Artificial Intelligence-Based Algorithm for Resources Allocation
Constantin Viorel Marian · 2022 14th International Conference on Electronics, Computers and Artificial Intelligence (ECAI) · 2022
In today's world the information should be available at any time and any place. From the user's point of view, these requirements are mandatory even if the traffic is growing and the user's experience is hyper-contextualized., personalized and mostly dynamic. From an application services provider point of view, in software defined networks and cloud intelligent managed data rooms, the networking architecture is required to combine central management and network residency. To provide the above mentioned services, the telecommunications networks have to provide maximum flexibility. The same mandatory functionality is required for clustered architectures used in cloud service providers' data rooms. This maximum resilience must involve two aspects: flexibility seen as accepting many different possible configurations and flexibility as the required time for these configurations to be applied to equipment in order to take effect. A possible solution to these requirements, is to separates the data plane forwarding the traffic from the control plane who's taking the decisions based on different criteria. In the meantime the control plane evolves and becomes a complete centralized management solution. This paper presents two new ideas and their implementation in an operational test data room. The first one consist of a data processing technique and algorithm that leads to dynamic resource allocation. The second implementation has as a main focus on the artificial intelligence AI-based trend prediction module that takes the decisions for the resources allocation algorithm. This research is part of a project developed by a joint team composed by cloud services provider in France and computer science specialists in Romania.