Companies and IT departments are under a great deal of pressure to improve the efficiency of their data centers. The reason may well be related to the publicly acknowledged efficiency and effectiveness of Google. In a mere decade, Google has become a hugely profitable organization, running the largest computer network in the world with hundreds of thousands of servers. Its data centers (for it has quite a few) are the envy of many chief information officers. These centers are efficient in terms of hardware use, power use, software architecture, and even location.

Google serves as an excellent example of the IT side of service management because for Google, the whole customer experience is IT. Even with the company's vast resources, customer satisfaction comes down to a single overriding issue: speed. Therefore, response time is the driver of Google's success. We humans are so impatient that we won't tolerate even short delays; if response time stretches out beyond 140–200 milliseconds, we're likely to take our business elsewhere. That's how it is in the search business.

Over its short history, Google's search latency has gone down from a whole second (1,000 milliseconds) to 200 milliseconds. Also, the company put the complete search index to the whole Web in computer memory shared across 1,000 machines that divide the query work. Every time you send Google a query, 1,000 computers jump into action.

Before you decide that your data center strategy should mimic Google's, you need to understand the context of that company's strategy. Google's IT strategy is fairly simple, because about three quarters of the company's revenue comes from advertising associated with its search service. It isn't a stretch to state that dominance in search is the pillar of Google's corporate strategy. In fact, you could boil down its IT service management strategy to the ability to provide the fastest, most accurate search capability in the industry. In effect, Google IT service management focuses on a single Web page. Like any other company, Google also has to manage its own security, compliance, and people.

Many business leaders look at Google's efficiency and effectiveness and want their own data centers to be equally well managed. Most companies have many more aspects to their business, however — and, therefore, to their service management strategy. If your company is a bank, insurance company, hospital, manufacturing concern, or retailer, providing the best possible service can have many dimensions and can involve a wide variety of activities in a multitude of contexts. Therefore, your IT strategy can be very complex because it likely involves many different systems.

Figure 18-1 shows the activities involved in data center planning, with a focus on both the operational and the evolutionary. Chapters 8 and 9 discuss many operational issues; in this chapter, we focus on planning for the long term.

Figure 18.1. Data center planning.

The figure indicates the following:

As a consequence, IT strategy can be created with some sense, and the company's efforts and investments can be focused correctly. It's nice to imagine a world in which corporate strategy and IT strategy are defined in fine detail, well documented, and published, but we don't live in such a world. Nevertheless, efficiently run data centers exist because their goals have been set out and data center planning has been done in the light of those goals.

We discuss these areas in detail in the following sections.

Project portfolio management

We devote most of this section to managing the data center by using a PPM approach. The biggest challenge facing IT is planning well and preparing for the unexpected.

When data center managers figure out how to finesse the planning, they can do two things: run an efficient operation and save money. But they don't have a wand that they can wave to achieve these goals, which require careful planning and the use of organizational skills.

Tip

We think that using a PPM approach goes a long way to meeting the goal of balancing IT with business needs (see Chapter 17). Good PPM helps you become proactive but isn't a project management system. Rather, it's a technique and set of best practices for planning and managing data center components. Often, these best practices are built into PPM software. Examples of this software include CA Clarity PPM, Compuware Changepoint, Hewlett-Packard's HP PPM, Planview Enterprise PPM, and Serena Mariner.

Putting PPM in context

In some ways, PPM is the opposite of the old-fashioned fire-drill approach to IT. Rather than thinking of a software upgrade, a new application, or a new server as a task, you think of the combination as a business project. A project might be virtualizing the server population as a way of optimizing power in the data center, and checking overall identity management to secure current and future business initiatives.

Note

In fact, enhancing any element of service management should automatically initiate a process of understanding the following:

• The current process

• The way data has been managed

• How other processes have been managed

Warning

This process isn't just an exercise. You simply can't make significant changes in your data center without managing the dependencies that will be affected.

Figure 18-1, earlier in this chapter, shows a set of user requests as inputs to PPM. We aren't talking about a plan for maintenance. Managing user requests in the context of PPM has a significant implication in planning for future capacity.

Managing change

Taking a PPM approach enables IT to consider the impact of data center changes on the business as a whole. New compliance requirements may lead to new software that manages those requirements, for example. The company may plan to add new online services, which will increase stress on existing servers and networks, and planned new partnerships will result in a different workload in the data center.

Typically, when large business initiatives occur, two types of IT projects occur: a software development project, and a data center delivery and implementation project. The data center portion can be a subproject of a larger effort, whereas initiatives to improve the data center may be projects in their own right.

Tip

Although anticipating future data center workload is never easy, a project orientation helps management plan better. Aside from efficiently managing data center evolution, PPM helps IT keep the data center from exceeding capacity or at least predicting accurately when capacity will be exceeded.

Note

Planning isn't just for new applications and projects. Existing data center elements, such as the service desk, also are affected as the business changes. Also, you have limits on what you can achieve in any given data center. These limits may include floor space, power supply, personnel, development resources, or financing.

Coordinating downstream activities

Figure 18-2 depicts the activity that occurs downstream from PPM.

Figure 18.2. Downstream activity.

Three major activities occur downstream:

• Capacity planning: This activity used to be relatively simple; when adding a new application, the company added new hardware. Adding capacity is more complex with the introduction of virtualization (see Chapter 15) because then IT is managed as a resource pool. Capacity planning involves workload modeling, the activity of statistically modeling the load on the resources available for groups of applications. The goal is to predict the resource needs of the whole network over time, allowing for factors such as variations in application and network traffic.

• Infrastructure projects: As IT moves away toward a more integrated infrastructure, it also must move toward a more integrated set of service management processes and technology. Consequently, infrastructure projects such as establishing a federated configuration management database (CMDB; see Chapter 9) or virtualizing desktop computers are almost certain to involve changes in IT processes, which in turn may mean using processes that involve workflow and are based on Information Technology Infrastructure Library (ITIL) definitions and models. (For more information on ITIL, see Chapter 5.)

• Systems development: Service oriented architecture (SOA) means both new applications and major changes to business processes. Because SOA makes application components available for use by other applications, it naturally creates new dependencies among applications or causes changes in existing dependencies. So when components are linked to build new applications or even to build end-to-end processes, existing software configurations are changed. This situation creates a testing issue, because now these components have to be tested to ensure that they work in all new contexts.

  As a general rule, SOA increases the dependencies among applications and application components, and the effect needs to be modeled and tested. The same end-to-end modeling needs to be used in capacity planning for the resources needed to support the SOA services. (See Chapter 6 for more information on SOA.)

Focus on the following when you're planning data center evolution:

We hope that we've given you a taste of some of the important issues and approaches that make the service management journey so exciting. There's certainly a lot to think about.