January 04, 2010

Set Up A Test Server System

By Elizabeth Millard at Processor Magazine

Key Points

• Testing is needed in order to prevent application outages and unexpected downtime.
• Direct replication of hardware and software is too costly for an SME, but virtualization can create testing benefits without significant expenditure.
• Having a consultant on call can be helpful for getting advice from experts who understand application design and installation issues.

When implementing new programs and system changes, many small to midsized enterprises understand the need for creating some type of test environment so they keep a “live” system out of danger in case of glitches. Although there are some instances where a direct replication of hardware and software can be overly costly, there are still ways to develop a network of test servers, staging servers, and production servers.

Seeing The Need

One of the most important aspects about a test server setup to understand is the necessity of such a system.

“Unfortunately, most SMEs think that setting up test and stage environments is too complex and costly for them to have in place in their corporate infrastructure,” says Dylan O’Connor, CTO of Staples Network Services by Thrive (www.thrivenetworks.com). “What most SMEs fail to realize is that by not addressing the need to test and stage releases of their software applications in a methodical manner, they expose their IT infrastructure to greater risks.”

Those risks include unexpected downtime or application outages, he notes, which can happen when a release is taken straight into the production environment without proper testing. O’Connor believes that most SMEs should think about their current IT infrastructure with an eye on replicating the infrastructure in a way that mimics production as closely as possible but also keeps their budgetary constraints in clear focus.

Using Virtualization

Direct replication on a one-for-one basis in both a stage and test environment shouldn’t really be considered for an SME, notes O’Connor, because the costs are just too prohibitive.

Also, that type of setup requires that a significant amount of manual configuration work must be done to keep the environments in sync, adds Tomi Kankaanpää, chief technologist at Symbio (www.symbio.com).

But an approach that leans on virtual infrastructure to replicate the majority of the infrastructure can be done on a very budget-friendly level.

“Eighty to 90% of the issues caused by new code or application release can be tested completely virtually at less than 20% of the cost of replicating the entire production infrastructure,” he says. “Given that most SMEs are doing little to no prerelease testing, getting that amount of testing capability out of that low [of] an investment is a simple and cost-effective approach to application release staging.”

Although the initial capital expenditure might seem daunting at first glance, O’Connor notes, the long-term cost savings achieved by leveraging the approach can more than pay for the initial investment. By using a virtual infrastructure, one single server can run many virtual server instances on one piece of equipment.

Also factored in is the ability to save templates of common server images—for example, an SQL server, a Linux Web server, and a server running Windows 2008—and this can create a solution that allows for test and staging environments to be set up on the fly, which may be necessary for instances in which new applications have to be installed quickly.

A virtualized test environment can be deployed multiple times, and different testing activities can be carried out in parallel. Another benefit is that such an environment is more transferable, so it can be utilized with outsourcing partners.

The same virtualized environment can be deployed in different hardware configurations and scaled according to the phases of the project, Kankaanpää says. Time and money are saved by maintaining the core server environment virtualization image instead of multiple physical environments, and the benefits can be even greater if the virtualization and hosting services are used for deployment.

Key points to consider when setting up this type of environment are performance, security, and reliability requirements, Kankaanpää notes. He says, “The best approach to determine whether to go for a full-blown environment is to perform combined ROI and risk analysis for different alternatives—such as virtualized vs. real environment—and make the determination based on the results of this analysis.”

Bringing In Help

Even with a strong test server environment, it can be helpful to bring in experts when a rollout or upgrade is major, says Shane O’Donnell, vice president of engineering for BlueStripe Software (www.bluestripe.com).

“In the cases where customers have problems moving apps from test servers and into production, the common theme is that they have tried to substitute someone who understands the application’s design with someone who can simply install the software,” O’Donnell says.

In general, test server setups require more planning than just getting the app up and running, he adds. Instead, test environments need to work hard to anticipate how a server-application combination will relate to the greater whole of data center resources.