Performance measurement as a tool for software engineering

Abstract

Some software development teams regard software performance measurement as a mere luxury. When it happens, it often tends to be infrequent, insufficient and subjective. Countless software projects were sent into an uncontrollable spiral of poor management and unsatisfactory results. By revisiting old ideas and policies, many companies have turned themselves around. To ensure that software engineering does the same, technologies and procedures have to be reevaluated. The fact that many companies have decided to cut costs on technology expenditure necessitates software development teams to look for alternative options for deploying high performance software systems. As many companies are moving into the electronic era and evolving to the next stage of evolution, electronic commerce, the more important it has become to apply these concepts on Internet development projects and procedures. The Internet market has shown that two software providers are aiming for worldwide domination of Internet server deployment, being Microsoft and Apache. Currently, the Apache web server is the most commonly used server on the Internet today (60%), with Microsoft's Internet Information Server (25%) in a strong second place. The need for higher throughput and better services is getting more with each passing day. It increases the pressure on these two software vendors to provide the best architecture for their clients' needs. This study intends to provide the reader with an objective view of a basic performance comparison between these two products and tries to find a correlation between the performance tests and the products' popularity standings. The tests for this study were performed on identical hardware architectures with one difference, being the operating system. By comparing the costly proprietary Microsoft solution with its cheaper open source rival, Linux, certain opinions were tested. Would a product developed by a software company that invests millions of dollars in their products perform better than this free-for-all solution, or would the selfless inputs of hundreds of developers all over the world finally payoff through the creation of the world's best Internet server? The results of these tests were evaluated through formal statistical methods, providing overall comparisons of several common uses of web servers. These results were implemented in a small field test to prove the findings in practice with some interesting outcomes in terms of supportive technologies, new rapid application development (RAD) tools and data access models. This research in itself will not change the mind of any Internet programmer. What it hopes to achieve is to demonstrate software engineers that current processes and methods of developing software are not always the right way of doing things. Furthermore, it highlights many important factors often ignored or overlooked while managing software projects. Change management, process re-engineering and risk management form crucial elements of software development projects. By not adhering to certain critical elements of software development, software projects stand the chance of not reaching their goals and could even fail completely. Performance measurement acts as a tool for software engineering, providing guidelines for technological decisions, project management and ultimately, project success.