Edge-fog-cloud continuum applications : analytical and hierarchical models for availability and performability evaluation

Abstract

Connectivity has introduced significant transformations in our society, requiring the continuous improvement of Quality of Service (QoS). The rise of various emerging tech- nologies has created a demand for networks capable of low-latency communication to facilitate real-time data processing. As our reliance on these technologies grows, it be- comes increasingly important to address the latency issue. While cloud computing envi- ronments offer high availability, reliability, and performance, they may not be suitable for applications that require low latency, such as disaster risk minimization, smart-traffic management, and crime prevention. For instance, numerous lives could be lost if a disaster risk minimization service delays in providing alerts about an earthquake. To overcome the challenges posed by latency and enhance computing capabilities between the cloud and edge devices (e.g., controllers, sensors, and smartphones), two complementary paradigms, namely edge and fog computing, have been proposed. However, evaluating the depend- ability and performance of distributed computing environments remains a concern due to the numerous challenges involved in supporting the required QoS of these systems. There- fore, this study aims to investigate the dependability of edge, fog, and cloud computing environments by assessing their availability and the resulting impact on performance. Additionally, we propose analytical and hierarchical models that facilitate scalability and capacity planning in these computing environments. The metrics considered in this study include availability, K out of N availability, capacity-oriented availability, as well as per- formance metrics such as utilization, response time, waiting time, and discard rate. Our proposed models serve as valuable tools for researchers, system designers, and practi- tioners in the field of edge-fog-cloud environments. By understanding the behavior and limitations of these systems, we can enhance their design, operation, and maintenance, ultimately leading to more reliable, efficient, and resilient infrastructures.