Tandem Capacity Model (TCM) Manual

Use MeasTCM to aggregate and analyze the samples into process category data in the form of
comma separated values (CSV) files. When you are finished collecting performance data, download
the CSV files created by MeasTCM to the custom folder on your workstation. For information about
MeasTCM, see Chapter 3: “Using MeasTCM”.
Apportioning Performance Data
The fourth step in the capacity planning process is apportioning performance data.
Typically, a TCM administrator translates the capacity planner’s apportionment methods into TCM
using the apportionment tables in the WA model. Chapter 5: “Creating a WA Model”, provides
instructions for each step.
To apportion performance data:
1. Create a new WA model and load the CSV file into the Workload Apportionment model.
(See “WA Model Flowchart” (page 75).)
2. Modify an Apportionment Weights Table (see “Modifying the Apportionment Weights Table”
(page 76)) and a Transaction Count Table (see “Modifying the Transaction Count Table”
(page 77)) for your application.
3. Update the capacity history database with this sample. To update the Capacity History
database, select History > Update History.
4. Update capacity history database for each CSV file downloaded. (See “Load Category Data
(F3)” (page 126)).
Checking an Apportionment Method
The fifth step in the capacity planning process is checking an apportionment method.
To ensure the accuracy of the TCM modeling results, you must examine the reasonableness of both
the performance data sample and the method used to apportion this data. Selecting an appropriate
method involves collecting and apportioning multiple performance samples and then analyzing
the results.
To check the workload apportionment model of your application, compare your results from all
samples by reviewing the Capacity History file. In general, an Apportionment Weights table can
be viewed as reasonable if CPU and disk demands per transaction are relatively constant across
many performance samples with varying transaction mixes and throughput levels. The requirement
for a variety of transaction mixes is crucial. If mixes do not vary, any apportionment method will
appear to be reasonable but might not be.
If your model is not reasonable, you need to collect performance data again or redesign your
Apportionment Weights or Transaction Count tables. For more information, see “Collecting Measure
Samples” (page 30), “Modifying Apportioned Data Using an .AWA File Not in the Current Session”
(page 242), and “Modifying the Transaction Count Table” (page 77).
If you use TCM Expert to generate a Workload Apportionment Weights table and to update the
History file, TCM Expert will check for inconsistent transaction costs in the History file. For more
information about TCM Expert, see Chapter 10: “Using TCM Expert”.
Performing What-If Analyses
The sixth step in the capacity planning process is performing what-if analyses.
Once you select a representative performance sample, you can use the TCM Performance model
to answer what-if questions such as estimated throughputs, host response times, and the number
of processors and disks required for different userdefined scenarios.
Select a representative Workload Apportionment model, also called a baseline, to use with the
Performance model. You can also use TCM Expert to pick the most representative CSV sample and
to load that sample into the Performance model automatically.
TCM and the Capacity Planning Process 23