An adaptive load balancing scheme for evolving virtual networks

Abstract

An algorithm to adapt dynamically virtual networks to additional resource requirements is proposed and evaluated. The optimization is achieved while balancing load and avoiding fragmentation in the infrastructure (often referred as substrate or physical network). The algorithm focuses on virtual nodes requiring more resources by extending their allocations and maintaining their connectivity (even if the node is migrated) to other resources while tidying up (or consolidating) the infrastructure. The algorithm outperforms existing approaches.

A self-stabilizing framework for dynamic bandwidth allocation in virtual networks

“…To do so, a substrate path connecting the substrate nodes hosting src l and dst l , and having enough available bandwidth (> req l ) should be found. We define the cost of embedding a virtual link l as the sum of costs of the substrate links hosting l, as in [14]. The aim is to find the most cost effective substrate path, in a distributed, and self stabilizing manner.…”

“…We adjusted the C++ simulator used in previous work [14] to fit our scenario: substrate nodes with local view, and a round per round algorithm execution. The GT-ITM tool ( [15]) is used to generate random topologies of the substrate and VN networks.…”

“…the two steps duration, called Convergence T ime. We set K to one, and T imer to 100 rounds, and make evaluation for three SN configurations: ALS = 78%, 64% and 50%, where ALS is the average substrate links saturation, as defined in [14]. We measure the time in seconds for more precision.…”