RAFALE: Rethinking the provisioning of virtuAl network services using a Fast and scAlable machine LEarning approach

Network Function Virtualization (NFV) has been identified to revamp the provisioning of next-generation network services. This new paradigm allows cloud and network/service providers to compose their network services, also known as service function chains (SFCs), in an agile way since the software of the network function is decoupled from the legacy hardware. To reap the benefits of this new technology, there is a need for novel mechanisms that help cloud and network/service providers deploy the increasingly complex virtual network services seamlessly, efficiently, and in a time-efficient way. Existing state-of-the-art techniques often rely on the Integer Linear Programming framework, heuristics/metaheuristics, and greedy methods to deploy the services function chains. However, these techniques although reasonable and acceptable, still suffer from several key limitations: convergence time and scalability. To this end, we propose RAFALE, a suite of solution techniques, to tame this complexity by leveraging the concept of similarity from machine learning and skip-gram modeling framework. To the best of our knowledge, we are the first to tackle these key limitations and propose a suite of solutions to them. RAFALE, a novel approach proposed to find the similarity between the new incoming virtual network service request and all the already-deployed services to learn from the previous experience of deploying techniques and use the same or close similar provisioning techniques. RAFALE is the first and the only method that develops the idea of detecting the similarity between virtual network services. Experimental results show that RAFALE reduces greatly the convergence time needed for provisioning virtual network services and can scale to 100 virtual network functions per virtual network service compared to the state-of-the-art. The Experimental results prove that RAFALE accomplished the NFV promises; decreasing the time and complexity of managing and deploying the virtual services, and providing a solution that is agile, faster, and scalable to deploy the new service requests by skipping one or more service provisioning steps (i.e., detecting and resolving the conflicts among policies, placement, and chaining) while satisfying the validated NFV policies.

     

Economical evaluation of the fluoride removal by nanofiltration

Economic evaluation was carried out for a plant of fluoride removal by nanofiltration having a capacity of 2400 m³/d (100 m³/h) corresponding to a water consumption for 50.000 capita following the Moroccan considerations in rural medium. The design of this plant was carried out for the predetermined optimized conditions corresponding to a recovery rate of 84%, a fluoride rejection of 97.8% and a pressure pump of 10 bar. The capital cost was estimated to 748,003 € and the calculated operating cost to 0.212 €/m³. These costs were briefly compared to other ones.     

Economic evaluation of fluoride removal by electrodialysis

An economic study was conducted to evaluate the costs of fluoride removal by electrodialysis on the basis of industrial and economic data. The investment and operating costs were estimated for an industrial plant with a capacity of 2200 m³/d water consumption for 50,000 per capita according to Moroccan standards for rural areas. The capital cost was estimated to be € 833,207 and the calculated operating cost to be € 0.154/m³.     

The effect of annealing on the optical and electrical properties of a-Si:H sputtered films

In this work we study the influence of an annealing treatment at high temperatures on the optical and electrical properties of a-Si:H sputtered films. The results show important changes of the different parameters determined after annealing. A possible model of the silicon matrix evolution during annealing is proposed, and the dependence of the energy gap on disorder and hydrogen concentration is discussed.