Availability and capacity modeling for virtual network functions based on redundancy and rejuvenation supported through live migration

Abstract

The success of server virtualization and cloud computing led to a subsequent network virtualization requirement, because the flexibility achieved by virtualized hardware resources could be impaired by static network interconnections. Network virtualization refers to the ability to execute virtual instances of routers, switches, and links on top of a physical network substrate. So, multiple virtualized networks can co-exist in a shared network infrastructure. Technologies such as Software-Defined Networks, Network Function Virtualization and Service Function Chaining have been launched to enable the replacement of traditional network hardware appliances by softwarized Virtualized Network Function (VNF)s chains. As a consequence, virtualized networks represent additional obstacles to the provision of high availability services, because it results in more layers of software: the increasing number of software components required to run virtualized systems also increases the number of possible failures. This thesis designed and evaluated a set of stochastic models to improve virtual network functions provision considering metrics of availability and capacity. The models can represent high availability mechanisms, such as redundancy and software rejuvenation, allowing to estimate the behavior of the studied metrics facing these mechanisms. The adopted methodology encompasses the assembling and configuration of high available cloud computing infrastructure. The implemented cloud supports the provision of redundant virtual network functions and service function chains, enabling the measurement of parameter values that were injected in the designed models. In order to show the applicability of proposed solutions, a set of case studies are also presented. The results demonstrate the feasibility in providing high available Virtual Network Functions and Service Function Chains in a cloud infrastructure for the studied scenarios. Such results can be useful for telecommunication providers and operators and their heterogeneous infrastructures.