Predicting capacity needs for user applications is a keystone activity for IT organizations who adopt an aggressive or proactive stance on capacity management. Whether your organization’s capacity planning methods are simple projections or utilize complex heuristics, the metrics involved are focused on the performance and efficiency of the infrastructure supporting the applications (CPU, storage, memory consumption, etc.). However, there is an additional metric source that’s both the key driver for growth and the reason the application exists in the first place: the users.
Aside from automated processes, user activity is directly correlated to resource consumption. This can be better understood by determining the impact that an individual user has on application resources. With this user-resource metric, or impact value, projecting the growth or decline of user activity becomes a great “finger in the wind” for understanding capacity supply and demand. In a virtual or private cloud environment, things are understandably more complicated, but that’s a topic for another time.
You can collect and analyze user activity just as you do performance data; giving you greater visibility into application performance and the opportunity to be more efficient with capacity planning. Establishing baselines on user metrics can help you catch unexpected growth or decline in user activity. This allows you to be more efficient with resource allocation as you can quickly respond to changes in demand.
User data can be collected through application management technologies such as WMI or JMX or more commonly through application logs. The problem with this data is that it’s often stored in separate silos from performance data in log- or application-only solutions. Further complicating this is that these solutions often reside within different organizations such as the application or security teams. There are solutions, such as SevOne, that are capable of collecting and analyzing the infrastructure and application performance data as well as the user data. These solutions help avoid the hassle of farming metrics from different silos or business units and in some cases avoid having to manually correlate the data.