Network Representation Learning Aided Resource Allocation in Software Defined Networks

Resource allocation is a difficult online decision-making problem in Software Defined Networks (SDN). Traditional algorithms usually formulated the network resource topologies as simple matrices or edge lists (e.g., adjacency matrices). However, the rapid growth of the network scale and data volume brings new challenges to the scalability and efficiency of these models. Recently, network representation learning (NRL) has been widely adopted in network modeling, which can embed network nodes and links into low-dimensional vectors. Specifically, as one of NRL techniques, knowledge graph (KG) integrates both pieces of knowledge and their relations, and therefore capturing relational information in the network. Therefore, in this paper, we adopt KG for SDN representation and propose a novel resource allocation scheme based on the relational information learned by KG embedding. We first extract relational information between network nodes, links and requests, and then embed them into low-dimensional vectors. Based on these vectors, we calculate resource allocation schemes by consecutively selecting relay nodes and links in consideration of both available resources and their relational vectors. The extensive simulations are conducted to evaluate our proposed algorithm in comparison to state-of-the-art schemes.