Software Engineering

199
11.3 
Performance allocation
11.3.1 
Allocate performance budgets
The execution performance of each function must be specified and allocated among 
its constituent subfunctions. Execution performance addresses the time that a soft-
ware function will take to be executed on the specified computing environment. The 
duration of software functions should not account for any delay associated with 
external users or systems interactions. The initial allocation of performance require-
ments should be considered budgets until the design of the software functional 
architecture is complete. The resulting functional specifications must then establish 
the performance requirements associated with every software function. While the 
functional architecture is evolving, the performance budgets may undergo continual 
change as the software implementation subject-matter experts refine their expecta-
tions associated with what can be achieved and delivered.
Functional timing may be specified as a constant or variable duration. Constant 
durations should be used when functional timing will be relatively consistent. 
Variable duration specifications should utilize probability distribution functions to 
represent the random execution duration of a function.
Analysis of the behavioral model may provide insight into the critical path of 
functional execution. Critical-path analysis is a powerful approach for identifying 
bottlenecks in highly concurrent systems, but typically requires detailed domain 
knowledge to construct the required event graph that identifies the dependencies 
and timing among events in the software behavioral model. The criticality of a par-
ticular function can be determined as the ratio of the duration of the function to 
the total time of the critical path. This metric gives a quick summary of the most 
important functions within a data processing sequence and contributes to the crit-
ical path. This analysis identifies the areas within the functional architecture that 
offer the largest opportunities for improving overall software performance.