Embedding Multicast Service Function Chains in NFV-Enabled Networks

Virtual Network Function Embedding with Quantum Annealing

“…The experiments have all been conducted with an experimental pipeline that is depicted in Figure 2 and is publicly available on GitHub together with additional online material. 4 In detail, the QUBO formulation module is the main one and is in charge of producing a suitable formulation given as input a specific VNFEP (i.e., topology and SFCs to be embedded). After this phase, a set of parameters (e.g., number of reads, chain strength) is provided to the three solvers so that they can be consistently configured.…”

“…Due to the complex nature of the problem, finding an effective solution is challenging as the solution space grows rapidly and exploring it exhaustively is impractical. In previous works, some heuristic strategies have been proposed to obtain good quality solutions in a reasonable time [1]- [4], as a compromise to reduce the computational effort. However, stochastic methods do not guarantee convergence on the optimal solution.…”

A comparative analysis of two‐stage approaches for embedding network function virtualization enabled multicast services

“…After solving (), since the decision variables $$x_{idf_n}$$ and $$x_{if_n}$$ are relaxed in LRP, the rounding process must be run. In the rounding technique (adopted from [18]), we concentrate to round the values of $$x_{idf_n}$$ and at the end, the integer values of $$x_{if_n}$$ are obtained by (). For rounding, first, the threshold $$\gamma _{df_n}$$ is applied roughly on each $$x_{idf_n}$$, $$$$\begin{equation} x_{idf_n} = {\begin{cases} 1 & \text{if } x_{idf_n} \ge \gamma _{df_n}\\[6pt] 0 & \text{if } x_{idf_n} < \gamma _{df_n} \end{cases}}, \end{equation}$$$$where $$\gamma _{df_n}=1/n_{df_n}$$, and $$n_{df_n}$$ is defined as the number of nonzero values on the vector $$A_{df_n}$$.…”

“…After solving (10), since the decision variables x idf n and x i f n are relaxed in LRP, the rounding process must be run. In the rounding technique (adopted from [18]), we concentrate to round the values of x idf n and at the end, the integer values of x i f n are obtained by (4). For rounding, first, the threshold 𝛾 df n is applied roughly on each x idf n ,…”

“…Then, these sub‐problems are solved using a simpler exact, approximate, or heuristic method in two consecutive stages. A two‐stage algorithm for solving the problem of NFV‐enabled multicast embedding is introduced by Asgarian et al., where, first a simple Linear Program (LP) for VNF placement is solved, and then, the multicast tree is found by solving another simple linear program [18]. Also, they proposed a two‐stage algorithm named TSDV (Two‐stage Source‐Destination‐VNF) based on the newly defined SDV (Source‐Destination‐VNF) graph to offer the trade off between complexity and efficiency [19].…”