Maximum User Concurrency and Blocking Probability for Managed and Open Access Applications

摘要

Maximum number of concurrent users is one critical factor for application design and capacity planning. It has a direct impact on the settings of key system parameters such as connection pool sizes, maximum number of threads and buffer sizes. Depending on the type of the workload, there are a number of theories and models that study user concurrency. Unfortunately, developers are not always aware of the appropriate approaches. Instead of systematically determining these parameters, we often see teams resort to empirical or even trial_and_error methods. This guesswork often leads to either excessive resources allocation or unnecessarily high blocking probability for users in production systems. In this paper, we discuss two typical use cases: managed access and open access applications. We describe the different theories and models applicable to these use cases and present the results of our experiments.