Dynamic QoS/QoE assurance in realistic NFV-enabled 5G Access Networks

Genetic Algorithm for Effective Service Mapping in the Optical Backhaul of 5G Networks

“…Figure 2 shows the corresponding values of percentage of active CPU cores. These figures show that GASM achieves the same service blocking ratio as the "MEC First" policy [7], [8] (Figure 1), but reducing the number of active CPU cores (and thus energy consumption) when compared to this policy (a reduction ranging from 45 to almost 10 percentage points, depending on the number of users, as shown in Figure 2) for up to 4,500 average users. When compared to the "CO First" policy, the latter requires the same or less number of active CPU cores than GASM, but GASM significantly improves the results in terms of blocking ratio.…”

“…Both types of cells are connected to a CO via 10 Gbps optical links. The nodes are equipped with the computational resources shown in Table 1 like in [7,8]. We assume that all the nodes are NFV-enabled.…”

“…A VNF is available if it processes less operations than its maximum number of concurrent operations per instance. The algorithm looks for VNFs at the local cell first, and only searches for VNFs in the CO when no available VNF are present at the local cell, like in [7], [8]. Once the algorithm finds an available VNF of the SC in the CO, it will not look for VNFs at MEC nodes again.…”

“…In order to improve the performance, two ad-hoc individuals are created in the first generation of the genetic algorithm together with other randomly generated ones. The first ad-hoc individual is created so that its solution is equivalent to following the policy used in [7], [8]. We call this policy "MEC First" since it first tries to instantiate or use VNFs at the radio node.…”

“…Last but not least, in [6] authors studied the impact of latency requirements on the placement of VNFs constituting an added value service in a static scenario. ONOFRE project (Optical Network Convergence of the Future Internet) proposed in [7], [8] an algorithm for service chaining allocation in a real-life scenario, with a very high density of population and offered traffic expectations changing rapidly; with the aim of reducing service blocking ratio due to either unavailability of optical network capacity or computing resources.…”