An experimental study on dynamic network reconfiguration in a virtualized network environment using autonomic management

In a virtualized network, network virtualization is achieved by creating multiple instances of virtual routers over existing physical routers. Critical aspects of network virtualization are resource allocation and resource re-allocation, especially in the case of router failures or re-allocation of routers due for maintenance. In this paper, we present a dynamic network reconfiguration scheme that can provision a new virtual network and recover it from router failures using autonomic management. In order to provide a highly available virtual network, our approach considers replacing failed routers with the most suitable standby routers. We have setup an experimental virtualized network on the GpENI Virtual Network Infrastructure (GpENI-VINI) testbed in which virtual routers are geographically distributed. Our autonomic management function invokes the replacement module in the event of a router failure. Our study considers not only end-to-end metrics (latency, bandwidth, loss-rate) but also the routing table convergence time to understand the transient behavior in this autonomic management environment.