Quality Assurance: Software Quality Assurance Made Easy pdfdrive com

Software Reliability Models
There are at least 200 software reliability models available since the 1970s.
However, no single model can satisfy the software complexity thus, assumptions
and constraints are included to quantify the process. Most models have
assumptions, a mathematical function, and factors that relate to software
reliability. In most cases, the mathematical function is logarithmic or
exponential.
The software modeling techniques fall under two categories: prediction and
estimation. Both techniques accumulate and observe failure data and analyze it
with statistical inference. However, the prediction model uses historical data and
used before the test or development phases. It is able to predict software
reliability in the future. On the other hand, the estimation models uses current
data and used at a later software life cycle. It can predict software reliability at
some future time or at present.

---

Software Reliability Metrics
Measurement is easy in most engineering fields. However, the same is not true
in software engineering. Measurement is frustrating and there is really no good
measurement method for software reliability. It is difficult because no one has a
good understanding of software. In fact, there is no appropriate measurement
method for software reliability unless experts are able to define the aspects
related to it.
Currently, there are four categories for software reliability measurement namely,
product metrics, project management metrics, process metrics, and fault and
failure metrics.
Product Metrics
Software size is a reflection of reliability, development effort, and complexity.
Intuitively, the Lines of Code (LOC) measures software size. However, there is
no standard method for measurement. Usually, non-executable statement and
comments do not count. Only the source code does. This way of counting cannot
compare with software written in other programming languages.
The function point metric measures the functionality of software development
based on interfaces, inquiries, master files, inputs, and outputs. It estimates
software size when functions are already identifiable. It measures business
systems but are not scientific.
Complexity, on the other hand, relates directly to software reliability. There are
metrics that measure complexity by determining the control structure of the
program and simplifying the code graphically.
Lastly, test coverage metrics estimate reliability and fault by testing the software
based on the assumption that once a part of software is tested, the software is
reliable.