Stochastic Model for Availability Analysis of Service Function Chains using Rejuvenation and Live Migration

Abstract

Service function chains aim to replace proprietary and expensive appliances by similar ones provided through virtual services running over some virtualization technique. Such a replacement put pressure over availability because proprietary devices are hardware-based, i.e., more reliable. To tackle this issue, we propose a modeling approach to produce recommendations of high available cloud infrastructures responsible for the provision of VNF chains. Our proposal is based on rejuvenation of both cloud infrastructure and SFC adopting conditional virtual machines live migration. We performed SFC migration experiments in an assembled high available private cloud based on openstack to measure live migration times and inject these values into proposed models. The results of the presented case studies show that so-called five 9's availability can be reached in a 2N redundant environment.