Solving Business Problems with TeamQuest Performance Software: Reporting on Application Throughput and Cost

Automatic Workloads are easy to establish with TeamQuest software. Based on server process data, properties such as user, group and command are used to create definitions that group processes together into "workloads." Further, it can leverage a process accounting facility where available. A typical Automatic Workload, for example, might be available every hour concerning the percentage utilization of CPUs (%CPU) in a series of servers. Alternatively, you can import your own data into TeamQuest using User Agents. Data is pulled from the source, typically through a command line utility. A User Agent is then defined in TeamQuest Manager and scheduled to collect data at a regular interval.

TeamQuest software release 10 can use Derived Tables as opposed to only offering statistics. This features means you can create reference fields from multiple tables and do calculations based on tables. In addition, you can create new tables from existing ones. TeamQuest agents can bring data into TeamQuest Manager from multiple databases throughout the enterprise. Analysts can then access reports in several ways: by report, by resource or both. These reports can give a summary, an overview, specific details, vital statistics, etc., as defined by the user.

Example Scenario
Let's look at an example of these tools in the real world. The reporting requirements were for Web-based, interactive reporting and the capturing of statistics such as:

• Total Work (X units of work completed)
• Rate (X units of work completed per second)
• Cost (X cpu seconds used per unit of work completed)

Several servers and applications were included in this example. Each application was isolated using Solaris Zones. In this scenario, TeamQuest IT Service Analyzer generated Web-based, interactive reporting of key metrics such as CPU usage, disk reads/writes, network reads/writes and memory paging. All it took was flipping a switch to activate the automatic workload set based on the Solaris Zone configuration

A "Unit of Work" User Agent was used to query the zones on each box to see how much work each was doing.

With Derived Tables, the performance data was compiled by agents and combined by TeamQuest into customized tables. This could be taken a step further into new calculated fields. For instance, unit of work (uow) statistics can be gathered and formulas developed to make additional calculations. In this example, uow completed/time interval provided a uow rate metric. Similarly, the total CPU stat/uow completed provided a new metric termed uow cost. These stats were used to provide more insight in terms of utilization rates, the costs of different workloads and more.

• uow_rate = uow_completed / Interval
• uow_cost = totcpu / uow_completed

Creating User IT Resources
In TeamQuest, IT Resources are automatically generated. However, you can also create your own. Open up the Create Report Templates in IT Service Analyzer to create the IT Resource types and groupings you desire such as sibling and parent, as well as their overall structure.

In this example, we decided to find out how much work was being done. A Derived Table was automatically calculated by TeamQuest to show the amount of work being done by each application and server. Another Derived Table showed how long it took each server to produce one unit of work. This highlighted the fact that one server was creating a lot less work than the other two, making it possible to see the cost of each unit of work.

Yet another Derived Table covered application-specific calculations of the amount of work. This also revealed that some applications were doing a lot more work than others. If preferred, such results can be shown in tables rather than charts in order to see all the details.

Now let's use this data to solve some of these problems. A CPU upgrade was performed to handle the poorly performing server. Before and after graphs demonstrated that this upgrade solved the issues on poor server and application performance. But the fact of an improvement may not be enough. Did you get the gain you expected? If you anticipated a performance leap of three times and only got a 2x gain, you could then drill down to uncover further bottlenecks.

Conclusion
If you routinely make use of the Solaris or AIX Resource Manager, you should seriously consider activating TeamQuest Automatic Workloads. TeamQuest User Agents make it possible to combine technical data and business data.