This course is run at the The University of Nottingham within the School of Computer Science & IT. The course is run by Graham Kendall (EMAIL : gxk@cs.nott.ac.uk)
If we assume we only have one processor and there are two or more processes
in a ready state, how do we decide which process to schedule next? Or more precisely,
which scheduling algorithm does the scheduler use
in deciding which process should be moved to a running state?
These questions are the subject of this section.
In trying to schedule processes, the scheduler tries to meet a number of objectives
1. Fairness : Ensure each process gets a fair share of the CPU
2. Efficiency : Ensure the CPU is busy 100% of the time. In practise,
a measure of between 40% (for a lightly loaded system) to 90% (for a heavily
loaded system) is acceptable
3. Response Times : Ensure interactive users get good response times
4. Turnaround : Ensure batch jobs are processed in acceptable time
5. Throughput : Ensure a maximum number of jobs are processed
Of course, the scheduler cannot meet all of these objectives to an optimum
level. For example, in trying to give interactive users good response times,
the batch jobs may have to suffer.
Many large companies, that use mainframes, address these types of problems
by taking many of the scheduling decisions themselves. For example, a company
the lecturer used to work for did not allow batch work during the day. Instead
they gave the TP (Transaction Processing) system all the available resources
so that the response times for the users (many of which were dealing with the
public in an interactive way) was as fast as possible.
The batch work was run overnight when the interactive workload was much less,
typically only operations staff and technical support personnel.
However, these type of problems are likely to increase as the world becomes "smaller." If a company operates a mainframe that is accessible from all over the world then the concept of night and day no longer hold and there may be a requirement for TP access 24 hours a day and the batch work somehow has to be fitted in around this workload.
| Last Page | Back to Main Index | Next Page |
Last Updated : 13/01/2002