Tandem Capacity Model (TCM) Manual

10 Using TCM Expert
TCM Expert is a feature of TCM that enables you to do the following:
• Automatically generate Apportionment Weights table from sample data
• Update capacity history with selected samples
• Create a default Transaction Count table
• Pick the most representative sample
• View and modify TCM Expert parameters
TCM Expert provides a way to do capacity planning when you have limited knowledge of the
application design or architecture. When you bring CSV data into TCM Expert, TCM Expert warns
you if the collected data is inconsistent with the current model and advises you when you should
generate a new apportionment scheme.
The TCM Expert tab enables you to access all TCM Expert features. You can use TCM Expert with
CSV files to generate an Apportionment Weights table automatically. However, you must still use
an existing Transaction Count table or transaction count files. If you do not have an existing
Transaction Count table or transaction count files, you must generate the default Transaction Count
table and modify it according to your application needs.
CSV Sample Data Required
TCM Expert needs to have enough CSV sample data to do its work. Depending on the quality of
your data, you might need at least 4 to 5 times the number of defined transaction names. For
example, if your model has two transactions, you would need at least 8 to 10 CSV samples for
the best results. The more CSV files, the better.
The CSV samples you collect should be independent of one another. Thus, the CSV samples should
have varying transaction mixes. To get independent CSV samples, you might have to take several
measurements. For instance, to get 12 independent samples, you might have to take 12 1-hour
intervals over a 12-hour period. For other applications, you might take measurements at the same
time but over a 12-day period to get 12 independent samples. In an automatic teller machine
(ATM) application, for example, you will get different transaction mixes on a traditional pay day
(such as the 1st or 16th of the month) versus another day. You would also get a different transaction
mix on a Friday when customers want to withdraw money for the weekend or the day before a
holiday.
CSV files should cover periods when the different transaction rates are not correlated. As part of
its analysis, TCM looks for linear relationships in the counts in your CSV sample data. For example,
if there are always twice as many withdrawal transactions as deposit transactions, the data is not
suitable for TCM Expert analysis. In this case, retake your CSV samples during different periods
when the established relationship is not as strong or is nonexistent.
If TCM detects any undesirable correlations, your CSV samples are not independent. A dialog box
appears describing the established linear relationship. However, you can continue if you want. If
you decide to continue with your original data and TCM Expert comes up with reasonable
apportionment weights, be aware that the results depend on the established linear relationship
continuing.
Data Splitting
TCM Expert uses the statistical concept of data splitting to give you a higher confidence level in
its results. When TCM Expert generates apportionment weights, it uses two-thirds of the CSV files
selected. TCM Expert does not use the other third of the CSV files selected but reserves those CSV
files as a test sample for consistency. When TCM Expert updates the capacity history, it uses all
CSV Sample Data Required 179