Virtual network function descriptors mining using word embeddings and deep neural networks

Agile automation of Virtual Network Functions (VNFs) deployment is a must in the future softwarized networks. The automation is generally enabled using descriptor files associated with the VNFs, called Virtual Network Function Descriptors (VNFDs). These descriptors define the resource requirements, operational behavior, and policies that are required for the deployment. We propose in this experience paper a framework for VNFD mining based on Word Embeddings and Deep Neural Networks. The goal of the framework is to automatically recommend VNF descriptors based on a given description or/and to complete the descriptors with appropriate data. The framework is based on two neural network methods: Long short term memory (LSTM) and Convolutional Neural Networks (CNN), which are both complementary in their abilities and could be combined to enhance the performance of the framework. The experiments show good and promising results.