A stochastic reward net-based assessment of reliability, availability and operational cost for a software-defined network infrastructure

Abstract

The networking infrastructure of a software-defined network (SDN) requires further study to achieve continuity and high availability of data transactions for cloud computing services. However, various types of failures on links or system components are encountered because of high-speed and complicated structures of hosts and network devices. This study examines the specific characteristics and impact of various failures on a typical SDN infrastructure. We propose a stochastic model using stochastic reward net by incorporating hardware failures (of hosts, switches, storage, and links) and software failures [virtual machines (VMs)]. The system model is analyzed based on steady-state availability under default parameters. Comprehensive sensitivity analyses are conducted to study the system behaviors with respect to different major factors of impact. A reliability analysis is also conducted to pinpoint the role of VM migration in extending the system lifetime. Furthermore, operational cost is analyzed in detail to identify the main sources and to observe the effect of major system parameters on costs over time. The analytical results are verified by a developed discrete-event simulation module based on CloudSim Plus simulation open-source package. This study provides a helpful basis for network design and implementation in SDN infrastructures.