Time-efficient distributed virtual network embedding for round-trip delay minimization.

Virtual Network Embedding (VNE) aims at determining the placement of virtual services both internally in the edge/cloud infrastructure (EC) and distributed, among distinct ECs, under several resource and network constraints. With regard to Network Function Virtualization technology, a virtual service is referred to as Virtual Network Function (VNF). Modern 5G applications related to the Internet of Things (IoT) are deployed as interconnected VNFs with specific execution sequence in the form of a Service Function Chain (SFC), composing a Virtual Network (VN), and the execution result has to be returned back to the original end-device. Thus, the minimization of the round-trip delay is of major importance to guarantee the performance requirements of IoTbased applications. Additionally, in such a dynamic environment, the re-optimization of VNE is often required to meet the network delay requirements due to end-device mobility, and other orchestration constraints related to the limited edge resources. In this paper, we propose a distributed VNE (DVNE) approach, which focuses on minimizing the round-trip delay. Initially, we introduce an algorithm, namely sV-VNM, that associates every VNF of the VN with a set of candidate ECs that can be hosted onto. Given this initial mapping, an efficient path-based algorithm, namely kSP-DVNE, is proposed to provide the DVNE solution, in polynomial time. Evaluation results show that the sV-VNM algorithm maps the VN onto the ECs, with decreased resource utilization within the infrastructure, while the kSP-DVNE achieves optimal or near optimal solutions regarding the round-trip delay minimization, with a significant reduction in the execution time, compared to existing relevant approaches in the literature.