A Linear Programming Model for Determining Distribution of User Sessions in a Voice Over Internet Protocol Network

Abstract

Telecommunications service providers build highly available networks using equipment redundancy and overload protection mechanisms. However, because of network equipment misconfiguration or a natural disaster, the network equipment failover could be unsuccessful. This article proposes a linear programming model for determining the optimal distribution of user sessions in a commercial network such that the impact of an unsuccessful network equipment failover is minimal. The model has been applied successfully during the design of a commercial network that supports approximately 5 million customers. The proposed model will be useful for managing the telecommunication networks in a way that is more robust to catastrophic failure, providing greater security and better service to users. The article also presents quantitative results that show a modest increase in end-to-end one-way speech transmission delay ranging from 3.5 to 8.1 ms. This delay is also known as mouth-to-ear delay. The article proposes improvements to the linear programming model by putting an upper limit on the potential increase in the mouth-to-ear delay.