Mobile Edge Vertical Computing over 5G Network Sliced Infrastructures: an Insight into Integration Approaches Roberto Bruschi * , Raffaele Bolla *£ , Franco Davoli *£ , Anastasios Zafeiropoulos § , Panagiotis Gouvas § * S3ITI National Lab., CNIT, Italy £ DITEN, University of Genoa, Italy § UBITECH, Greece Abstract–5G is designed to leverage on network softwarization technologies, like Network Functions Virtualization (NFV) and Mobile Edge Computing (MEC), to expose customized network instances and resources, at the edge of the infrastructure to vertical stakeholders. Most of the 5G success will depend on the ability to attract vertical stakeholders acting in the cloud, enabling them to smoothly port cloud applications to 5G, and to add performance and cognitive capabilities not supported in cloud environments. To this end, this paper provides an insight on the possible architectural approaches to fully integrate Vertical Applications (vApps) into the 5G infrastructure. The paper follows a top-down approach. First, it provides an outlook on the state-of-the-art in cloud application design, and on the MEC and NFV new capabilities. Then, on this basis, the analysis is devoted to identify integration issues not yet fully addressed in standard specifications. Two alternative architectural approaches are discussed. I. INTRODUCTION A key 5G objective resides in the enablement of a new class of vApps with heterogeneous and extremely challenging requirements [1-2]. To this end, the 5G community is embracing well-known technologies, like NFV and MEC. Both these frameworks are based on the unrestrainable “softwarization” process, which is going to transform network operators’ infrastructures into distributed datacenters with advanced virtualization and software-driven capabilities. MEC and NFV frameworks will have clear and well-separated objectives. As stated by the ETSI MEC working group (WG) in [3], “MEC uses a virtualisation platform for running applications at the mobile network edge. NFV provides a virtualisation platform to network functions.” As the infrastructure requirements of both approaches are quite similar, the use of a converged virtualization infrastructure would be beneficial. These frameworks will be key enablers for flexible customization of mobile network slices to the needs of vApps [4-5] and their provision with full network-awareness and zero-perceived latency. Radically new applications [6-7] can be made viable through the joint adoption of these technologies. As defined by 3GPP and MGMN [2,4], a network slice is a logical end-to-end network providing specific 5G network services, offered as-a-Service by a Telecom Service Provider (TSP) to Over-The-Top (OTT) players, such as Vertical Industries. The TSP should support multiple network slices from different OTT players at the same time, and dynamically realize each of them through the composition of shared/isolated 5G functions’ instances [8]. Notwithstanding the high complementarity between NFV and MEC, their integration and interplay in the 5G ecosystem is still largely unexplored. The objective of this paper is to identify the possible approaches, to highlight their main advantages and drawbacks, as well as to introduce relevant integration issues. To this end, the paper will follow a top-down approach: starting from current trends in cloud application design, we will identify the main evolution routes towards 5G, mapping them onto the emerging technological paradigms offered by the 5G infrastructures. From this mapping, two possible architectural approaches to integrate vApps with edge computing and NFV facilities are finally introduced and discussed, by outlining the roles of involved stakeholders, and the main induced benefits and drawbacks. The remainder of this paper is organized as follows. Sect. II provides a short discussion on cloud vApps, and on how they have to evolve in the 5G scenario. Sect. III introduces the main building blocks composing the 5G architecture, and Sect. IV discusses possible integration approaches. Conclusions are drawn in Sect. V. II. FROM CLOUD TO 5G-READY APPS In cloud computing, state-of-the-art software engineering trends are based on the microservice concept. To achieve high scalability and agility levels, applications are decomposed into a mesh of “cloud-native” microservices, each one with specific and “small-scope” processing objectives, instantiated even multiple times, and packaged on independent virtual execution environments [9- 10]. A central entity, named “application orchestrator,” is in charge of realizing the application business logic, by managing the lifecycle and the mesh interconnection of such microservices over cloud resources. Depending on the nature of the application, the orchestrator can alter the application graph [10], by adding/removing: (i) new types of microservices to enable/upgrade specific application functions, (ii) instances of existing types of microservices to scale the overall application processing capacity, where needed.