Application of a Long Short Term Memory neural predictor with asymmetric loss function for the resource allocation in NFV network architectures

Abstract

Traffic and cloud resource prediction methodologies have been recently used in Network Function Virtualization environment for cloud and bandwidth resource allocation purposes. Both traditional and innovative prediction methodologies have been proposed for the application of allocation procedures. For instance Long Short Term Memory-based prediction techniques have been shown to be very effectiveness to allocate the resources. All of these techniques are based on the minimization of a symmetric cost function as the Root Mean Square Error that equally weights positive and negative prediction errors. However the error sign can differently impact the cost increase due to prediction errors. For instance when the Quality of Service degradation cost due to traffic loss is prevalent with respect to the cloud resource allocation cost, an algorithm is preferable that overestimates the offered traffic; conversely the traffic underestimation is preferable in the opposite case when the cloud allocation cost is higher than the QoS degradation one. For this reason we propose an Asymmetric LSTM traffic prediction procedure in which the cost function is defined so as to take into account both the QoS degradation and cloud resource allocation costs. In a typical network and traffic scenario, we show how the proposed solution allows for cost decrease by 40% with respect to classical LSTM prediction methodology based on the Root Mean Square Error.