Availability-Based Resilient Virtual Optical Network Mapping
Abstract
With the growth of bandwidth-demanding and dynamic network services such as cloud computing, big data, and machine learning applications, a key challenge for network operators is how to efficiently deploy the dynamic network infrastructures to support these various services and applications. Network virtualization provides a feasible way to share computing resources and network resources among virtual network requests. Availability is a key factor in the service level agreements (SLAs) when a network service provider provides the computing and network resources to the customers. A virtual optical network (VON) request may consist of many services which depend on each other, and the failures of single or multiple network elements may lead to the inability of the VON to support these services. Therefore, it is important to evaluate a VON based on its availability, where a VON is defined to be available only when all of its virtual nodes and virtual links are available at the same time. We first study the availability-guaranteed VON mapping problem over the physical optical network where the network elements suffer independent failures. The objective is to minimize the total link cost of a VON mapping while guaranteeing the VONs availability requirement. Based on the proposed availability analysis model, we propose an ADAPTIVE algorithm using a new cost metric to provide selective dedicated protection. We also provide shared backup path protection for VONs by finding the maximum availability VON mapping based on a group node-weighted Steiner tree solution. We then address high-availability VON mapping over the physical optical network where its physical links may suffer correlated failures. We propose the Group Steiner tree problem under correlated cost (GSTP-CC) to determine the maximum availability of the VON mapping without any protection, and a dedicated protection mechanism is provided to meet the availability of each VON, if needed. Finally, we focus on availability-guaranteed service function chain (SFC) mapping in interdatacenter networks. We propose different coordinated protection mechanisms that adopt both backup path protection in the network and VNF replicas at nodes to guarantee a SFCs availability while minimizing the total computing and link resources.