Using Unicenter TNG to Enhance Performance Management
The Unicenter TNG performance management tools compile system utilization data and accounting data, enabling activities such as capacity planning, accounting charge-back for allocating computer system costs, and long-term planning and trend analysis. These Unicenter TNG performance management activities involves data collection and analysis and helps you make decisions while building efficiencies into the system.
Unicenter TNG helps take the guesswork out of performance management. An added benefit is that the reports and charts from the performance viewer can be used to visualize resources and justify decisions.
For example, Steve is trying to decide whether his group needs to purchase more workstations or if the group's members can somehow squeeze more performance out of the ones they have. Unicenter TNG routinely collects all types of data on enterprise resources. Steve can extract the data that pertains to a particular set of machines and determine how much demand is placed on those systems in a 24-hour period. By comparing the CPU usage of several workstations, for instance, Steve can determine how system workloads may be redistributed, leading to a more efficient use of resources. Of course, his conclusion may be that all systems are nearing capacity, and it's time to order new equipment. He can then use the supporting data to justify his purchasing request.
Data collection and trend analysis
Unicenter TNG performance management enable you to compare real-time performance data with data from a previous time period — yesterday, last week, last quarter, or two years ago (assuming such data was being collected two years ago).
The advantages of comparing new and old data include:
- Anticipating problems: Side-by-side comparisons let you observe changes in system resources and potential problem areas, such as memory thresholds, disk usage, caching, and the like.
- Stronger analysis: Over time, as the historical database expands, capacity planning and analysis capabilities become stronger. Online charting and reporting capabilities help you decide whether systems are performing within acceptable limits.
- Empowered decision making: The IT group controls what type of data is collected and how it is collected. The administrator can decide which resources to monitor, when data should be collected, and how long data should be retained.
Unicenter TNG uses two types of agents for collecting performance-related data:
- Real-time performance agent: Also called performance scope, this agent gives you a real-time view of system performance.
- Historical Performance Agent: The HPA collects time-banded data, which is stored in performance cubes.
- The time-banded data collected by the HPA is performance data sampled at a predetermined interval and represents a snapshot of how the systems are behaving at a specific period of time. The HPA queries the operating system every n minutes (typically every 5, 10, 15, or 20 minutes) to collect performance data and write it to a cube file. The administrator can define how often data is collected and vary the frequency according to the type of data. So, for instance, Unicenter might poll memory capacity every 3 minutes but collect data on how many users are logged in every 20 minutes.
- Data collection can be resource-intensive. Frequent polling (such as a 60-second or 3-minute interval) uses more systems resources than a longer interval (such as an hour, or the 20-minute default provided by Unicenter). Administrators should consider how frequently different resources need to be polled and use the collection interval that is most appropriate to the organization's needs. You can accelerate poll time for special purposes, such as testing, but be sure to slow the pace when the test is complete.
Historical performance agent (HPA)
Unicenter uses the HPA to cull data. The HPA performs several important functions:
- Collects data: The HPA periodically collects all types of systems data across the IT enterprise. Such data includes memory usage, disk usage, cache, connections in use, the number of users logged in, and other systems resource information.
- Generates performance cubes: The HPA writes the collected data to a performance cube file, which is a comma-delimited file with date, time, measured parameter (such as memory usage), and value fields.
- Distributes cubes: The HPA transfers the cube file to any number of managers by using CA Messaging (CA-M) and CA File Transfer (CA-FT).
The performance cube is actually comma-delimited data that can be viewed through Excel. Unicenter TNG provides custom templates and menu items that enable you to view and graph the cube files.
Types of performance cubes
Depending on time period and scope, the HPA will create three types of performance cubes:
- Daily performance cube: This cube encompasses data from a period of up to 24 hours, and it can be combined with other daily cubes to produce an average of many days.
- Period performance cube: The period cube comprises daily cubes compiled from one machine. This cube covers more than one day and could encompass weeks, a month, a year, or more.
- Enterprise performance cube: The enterprise cube comprises daily or period cubes collected from more than one machine.
Unicenter TNG helps maximize the use of resources with its capacity planning capabilities. With performance management, you can figure out how systems are being utilized. You may observe, for example, that when running batch jobs after hours, the server in accounting frequently encounters memory problems, exceeding thresholds that have been defined. You determine that the server in the human resources department is rarely used at night. Based on this observation, you can reroute some of the accounting jobs to run at night on the HR machine.
By considering performance metrics, such as response time, throughput, and concurrence, Unicenter TNG's capacity planning lets you:
- Assess server usage.
- Determine which machines are overused and which ones may be underused.
- Examine trends in usage (such as whether a server is being used more or less frequently over time).
- Identify bottlenecks, such as batch jobs piling up in a queue or excessive network traffic degrading response time.