Network primitives based on latency and recovery time in orchestrated multi-layer networks
Ćiril Rožić, Chris Matrakidis, Dimitrios Klonidis, Ioannis Tomkos · 2017
Other than merely a bandwidth pipe, current and future network applications will have other requirements such as maximum end-to-end latency and the type of recovery from failure. The added complexity will require a central intelligence such as a software-defined networking (SDN) orchestrator to compute and keep track of the resources allocated to such applications. In such an SDN-orchestrated network, we examine a planning scenario where not all the requirements are known in advance. Instead, they are computed by the network orchestrator and referred to as network primitives. The primitives are then presented to the applications so that the applications know what type of service the network can offer. With our planning algorithm, we show an example of computing the primitives by analysing the guaranteed maximum end-to-end latency and post-failure recovery time.