The Impact of Soft Resource Allocation on n-Tier Application Scalability

摘要: Good performance and efficiency, in terms of high quality service resource utilization for example, are important goals a cloud environment. Through extensive measurements an n-tier application benchmark (RUBBoS), we show that overall system is surprisingly sensitive to appropriate allocation soft resources (e.g., server thread pool size). Inappropriate can quickly degrade significantly. Concretely, both under-allocation over-allocation lead bottlenecks other because non-trivial dependencies. We have observed some non-obvious phenomena due these correlated bottlenecks. For instance, the number threads Apache web limit total useful throughput, causing CPU C-JDBC clustering middleware decrease as workload increases. provide practical iterative solution approach this challenge through algorithmic combination operational queuing laws measurement data. Our results plays central role scalability complex systems such applications environments.