Network function Virtualization (NFV) is expected to accelerate service deployment and enable service agility for Telco operators. NFV builds on virtualization technology and off-the-shelf general purpose hardware, shifting previously dedicated Telco appliances towards the cloud ecosystem. This paradigm shift brings new concepts such as dynamic service chaining that necessitates to rethinking the Network Management approach. In this paper we present CogSLA, a data-driven solution for SLA enforcement in an NFV deployment. CogSLA uses Deep Feedforward Neural Network or Multi-Layer perceptron (MLP) for achieving proactive identification of SLA violations. The key contribution of this work is to proactively change the network state, anticipating and avoiding foreseeable SLA violation