs that arise from building and managing a cloud, you need to think about it from two angles: first, the upfront and monthly costs of software, and second, the costs of paying employees. As an example, let's compare two options for managing a private cloud.
Option one is to pay a monthly fee to purchase premium management tools, such as those RightScale offers. Fees for these management tools can run anywhere from $500 to $1,000 or more per month, depending on the package.
Enterprises have another option: enable their own IT teams to use various application programming interfaces (APIs) to write their own tools, thus avoiding a vendor's fees. But sidestepping the cost of a premium management tool doesn't mean an in-house option is free.
Though open source cloud options are technically 'free,' that doesn't mean they won't cost an enterprise.
An IT team might spend, for example, 200 hours building the tools. Depending on the hourly rate of IT employees, the man-hours involved could run a company $10,000 or even $100,000. In the short term it might seem like you're getting a deal by having in-house staff develop the tools, but after factoring in those costs with the ongoing costs of maintaining the tools, enterprises may not be saving money after all. Additionally, those in-house management tools may not have all the capabilities of premium tools on the market.
When deciding if you should develop in-house tools, ask yourself, "Are we in the business of building software?" For an enterprise with little experience constructing its own software, costs can easily spin out of control. For example, a nursing company decided its in-house IT team would roll out its own internal software. The problem was that upper-level management knew nothing about how to manage a software team; it simply wasn't their business. After two years of struggling through the development process, the team was unable to deliver a final product, and the company lost hundreds of thousands of dollars on wasted development efforts. If the company had examined its own capabilities before embarking on this project, it would have determined that the most cost-effective option would have been to purchase an existing product or hire an outside firm to build the software.
Factoring in the open source cloud option
Though open source cloud options are technically "free," that doesn't mean they won't cost an enterprise. Any software, including open source, can create issues for untrained or inexperienced staff, so it's vital to factor in support costs.
Scalr is a popular cloud management platform that includes an open source option. It offers a set of monthly plans, each with its own pricing structure, in addition to a "free" open source version. Enterprises can either pay for the software and the included support team or use in-house IT admins to install, configure and maintain the system. With the second option, training costs and man-hours paid can often overwhelm any agreed-upon and expected monthly charges.
Clearly, nothing is truly free. You must factor in not only the costs of the products and support plans, but also the time and effort it will cost your company to build and maintain the systems. Weigh the costs of monthly plans against the costs of maintaining an in-house support team.
Finally, factor in the stress involved. You may find it is truly less expensive to choose open source software with an in-house support team. But be sure your staff is really prepared for it. Or will you need to hire more employees?
In the end, the cheapest option might not be the most desirable. Controlling cloud computing costs is no simple matter, but it can be done, provided you factor in all the components up front -- and realize what you're getting in to.
This was first published in November 2012