IBM Cognos, User Manual

Page: 222 / 411

Using the Log File to Analyze Processing Bottlenecks
The Cognos Transformer log file provides useful information to help you diagnose
the cause of processing bottlenecks during cube builds.
Complete the following procedure:
1. Import these delimited-field data log files into a spreadsheet, such as Microsoft
Excel:
LogFileDirectory=f: \logfiles\
2. In the resulting log file import, select the header of the E column and, if you
are using Microsoft Excel, from the
Data menu, click Filter and select the
AutoFilter option.
3.
In the list, select either
(NonBlanks)
or one of the three phases in the process.
The spreadsheet now shows only the timing information for your selection.
4. Select several cells in the F column and select the Sum command, to add the
timing values for your selected cells, as displayed in the lower toolbar.
5. Repeat the above steps to examine each subphase. These appear in distinct
blocks as follows, with each completing before the next subphase begins:
v Data Read
INITIALIZING CATEGORIES
OPEN DATA SOURCE
READ DATA SOURCE
MARKING CATEGORIES USED
v
Metadata Update
SORTING
UPDATE CATEGORY AND PROCESS WORK FILE
METADATA
v
Data Update
CUBE UPDATE
CUBE COMMIT
Note: If the timing shown for TOTAL TIME (CREATE CUBE) is different from
the timing for the separate phases, you can adjust for the time difference in the
Cube Update phase.
6. Assess how much time is spent on each phase. If you notice that, over time,
cube builds are taking longer to complete even though the data set is relatively
constant, compare each phase in the log files, to see where increases in build
times are occurring.
208
IBM Cognos Transformer Version 10.1.1: User Guide