One of the main goals of 5G networks is to support the technological and business needs of various industries (the so-called verticals), which wish to offer to their customers a wide range of services characterized by diverse performance requirements. In this context, a critical challenge lies in mapping in an automated manner the requirements of verticals into decisions concerning the network infrastructure, including VNF placement, resource assignment, and traffic routing. In this paper, we seek to make such decisions jointly, accounting for their mutual interaction, and efficiently. To this end, we formulate a queuingbased model and use it at the network orchestrator to optimally match the vertical's requirements to the available system resources. We then propose a fast and efficient solution strategy, called MaxZ, which allows us to reduce the solution complexity. Our performance evaluation, carried out accounting for multiple scenarios representative of real-world services, shows that MaxZ performs substantially better than state-ofthe-art alternatives and consistently close to the optimum.
INTRODUCTION5G networks are envisioned to provide the computational, memory, and storage resources needed to run multiple third parties (referred to as vertical industries or verticals) with diverse communication and computation needs. Verticals provide network operators with the specification of the services they want to provide, e.g., the virtual network functions (VNFs) they want to use to process their data and the associated quality of service.Mobile network operators are in charge of mapping the requirements of the verticals into infrastructure management decisions. This task is part of the network orchestration, and includes making decisions concerning (i) the placement of the VNFs needed by the verticals across the infrastructure; (ii) the assignment of CPU, memory and storage resources to the VNFs; (iii) the routing of data across network nodes.These decisions interact with each other in ways that are complex and often counterintuitive. In this paper, we focus on the allocation of computational and network resources, and make such decisions jointly, accounting for (i) the requirements of each VNF and vertical; (ii) the capabilities of the network operator's infrastructure; (iii) the capacity • S. Agarwal is with IIT Guwahati, India. F. Malandrino and C.-F. Chiasserini are with Politecnico di Torino, Italy and CNR-IEIIT, Italy. S. De is with IIT Delhi, India. • A preliminary version [1] of this work was presented at the IEEE INFOCOM 2018 conference.