Çoklu-erişim kenar bilişim sistemlerinde işlem yükü devri karar yöntemlerinin karşılaştırılması
Özet
In recent years the increase in the performance expectations of the users from smart devices together with the widespread use of the Internet of things devices has led scientists to seek new methods in order to quickly and effectively conclude the processing loads on the devices. Cloud Computing has been used for many years to solve this problem and it still continues to be used. However, Cloud Computing cannot meet the expectations due to its remote centralized structure from the user. On the other hand, Multiple-Access Edge Computing systems are capable of solving the problem due to the fact that they are installed on base stations at the edge of the network close to the end user. However, the capacity of Edge Computing servers is not unlimited. In this context, a new problem emerges: how to distribute the limited processor capacity among users. Within the scope of this thesis, some methods proposed to solve the problem in question were compared according to various criteria. These methods are; Minority Game Method, Q-learning, Win Stay Lose Shift Strategy, Roth-Erev Learning and Learning Automata. The problem was modeled as a resource allocation problem and a simulation study was carried out in this framework. In the studies, first of all, the most appropriate values of the parameters in the methods used were tried to be found, then these parameters were used and the methods were compared according to various criteria. In the studies carried out, the methods were examined in terms of volatility, average welfare, fairness, average completion time of a round and average energy spent in a round.