Optimization Of Gateway Settings For Ibm Cognos Series 7 Iqds; Keeping Model And Cube Sizes Within Practical Limits - IBM Cognos User Manual

defined in the cogtr.sh file, or the environment variables TMPDIR, TEMP, and
TMP as defined by the operating system. If multiple variables are defined, Cognos
Transformer uses the first one in the list.
The UNIX or Linux command line syntax for specifying global preferences is
cogtr -fpreferences.rc -mmodel.mdl
The Windows equivalent is
cogtr.exe -n -fc:\preferences.prf model.mdl
Optimization of Gateway Settings for IBM Cognos Series 7
IQDs
To further shorten the data read phase for IBM Cognos Series 7 IQDs, you can
change the database-specific settings found in the gateway .ini files included in the
Cognos Transformer installation directory.
Search for a file name such as cogdm*.ini, where the asterisk represents a specific
database version. The entries in each gateway .ini file are different, depending on
the database type.
Note: For IBM Cognos data sources, see the Architecture and Deployment Guide.
Example - Change Oracle Database Settings
Oracle uses the cogdmor.ini gateway file for database-specific settings. Consider
adjusting the following settings:
v Fetch Number of Rows
Increasing the number of rows to fetch in each fetch operation can improve
performance on some systems. Although the current limit for this number is
32767, numbers larger than the default (100) may degrade performance on some
systems.
v Fetch Buffer Size
Increasing the size of buffer used during fetch operations from the default (2048
bytes) can improve performance on some systems.
Where both entries have been changed, the row setting takes precedence over the
buffer size setting.

Keeping Model and Cube Sizes Within Practical Limits

There are practical limitations on the file size of production models and cubes,
based on typical memory capacities and run-time performance requirements, in the
production environment.
For example, you can limit the size of your models to 2 million categories each.
Pay particular attention to bloating due to excessive use of labels, descriptions,
short names, category codes, and custom views. Metadata associated with your
structural data source dimensions, levels, and categories can contribute
significantly to overall storage requirements.
In virtual memory terms, models held in allocated memory are limited by the
available address space. Also, performance is severely impacted if a model is larger
than the available physical memory.
Chapter 9. Guidelines for Optimizing Production Environments
207