A Novel Reactive Survivable Virtual Network Embedding Scheme Based on Game Theory

In this paper, we tackle the virtual network (VN) embedding problem within Cloud's backbone network by taking into consideration the impact of physical equipment outages. Our main focus is to improve the Cloud provider's (CPs) revenue by: 1) maximizing the acceptance rate of VNs within the Cloud's backbone and 2) minimizing the penalties induced by service disruption due to the hardware outages. This optimization problem is non-linear multi-objective and it has been proven to be NP-hard. To cope with this complexity, we propose an advanced coordination game for VN embedding (Advanced-CG-VNE). In this mapping (i.e., embedding) game, fictitious players are playing on behalf of the CP in order to maximize the turnover. The decision makers cooperate in aim to converge to a Nash Equilibrium that we prove the existence of and the matching with a social optimum. Two variants of Advanced-CG-VNE are proposed according to the virtual links embedding approach. The first one, denoted by Advanced-CG-VNE-unsplittable, embeds each virtual link in only one substrate path. The second variant, denoted by Advanced-CG-VNE-splittable, dispatches the required bandwidth of a virtual link among a set of substrate paths. Advanced-CG-VNE does not allocate any backup to handle service interruption caused by hardware failures. Our proposal adopts preventive and reactive mechanisms to palliate substrate failures. Based on extensive simulations to gauge the effectiveness of Advanced-CG-VNE, the obtained results show that our proposal outperforms the most prominent related strategies in terms of: 1) rejection rate of VNs; 2) rate of VNs impacted by physical failures; and 3) CPs turnover.