Building an Intelligent IT Infrastructure

The inherent shortfalls of first-generation approaches have motivated the emergence of a second generation. An important example is "hive" computing, where the underlying infrastructure becomes part of a software-managed fabric. The fabric appears to developers, administrators, and users as a single system, independent of the physical hardware configuration.

Hive computing focuses on establishing relationships between the individual computers in a "hive," allowing them to reliably process requests. These relationships define a variety of variables, including how the hive distributes incoming work across the various machines (load balancing); how machines cooperate to compensate for failed hardware; and how they organize to maximize a hive's scalability. These relationships are established dynamically, in a fully distributed manner, which avoids any single point of failure within a hive.

The intention of hive computing is also to allow new computers to enter and exit a hive on the fly, without manual configuration changes, while an application is running. Hive software must ensure that state information is always available in multiple places within the hive; this makes it possible for the fabric to successfully complete every accepted request. As an amalgamation of the core benefits offered individually by first-generation technologies, hive computing's objective is to enable business-critical applications to run dependably on inexpensive, standards-based hardware running commodity operating systems, such as Linux or Windows.

Pathways to Adoption

Each of the intelligent infrastructure approaches we've outlined requires an organization to change the way it approaches managing — and often developing — its business-critical applications. The barriers to change in applying first-generation approaches, not to mention stepping into the second generation often stop progress before it starts. However, strategies and methodologies exist that can act as your roadmap to implementing any of the intelligent solutions described. These methods will help you to maximize the benefits for your company.

One of the most important strategies to consider is risk management — especially when you're talking about migrating business-critical applications. As part of this strategy, it's important to select a single application as a pilot migration project, to be run in parallel with its counterparts. Pilot projects are an effective way to minimize and manage a wide variety of risks, both technology and nontechnology related. They will also lead to a deeper understanding of the issues and challenges associated with transitioning your broader suite of applications and provide insight into overcoming them and achieving the benefits.

Selecting a suitable pilot project will help you to evaluate potential business risk, as well as combat any management skepticism. Various methodical approaches can identify high-payoff pilot projects. An important one to consider is the Pareto selection method. Using this method, you can narrow your search to the 20 percent of applications that tend to cause 80 percent of your current problems.

Divide and Conquer

Get started with the Pareto method by identifying waste streams within your organization and its processes, as well as the costs associated with each. When assessing your organization's applications, identify the individual issues that are negatively affecting your bottom line. Once you've selected some preliminary business processes and associated applications, identify bottlenecks and vulnerabilities and assign a dollar value to these infrastructure shortcomings. For example, if you have a customer-facing application or Web service for which you're struggling to meet service-level commitments, the costs might include those tied to system administration and management, revenue lost to downtime, mandatory penalties paid to customers, or loss of customer goodwill.

The Pareto method allows you to assign numbers to your decision, yielding quantitative data to help you make an otherwise qualitative decision. In addition to this numbers-based approach, identifying a clear statement of the desired outcome will add further validity to your decision.

As an example, an ideal pilot project outcome would be to reduce the dollars spent on an application prior to its migration. Objectives might be to anticipate likely future changes and spending on that application post migration, and second, to reduce critical path steps in the application's key operations. The outcome should also provide increased quality and improved service levels (as perceived by customers). The project implementation should focus on delivering real results in three months or less, provide readily available historical data on process performance, and deliver a true organizational learning experience. We also recommend selecting a project that addresses an ongoing problem that previous efforts were unable to resolve. The resources required to support a project through to its fruition must also be readily available and should affect your decision-making process.

Before initiating your pilot project, make a list of the processes that the project will affect and estimate the effort required to effectively introduce and execute your project. Depending on the intelligent solution that you're migrating toward, program changes to the constituent applications are likely. Your analysis will provide you with a good feel for the scope of your undertaking.

Setting the Stage for Success

It's important to create a test bed for your planned pilot project, so be sure to create a test version of production databases to ensure your project is executed under real-life conditions. Consider network conditions, such as WANs, bridges, routers, and wireless technologies to ensure that the test bed is representative of the production environment.

Analyze and measure the current IT environment to establish criteria for success relating to your project. This process will include quantifying the application's current performance and carrying out a benchmarking exercise. You can then establish more accurate target criteria, which play an important role in justifying the results of your pilot. We also recommend formulating a project objective statement to clarify your goals, keep your team on track, realize the benefits of your plan and help gain general buy-in. The project objective should incorporate and demonstrate a quantifiable payoff.

Effectively reporting the results of your pilot project is as important as obtaining the results themselves. How you relay your successes will affect future management support of intelligent solution adoption — your ultimate goal. This area is where measuring the new environment and benchmarking can really come into play. By preparing a before and after case study that assigns dollar values to the new performance level attained, you'll be better positioned to convince those who hold the purse strings.

The Final Countdown

Before embarking on your journey to the new world, ensure the fundamentals are in place and don't get lost or distracted along the way:

Transition Time

Once you've mapped out your journey and taken the necessary steps to avoid or reduce potential hazards and risks, you're ready to consider transitioning a broader suite of your applications to the new environment. You'll use many of the strategies already described, but to reduce and eliminate the larger risks associated with a full transition, you should also consider identifying and establishing technical plateaus of stability and operability. These technical plateaus will help to prevent or minimize costly data loss and protect the availability of mission-critical applications.

The challenges of migrating your applications to an intelligent infrastructure aren't limited to technology choices. The road to this ultimate goal is rarely smooth, but you can take practical steps to minimize the risks and maximize the potential gains. Like every successful IT project, it's important to plan the work and then work the plan.

Roy Ferguson is founder and CEO of Newmetrics Corp., an operations research consultancy that specializes in optimizing and automating mission-critical business processes.

Samuel Charrington is a vice president with Tsunami Research, a software company focused on hive computing. He is responsible for the company's product and market strategies.