Capacity Advisor Overview

Capacity Advisor is a component of the VSE Management Software as described in Chapter 17, "VSE Virtualization Manager." The Capacity Advisor tool collects workload utilization data for every workload in the enterprise data center. This data is collected on a daily basis from each system and is stored on the HP Systems Insight Manager CMS. The collected workload utilization data is referred to as a workload's historic profile.

A workload's historic profile is its resource consumption history. This information consists of historic metrics for CPU, memory, disk I/O, and network I/O utilization. Reports and graphs are available for each metric in the workload's historic profile. In addition to generating reports and graphs, a historic profile can be edited using the Historic Profile Editor.

The historic profile editor can be used to mark data points as invalid in a workload's historic profile. Atypical events occur in computing environments that skew the actual resource requirements for a workload. For example, an intermittent hardware failure can cause an application to require unnecessary retries and thus record unusually high utilization metrics. For these events, the historic profile editor in Capacity Advisor can be used to invalidate the unusually high data points. The unwanted data points can be invalidated according to a given date range or it can discard values above a certain threshold.

After ensuring that the historic profile accurately reflects the nature of the workload, the next step is editing the workload's forecast. A forecast is an extrapolation of the baseline data into the future. Forecasting allows growth trends to be input in order to adjust for expected increases in resource requirements. In situations where the resource requirements for a workload are expected to change, the forecast editor allows the forecast for a workload to be modified. For example, if a workload is expected to grow 10% over the next 12 months, the forecast editor can be used to scale the workload accordingly.

As part of editing the workload's forecast, the workload's baseline can also be tailored. A baseline is a workload's resource utilization blueprint based on its historic profile. One way to create a baseline is to select a specific day in the workload's history as the representative sample to be used for planning purposes. Workloads that have daily repeating usage patterns such as backup systems are candidates for this type of baseline. Another way is to create a baseline from an arbitrary period of historic data. Workloads that are cyclical on a weekly basis are best modeled with this type of baseline. Several more options exist for creating baselines. It is important to remember that baselines should be created using the appropriate dataset for the workload. Only baselines that accurately reflect the business's use model of the workload are useful for capacity-planning purposes.

Editing of the baseline is performed using the Capacity Advisor baseline editor. Unless you made modifications, Capacity Advisor derives the baseline directly from historic information. In some cases the baseline does not require modification. However, most of the time workloads are event driven or cyclical in nature and adjusting the baseline results in a much more accurate and reliable model for capacity-planning purposes. For example, a payroll workload that runs on a biweekly basis requires editing of the baseline in order to accurately plan for the workload and its requirements in the future. Consider a situation where a payroll workload is being consolidated with a workload that runs at the end of each month. Using the baseline editor for each of these workloads to specify the exact nature of the workload makes it readily apparent that consolidating these workloads might not be feasible because some days payroll is processed on the same day as the month-end processing. This simple example illustrates why it is important to create an accurate baseline. If the baseline for the payroll workload does not accurately reflect the workload, the capacity-planning scenario evaluating this consolidation might overlook the problematic times where the peaks for the two workloads are aligned.

The final aspect of editing a workload's forecast is accounting for expected events in the workload's future. Events such as obsolescing a workload or doubling a workload's user base due to a merger can be specified as part of the forecasting process. These events further ensure that the capacity-planning activities performed on a workload are reliable and properly account for known events in a workload's future.

Once the workloads' historic profiles and forecasts have been customized, the next step is creating a Capacity Advisor scenario. This process allows many workloads to be evaluated as a group. A scenario is a set of systems and workloads combined together for capacity-planning purposes. Using the workload forecasts, what-if scenarios can be evaluated to determine if workload migrations, consolidations, or introductions will be successful. Hypothetical systems and workloads can also be defined within a scenario to provide what-if planning facilities. The hypothetical systems can be used to assist in answering questions such as "Will these workloads perform acceptably on the system I'm purchasing?" Additionally, scenarios can be used to evaluate the effects of moving a workload from one system to another. Essentially, capacity-planning scenarios provide a test-bed for performing analysis using both real and hypothetical systems and workloads. These scenarios are reliable because they are based on historic workload utilization data. The historic data is then used to create a forecast and baseline for the workload. The Capacity Advisor product allows capacity planners to effectively and accurately plan for changes in the datacenter.