Nowadays, networks are present everywhere. They had become important part of our lives. They have to serve continuously increasing demands, so they themselves are getting bigger and more complex. With the emerging of IoT (Internet of Things), the demands are likely to follow this increasing rate.
The management of the networks often still happens in the old-fashioned, centralized way. In this approach, the nodes communicate with a centralized management station. However, this method is far from optimal: a lot of control data has to travel through the network, wasting bandwidth. If the management station is overwhelmed, the response times will be long. And with the increasing heterogeneity of the networks, the centralized management also tends to become way too complicated.
The mobile agent based network management offers a distributed alternative: intelligent programs capable of making decisions on their own travel the network and solve problems locally. Because of the distributed working, fast reactions to dynamic changes are possible. We are also able to use the bandwidth sparingly, because we only have to transfer the code of the agent from one node to another, what can mean much less data than whole SNMP tables used in the centralized approach.
The number of the agents present on the network can be adjusted with different population control algorithms. The goal of these algorithms is to reduce or increase the number of agents, converging to the optimal number. Less agents than the optimal may result in poor service of the nodes, while overpopulation can lead to wasting of resources.
To examine these algorithms, I made a simulation environment. Using it, I analyzed different algorithms and investigated the results.