ACT CPU Occupancy Time
CPU occupancy time, also known as pseudo-elapsed time, approximates how long a job would be in the system if it ran all by itself, without the influence of the multiprogramming mix. Two basic methods compute pseudo-elapsed time: the algorithmic and wait-time methods.
rm
CPU occupancy time, also known as pseudo-elapsed time, approximates how long a job would be in the system if it ran all by itself, without the influence of the multiprogramming mix. Two basic methods compute pseudo-elapsed time: the algorithmic and wait-time methods.
Algorithmic Approach
The algorithmic approach uses the following type of formula to approximate the occupancy time:
Occupancy Time = CPU Time + (dms * Number of I/Os)
- dms
- is the average duration in milliseconds of one input/output (I/O) operation.
The error in this approach, however, is that the user is charged the full system rental, when the job may have required only minimal system resources. For example, suppose you submitted a tape copy program that would run for ten minutes if it was the only job in an environment with thirty tape drives and numerous disk drives. You would pay for much more than the job was consuming.
Another problem with the algorithmic approach is its inability to consider other important characteristics such as device speed, overlap, channel separation, and access methods. For instance, an indexed-sequential update operation and a sequential read operation would be charged exactly the same amount because they are one I/O operation each. In actuality, however, the index-sequential is much more time-consuming than the sequential operation.
Wait-Time Approach
The wait-time approach is based on the operating system measurement wait time. Wait time is calculated with the following formula:
Occupancy Time = CPU Time + Wait Time
- Wait Time
- is the amount of time that the task was voluntarily waiting for its own I/O function to complete.
In theory, the wait-time approach is the most accurate. However, measuring wait time accurately and equitably presents another problem. In most uses of this approach, wait time is not determined solely as the amount of time the task required but includes factors such as access contention that is not the fault of the task.