9.4 Simulation of Software Process

Sadaf Mustafiz has built a software process model [50] for SVM. The development process is modeled as several tasks, each of which is ``an entity (a real object that exists and has an extended lifetime)'' (Sadaf Mustafiz).

Table 9.1: Rounds and tasks in a software development process

Round	Task	Hours
1	developing code	12
1	developing tests	8
1	running tests	1
1	analyzing problems	3
2	developing code	6
2	developing tests	4
2	running tests	1
2	analyzing problems	2
3	developing code	3
3	running tests	1
3	-- passed --

To model those tasks with DCharts, each of them corresponds to an active state. For example, a software development process consists of three rounds, as shown in Table 9.1. In the first round, the following tasks are scheduled sequentially: developing code, developing tests, running tests and analyzing problems. The distribution of hours among them is: developing code takes 12 hours, developing tests takes 8 hours, running tests takes 1 hour, and analyzing problems takes 3 hours.

Because problems are discovered in round 1 during the running tests task, another round must be added to fix those problems. Before round 1 finishes, the task of analyzing problems is undertaken to analyze the problems that are to be fixed and the cost to fix them in the next round.

Round 2 is based on the results of round 1. It strives to fix the problems discovered in round 1, as well as to improve the functionality of the system. The development becomes faster: developing code takes 6 hours, developing tests takes 4 hours, running tests takes 1 hour, and analyzing problems takes 2 hours. Because there are still problems found in the running tests task, round 3 is required, which only aims at fixing those remaining problems.

In round 3, the problems are fixed, and all the tests are passed. The development process successfully finishes.

Figure 9.2: Traces of the software process model simulation

Sadaf Mustafiz has modeled this process with DCharts. The model is simulated with SVM. The output trace is written in text files. The plots of the output trace is shown in Figure 9.2.

More information about the above software process model can be found at the Modeling and Simulation Based Design course homepage:

http://moncs.cs.mcgill.ca/people/hv/teaching/COMP762B2003/