Exploring Emerging Industry Trends for Large-Scale Software System Performance Predictability: The Role of Apache and Xen
Zahra Nikdel, Stephen W. Neville · 2024
Modern societies critically depend on numerous large-scale cloud-deployed distributed software systems (LDSSs), whether in social media and on-line games, banking and finance, business-to-business systems, data science and AI, etc. This core reliance is accelerating the software engineering need to capital-“E” Engineer system such that they behavior predictably in the real-world at their full operational scales. This work applies Monte Carlo simulation to assess and quantify the impacts of recent industry technology trends on LDSS performance predictability. Specifically, we show that technologies such as Apache Storm, Apache Spark and Xen's recent real-time operating system (RTOS) hypervisor introduction, work to produce more predictable LDSS run-time behaviors. The implications of these observations on LDSS management approaches, such as Kubernetes and Docker Swarm, are then discussed. All simulations are conducted via OMNet++ and its INET network framework for an industry held LDSS against a selected range of commonplace scenarios. To our knowledge this is the first work to seek to quantify emerging industry solutions against the specific concern of their impacts on LDSS run-time performance predictability.