Virtualizing Network Services & Cloud Reliability

In a communications network, several physical appliances are used to manage and operate the network. A few examples of these appliances are equipment used as load balancers, firewalls and intrusion detection systems. With the advance of virtualization and cloud computing, these physical appliances are being replaced by virtual servers, in a process called network function virtualization (NFV). Using NFV has several advantages over employing physical appliances: increased modularity, reduced costs and easier scaling. In this talk, I will focus on a related aspect: reliability. Years of research and engineering efforts have been spent to ensure that the physical appliances will perform their duties in a reliable way. Although this work has resulted in high-quality equipment, it is fairly unknown how the same quality and reliability can be achieved using the cloud when virtualizing network functions. To this end, I will cover recent related work in cloud reliability, including some of our studies, identifying root causes and potential improvement areas. I will also discuss some recent efforts to ensure high reliability when deploying applications in the cloud.