Bandwidth-Aware Allocation of Resilient Virtual Software Defined Networks Rafael L. Gomes a,b,1,∗ , Luiz F. Bittencourt, Edmundo R. M. Madeira. b , Eduardo Cerqueira. a,c , Mario Gerla. a a University of California (UCLA), Los Angeles, USA b University of Campinas (UNICAMP), S˜ ao Paulo, Brazil c Federal University of Par´ a (UFPA), Par´ a, Brazil Abstract Currently, it is hard to imagine our lives without the Internet, where services are accessed and shared by billions of users every day. However, even after many years, the Internet cannot guarantee Quality of Service (QoS) for the main services to current and future clients. To deal with this problem, clients establish Service Level Agreements (SLAs) with their Internet Service Providers (ISPs), including resilience parameters. The Internet is recognized to be resilient enough for many services, but it is still sensitive to failure events that affect the performance of these services. Software Defined Networks (SDNs) together with Virtual Network (VNs) approaches aim to enhance the management, planning and resource usage of networks. When both approaches are mixed together, we have Virtual Software Defined Network (VSDN). However, the allocation of VSDNs considering resilience issues is still an open issue. Within this context, this article presents an algorithm for VSDN allocation, called Bw-Risk-Ratio, that considers resilience factors, as well as it deploys the VSDN according to QoS parameters defined in the SLA. Experiments using a real network topology show the effectiveness of the algorithm to deploy the VSDN resilient to failure events when compared to existing solutions. Keywords: Virtual Networks Allocation, Software Defined Network, Service Level Agreement, Resilience. 1. Introduction Nowadays, Internet services over communication networks play a vital role in our modern private, corporate, and institutional lives. As a consequence, users become frustrated when the Internet access fails and/or when application quality level drops. Currently, users share lots of content, which accounts for an expressive part of Internet traffic. In general, Internet traffic is composed of a variety of applications, which need high bandwidth and are sensitive to delay and packet losses. Usually, to pursue Quality of Service (QoS) between companies/clients and Internet Service Providers (ISPs), a Service Level Agreement (SLA) is established to specify network parameters to be fulfilled by the ISPs [1], where resilience is a key requirement to avoid service disruption of networked applications. Network resilience has been defined as the capacity of the network to provide a minimum specified level of service in situations of faults in standard operation [2]. The concept of resilience involves not only incorporating reactive actions to manage post- event consequences, but also pre-event strategic planning. Moreover, reliability and traffic congestion concepts are important aspects of resilience to be considered during the service provisioning process [3]. Resilience and Bandwidth (Bw) are metrics closely related to the QoS, and they have direct impact in the user’s impression/satisfaction about the service provided by the ISP. However, the current Internet design does not support both QoS and resilience guarantees, being necessary to improve the management and planning of ISPs to allow a better access to the Internet [3]. The implementation of these management and planning features involves ensuring SLA specifications, and thus impacts the user’s experience during possible failure events in the ISP. ∗ +55 19 982247245 1 rafaellgom@ic.unicamp.br Preprint submitted to Elsevier January 21, 2016