Abstract
Network Function Virtualization (NFV) can lower the CAPEX and/or OPEX for service providers and allow for quick deployment of services. Along with the advantages come some challenges. The main challenge in the use of Virtualized Network Functions (VNF) is the VNFs’ placement in the network. There is a wide range of mathematical models proposed to place the Network Functions (NF) optimally. However, the critical problem of mathematical models is that they are NP-hard, and consequently not applicable to larger networks. In wireless networks, we are considering the scarcity of Bandwidth (BW) as another constraint that is due to the presence of interference. While there exist many efforts in designing a heuristic model that can provide solutions in a timely manner, the primary focus with such heuristics was almost always whether they provide results almost as good as optimal solution. Consequently, the heuristics themselves become quite non-trivial, and solving the placement problem for larger networks still takes a significant amount of time. In this paper, in contrast, we focus on designing a simple and scalable heuristic. We propose four heuristics, which are gradually becoming more complex. We compare their performance with each other, a related heuristic proposed in the literature, and a mathematical optimization model. Our results demonstrate that while more complex placement heuristics do not improve the performance of the algorithm in terms of the number of accepted placement requests, they take longer to solve and therefore are not applicable to larger networks.In contrast, a very simple heuristic can find near-optimal solutions much faster than the other more complicated heuristics while keeping the number of accepted requests close to the results achieved with an NP-hard optimization model.
Highlights
The use of Network Function Virtualization (NFV) can bring advantages and challenges at the same time
Unlike AMPL, which is designed for solving optimization models, MATLAB allows us to develop our heuristic algorithm for Virtualized Network Functions (VNF) placement
There exists a wide range of heuristics proposed for VNF placement, none has focused on design a simple heuristic that is time efficient and scalable
Summary
The use of Network Function Virtualization (NFV) can bring advantages and challenges at the same time. NFEP can be solved by modeling it as an optimization problem that can be solved using different Linear Programming (LP) solvers/tools [1]. Such optimization models are proven to be NP-hard and are not scalable. While there have been numerous proposed heuristics for NFEP in wired and wireless networks, the focus has mostly been on providing results close to an optimal solution. The proposed heuristics became quite complex and highly time-consuming for larger networks, limiting their scalability. We aim to design a heuristic that is as simple as possible, allowing it to be time-efficient/scalable and applicable to larger networks. We are interested in the case of multi-hop wireless networks, with their more severe bandwidth-constraints, but believe that the insights from this work could apply to other networks where NFs have to be placed, such as data centre networks, wired access networks, or the generation of cellular networks
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
