Letting a user generate tasks in a multitasking computer environment will result in a serial stream of tasks, and thus a less than optimum use of the available computing and storage devices. When presented with a number of concurrent tasks, the multitasking operating system will be able to schedule the tasks in order to optimally fill the resources and so improve the throughput.
In a programming environment, serial execution is often initiated by the well-known Unix tool Make. The concurrent equivalent is the newly developed CASE tool Cake (Concurrent Make). This CASE tool is developed to describe a network of tasks and present the operating system with a flow of concurrent tasks. The overall response time of the network is less than the corresponding serial execution of the tasks.
Measurements in an OS/2 environment show that for a characteristic program development project, the overall throughput increases when offering the tasks in parallel instead of in sequence. By converting a standard Make description me in a Cake description me, the time needed to perform all tasks can be decreased by 25%. Three concurrent tasks are enough to almost optimally use the available resources in a single processor environment.
Modelling the Cake system using a simple queueing model results in a tool that can be used to predict the behaviour of the Cake performance when changing parameters such as the number and speed of the processing and storage devices. Comparison of the model results and measurements confmn that the model correctly predicts the behaviour of the system.