Tivoli Enterprise
Monitoring Server tuning recommendations for large-scale environments
This section provides tuning recommendations you might
consider to improve either hub or remote monitoring server performance
in large scale environments.
Review these topics and considerations for relevancy to your environment.
Avoid distributing unused situations
To reduce Tivoli® Enterprise
Monitoring Server memory usage and minimize the amount of CPU processing
when many agents connect, consider reducing the total number of situations
distributed by avoiding distribution of situations that are not being
used.
Check ulimit settings for open file descriptors (UNIX/Linux only)
The Tivoli Enterprise
Monitoring Server can use many file descriptors, especially in a large
environment. On UNIX and Linux Tivoli Enterprise Monitoring Servers, the
maximum number of file descriptors available to a process is controlled
by user limit parameters. To display the current user limits, use
the ulimit –a command.
Disable extra fsync() calls by specifying KLGCB_FSYNC_ENABLED=0 (UNIX/Linux only)
The KGLCB_FSYNC_ENABLED parameter specifies whether the
fsync() system call should be invoked after writes to the file system.
The default value is 1, causing fsync() to be called after every write
for maximum reliability. When KGLCB_FSYNC_ENABLED=0 is set, fsync()
calls are omitted after every write.
Avoid storing short-term history data at the Tivoli Enterprise Monitoring Server, if possible
For best performance, when configuring historical data
collection, store the short-term history data at the agent rather
than the Tivoli Enterprise
Monitoring Server when possible. Remember that for some agent types,
data is required to be stored at the Tivoli Enterprise
Monitoring Server.