Bringing ITIL to Life: Automating IT Capacity Management

Facing hundreds of servers supporting vital business functions, capacity management automation has become a must.

“The sheer volume of information required to do capacity management in today’s highly complex IT infrastructures makes automation a necessity,” said Ed Holub, an analyst at Gartner. “Even with automation in place, however, there still is a lot of effort required by senior IT professionals to effectively manage capacity.”

In capacity planning, data collection tools are first put in place. This enables organizations to gather performance and capacity data, which can be analyzed to build a baseline view of where the current infrastructure stands. With this in hand, the organization can better understand existing business plans and their potential impact on the existing IT infrastructure.

This highlights any shortfalls based upon the prediction of future resource needs. With these results reported to management, the cycle continues: Capacity and performance data is gathered on the upgraded infrastructure, which can then be analyzed and new baselines established. Thus capacity planning is a continuous process.

As workloads change, hardware is added or networks are reinforced, new baselines must be isolated and future needs forecasted with accuracy.

“The first element of capacity management is visibility of the infrastructure in your environment and knowledge of how the elements are connected together to deliver business services and the associated service levels,” said Rob Stroud, an IT Service Management evangelist at CA. “The second element is to understand the demand on your environment.”

Any organization beginning capacity planning activities for the first time faces a daunting prospect—the entire enterprise lies before them. Every process, every resource, every system and every building is a potential target.

Getting Started

The best approach is to prioritize capacity planning efforts based on mission-critical needs. That means focusing on infrastructure components supporting those applications necessary to business survival first. Typically, this centers around order processing, order fulfillment, manufacturing and customer service, depending on the business.

Once priorities have been established, the capacity planner should begin with a resource view to gather data, look for outliers and find out more about them. With that data in hand, the next step is to build profiles for each component or groups of components such as clusters, banks and mirrors.

The capacity planner should also dig in to locate repetitive cycles. For example, there might be a spike on server usage every Friday afternoon caused by everyone logging on to check messages and complete tasks before the weekend. Monthly, quarterly and annual processes can also be tracked. Capacity planning efforts can be thwarted by a failure to take these repetitive cycles into account.

Further, the capacity planner must determine representative time frames. This is meant to discern usage levels that fit various time frames: How many workstations will be in use at any one time? How will usage patterns shift over time? Similarly with servers, representative time frames must be established to take into account usage and other metrics.

Obviously, such tasks require automation. But rolling out performance data capture software across several thousand servers can be a daunting task. Even if agents are used, they still need to be configured in order to customize the data collected and the way it is aggregated for reporting purposes.

Further, associating business events to usage can be problematic. Performance data, after all, is of little use if you can’t determine the business events associated with the usage. Large organizations with several hundred applications, for example, make this task complex and extensive.

Such challenges can be overcome by using installation scripts that can be easily integrated into existing software distribution tools to help automate installation. Centrally based administration can also facilitate configuration by propagating commonly used configurations across large number of servers. For example, operating system component usage may be accounted for in an “overhead” category and a database management system accounted for in a “DBMS” category.