Profiling of Real-World Web Applications

摘要

Non-functional requirements of enterprise web applications, like performance, are usually assessed and analyzed in simulated environments before being released into production. However, it is not easy to simulate a real-world environment and the effective use of the system, leading to poor and expensive performance data collection. Moreover, in agile methodologies, where development is focused in the fast time to market and getting early feedback from end-users, upfront estimation and forward thinking about scalability is not in the top priorities. This constrains even more performance analysis and tests, as developers are only aware of performance issues when critical feedback from production systems is given back to development. This commonly leads to enterprise web applications with scalability problems, and low responsiveness. This paper presents a structured way of giving continuous and real world performance feedback to developers of enterprise web applications. By having early access to real performance metrics, developers easily detect stress points in applications, allowing for timely tuning actions, before reaching critical conditions for end-users. Metrics also help developers assessing the impact of changing intensively used parts of existing applications.