An architecture for self-tuning cellular systems

As cellular networks grow in size and complexity, the process of managing them becomes increasingly costly for mobile operators. It is therefore important to focus on simplifying or automating these tasks. This paper describes an architecture that supports automated management of a cellular network. Automation can be done by gathering and processing measurements from the network resources and, based upon the results, feeding back configuration commands. Due to potentially large networks, the large number of measurements needed, and the distributed nature of a cellular system, there is a need for a distributed architecture. This paper proposes such an architecture and describes evaluation results using this architecture in a field trial at the operator SmarTone in Hong Kong. Results from the field trial show that the suggested agent/task pattern is well suited to implementing self-tuning algorithms