Demand Engineering: IP Network Optimisation Through Intelligent Demand Placement

Abstract

Traffic engineering has been used in IP and MPLS networks for a number of years as a tool for making more efficient use of capacity by explicitly routing traffic demands where there is available network capacity that would otherwise be unused. Deployment of traffic engineering imposes an additional layer of complexity to network design and operations, however, which has constrained its adoption for capacity optimisation. The rise of Software Defined Networks has renewed interest in the use of traffic engineering approaches leveraging centralised network controllers for capacity optimisation. We argue that future networks can realise the network optimisation benefits of traffic engineering without incurring additional network complexity through closer coupling between the network and the applications and services using the network. This can be achieved through leveraging a network- and traffic-aware controller to directly influence where applications and services site or locate service instances, i.e. which implicitly impacts the paths that the applications or services traffic demands take through the network. We call this technique Demand Engineering. Demand Engineering has the additional benefit of providing an admission control capability, i.e. which can provide an assurance that network SLAs can be met. In this paper we describe the concept of Demand Engineering, give examples of its use and present simulation results indicating its potential benefits. We also compare demand engineering to traffic engineering.