Process scheduling

Configuring the scheduler

The default configuration includes both the time-sharing and the fixed priority scheduler classes. The time-sharing class is tuned for representative system workloads. Such workloads have a high proportion of interactive processes that sleep often. The fixed priority class can be configured in for applications that need it.

For traditional time-sharing uses such as software development, office applications, and document production, fixed priority processes probably are not necessary and they may be undesirable. First, they consume memory that cannot be paged since the u-blocks of fixed priority processes are never paged out. Second, they introduce new ways to cause performance problems, since a high-priority fixed priority process can block out all other processing. In a computing environment where only time-sharing is needed, you may want to remove the fixed priority scheduler class from your configuration; see ``Changing scheduler configuration''.

On the other hand, if a machine is running an application that has strict requirements for the sequence in which processes must run, then the fixed priority scheduler class provides the only way to guarantee that those requirements are met.

NOTE: Fixed priority processes can have a dramatic negative effect on time-sharing performance.

This section describes the parameters and tables that control scheduler configuration and tells how to reconfigure the scheduler. A basic assumption is that your workload is reasonable for your system resources such as CPU power, primary memory, and I/O capacity. If your workload does too much computation or too much I/O for your hardware, reconfiguring the scheduler won't help. See ``Managing system performance'' for more information.

Next topic: Default global priorities
Previous topic: Effect of priority on swapping

© 2004 The SCO Group, Inc. All rights reserved.
UnixWare 7 Release 7.1.4 - 22 April 2004