Edge computing enhances cloud architectures through the integration of distributed platforms that operate closer to end-users. With dynamically changing network topologies, user mobility, multiple access networks, demand-driven service orchestration, and consequently several distributed service instances that are available, methods for selecting edge resources must adapt to these conditions. This consideration is particularly relevant for edge computing, where target applications often have strict Quality of Service (QoS) constraints. This work proposes an edge resource selection scheme that considers several QoS parameters, computing load, and service usage cost. These metrics are employed to derive a selection score through four distinct algorithms. The overall performance of each algorithm is evaluated by scoring examples, a sensitivity analysis, and in a system-level simulation of a fifth generation (5G) cellular network with user mobility and multiple Multi-Access Edge Computing (MEC) platforms. The results show that all approaches perform well under unambiguous QoS and load conditions, i.e., very good or very poor, but some lose accuracy under moderate, but still acceptable, conditions. For the goal of cost-optimized selection while meeting QoS constraints, one novel method presented is significantly superior to the others. Used in simulation, it generates 12% more computing load on cheaper platforms than its closest competitor, and maintains significantly lower loads on high-cost platforms.

The Multi-Access Edge Computing (MEC) paradigm enables services that have strict network requirements like a maximum tolerable latency or a minimum bandwidth requirement. This consequently leads to a tight coupling of networks and computing, since the service can only function correctly if its communication requirements are met. Numerous existing strategies for MEC service deployment or resource selection put the characteristics of computing demands correlated with the free capacity of nodes or inter-service dependencies in their focus. In contrast, our work is a complementary attempt to enable especially the consideration of the strict communication needs of MEC services. It shows how edge features, i.e., knowledge of edge network conditions and service requirements, can be used to support resource selection and orchestration strategies. We consider and discuss how our model meQ can be applied in real communication systems as well as in heterogeneous networks, where clients may have multiple uplink connections. To enable communication-optimized selection strategies, meQ consists of two major components. A sharable, graph-based network representation and online searches to determine resources. The model benefits from existing work and limits the search space with the use of edge features. A performed simulation study demonstrates the benefits of meQ. The results show that it is possible to significantly reduce the search space, leading to lower search efforts also in large network graphs. An additional system-level simulation proves integration and applicability under realistic conditions.

The question at which layer network functionality is presented or abstracted remains a research challenge. Traditionally, network functionality was either placed into the core network, middleboxes, or into the operating system - but recent developments have expanded the design space to directly introduce functionality into the application (and in particular into the browser) as a way to expose it to the user.

Remote infrastructure is going to assist prospective mobile services, but requires reliable and efficient network resources. Multi-Access Edge Computing (MEC) and cellular networks starting from the fifth generation (5G) are two of the enabling technologies that are necessary to meet strong service level requirements. When service constraints cannot be met all the time, the service may be malfunctioning or even impossible. Impact factors can be user mobility, the network architecture, flexible topology changes, and orchestration decisions. We discuss these challenges while considering the technical integration possibilities. Mainly, we develop a MEC network simulation with a 5G cellular network and user mobility. Our aim is to respect and focus on network conditions for MEC-dependent services and to generate performance indicators for an example task offloading application. We vary the available 5G resources, the number of involved MEC nodes, and application-level behavior like the offloaded task complexity, i.e., computing duration and offloading interval. Additionally, we examine connection-oriented and stateless communication. Our results show that a dense MEC network is able to improve round-trip time (RTT) and jitter in high-load situations and can have a positive impact on service availability and reliability.

Network testbeds have empowered networking re-search and facilitated scientific progress. However, current testbeds focus mainly on experiments involving the current Inter-net. In this paper, we propose SCIONLAB, a novel global network testbed that enables exciting research opportunities and experimentation with the SCION next-generation Internet architecture. New users can join SCIONLAB as a full-fledged autonomous system with minimal effort and administrative overhead, and directly gain unfettered access to its inter-domain routing system. Based on a well-connected network topology consisting of globally distributed nodes, SCIONLAB enables new experiments, such as inter-domain multipath communication, path-aware networking, exploration of novel routing policies, and new approaches for DDoS defense. SCIONLAB has been operational since 2016 and has supported diverse research projects. We describe the design and implementation of SCIONLAB, and present use cases that illustrate exciting research opportunities.