Year of Graduation
Building a Neural Network for Load Balancing on the CDN
The goal of this work is the development and implementation of a new balancing algorithm for the existing CDN-network and comparison of the results of its work with already implemented methods. In the course of this work, the principles of the NS-based and End-user balancing were thoroughly studied. The problem of constructing a neural network load balancing algorithm for CDN is considered and its practical feasibility is not realized because it is impossible to create a training sample that does not reflect the CDN balancing method used. The method of CANS-balancing is developed. Programs that collect and analyze data is developed. The control measurements of the results of the three methods of balancing were carried out: NS-based, End-user and CANS. It was found that CANS-balancing manifests itself best in certain situations, when End-User balancing is ineffective. The results of the work can be useful when you cannot use the EDNS0 extension.