Facts and Fiction about Developing Reliable Software
|
|
MetroCon 2007 |
|
“Innovating for Society” |
|
Emerging Technologies |
|
For the last several decades there have been many ideas on how to develop reliable software. Many of these ideas are based more on opinion than on hard facts. Some of these opinions have been repeated so many times that the software community simply assumes that they are based on proven fact. Since 1993, this author has been collecting real data from real companies in an effort to quantitatively separate fact from fiction. This author has collected more than 90 complete sets of latent software defects from a spectrum of real organizations developing non-academic mission critical applications. This paper will expose the quantitative proof surrounding popular fiction as well as the top 10 list of the most effective and efficient software development practices to embrace. It will also cover the biggest obstacles to avoid when developing reliable software. Finally a method to predict how many defects will escape the development and testing process will be illustrated. This method was developed by examining what each of the 90+ software organizations did and did not do while developing and testing and how many normalized defects subsequently escaped. If you answer a survey the method will yield one of seven defect density ranges from "World Class" to "Ugly". This range is then used to predict defect density and then ultimately the total escaped (latent) defects. This method includes a ranking of cost, time and impact for each of the development practices thus allowing one to identify the most efficient quantitative defect reduction scenario. Finally, the author will discuss the prerequisites required for each development practice. My benchmarking illustrated that often times a development practice is ineffective because it is implemented too early or too late or without the necessary prerequisites. A prerequisite tree will illustrate the most effective order of the development practices as well as the most effective order to eliminate development obstacles. |
|
About the Speaker: |
|
Ann Marie Neufelder is the President of SoftRel, LLC which is located in Lewisville, Texas. Ann Marie is a 1983 graduate of Georgia Tech and has spent the last 24 years developing and managing software as well as executing software reliability prediction and management on real world applications. Since 1988 Ann Marie has taught more than 2000 professionals practical software reliability. Ann Marie has authored a Military guidebook, an industry guidebook, a toolkit, a software package and numerous articles and papers on the subject of software reliability. In addition Ann Marie has a US patent for a method to predict the time to Rastor Image Process. |
