A compositional modelling approach for live migration in Software Defined Networks

Abstract

Recent advances offered by Software Defined Networking and virtualization techniques are creating the favourable conditions to design and develop Future Networks where network functions and services can be virtualized, dynamically instantiated and moved across networks. The ability to dynamically allocate virtual nodes across distributed physical hosts and even more the live migration of the Virtual Machines which perform such virtual network functions are driving current network infrastructures towards “programmable” networks. To be adopted as a deployable feature on a Carrier's Network, live migration performance need to be analysed and tested. Therefore a compositional modelling approach could provide early measures by evaluating the impact of these new technologies on the performance of Future Network systems. The main contribution of the paper is the definition of a general modelling framework to integrate simple models representing the main components and features of a Future Network architecture. Starting from this model composed by atomic sub-models, we conduct the performance analysis of the live migration of a single Virtual Machine between two hosts located in different networks. The Mobius tool has been used for developing the model as well as for studying its behaviour and performance. Finally, some simulative results are provided in order to show the feasibility of this approach.