Proactive network design

This paper describes a process for establishing an optimal network design. High-level system requirements specified by a customer are used to derive performance requirements that bound the network parameters. A deterministic static model is used to derive the bandwidth requirements, e.g., identify performance requirements to address the worst-case conditions. The static model is composed of inter-related mathematical equations derived for network performance from source to destination. A probabilistic dynamic model is used to verify and fine-tune the results of the mathematical model and obtain detailed network parameters at the protocol and application layers, e.g., identify the performance requirements to address the expected-case conditions. The dynamic modeling tool used for this work is MIL3´s OPNET Modeler/Radio, which achieves realistic results through simulating traffic flow to the packet level using random distributions of traffic and file sizes around a predetermined mean. The results of these models are the optimal network parameters used in the proactive network design. A proactive network design is established using static and dynamic models and simulations to predict optimal network parameters such as bandwidth requirements and traffic load. A proactive network provides its users with connectivity and sufficient bandwidth to accomplish predetermined needs; that is, the network maintains stability, a non-congestive state, and meets the end-to-end delay requirements for a particular system. Network stability is the state in which traffic intensity does not exceed network resources. A noncongestive state is a stable state. End-to-end delay is the time associated with the duration of a packet transmission from source to destination. This process was applied to a classified government network to: (1) specify the performance requirements, (2) support development of the network system-level design, and (3) assess the performance impact of changes to the network during detailed design and implementation