NetBatch Manual

Scheduler Planning, Configuration, and
Management
NetBatch Manual—522460-004
3-3
Task 2. Determine Scheduler Numbers and
Attributes
General Rule
As a rule, plan no more than one scheduler for each node where you want jobs placed
under NetBatch control. A single scheduler on a node should be enough to satisfy
users’ requirements on that node.
The benefits of limiting schedulers to one per node are simplified job and system
management are:
•
You can manage CPU resources and job flow more efficiently and easily with one
scheduler than with several.
•
Having a single scheduler means there is only one scheduler database to maintain
and one scheduler log file to examine.
Of course, NetBatch does not prevent you from having more than one scheduler per
node. As you increase the number of schedulers, however, job and system
management become more complex.
Exceptions to the Rule
The one scheduler per node rule has some exceptions:
•
For user-training and job-testing purposes, plan at least one extra scheduler on
each node where those activities might occur. The extra schedulers enable users
to familiarize themselves with NetBatch functions and to test jobs without
disrupting your production schedulers.
•
For nodes where more than 500 concurrent jobs are likely, plan as many
schedulers as you need to service the job load. For example, to run 600 jobs
concurrently, you need at least two schedulers. For more information on concurrent
jobs, see Task 1. Compile Job Information on page 3-1 and Task 3. Establish Class
and Executor Configurations on page 3-6.
Determining Your Schedulers’ Attributes
Every scheduler has a permanent set of attributes that:
•
Specify the preferred CPUs for the scheduler’s backup process
•
Specify default values for some executor and job attributes
•
Determine the number of tape drives available for use by jobs
•
Enable or disable job submission, job initiation, and event-message generation
•
Control use of the AT job attribute and the accumulation of run backlogs
•
Determine whether the scheduler treats jobs submitted from licensed requesters
on remote nodes as local jobs, not as remote jobs
•
Specify the scheduler’s concurrent-jobs, temporary-executors, and execution-
priority limits