Optimized Transition Planning ensuring the Reliability of INP with Stochastic Contexts
A key task of in network processing (INP) / network function virtualization (NFV) is the virtual network embedding problem, i.e., to place virtual networks of logical, interconnected function blocks into a substrate network of physical compute nodes and communication links. The specific target of this subproject is to ensure high run-time reliability of the embedded functionality given dynamic changes of the substrate network, its loading and environment, the so-called context.
For each logical function block and communication link we assume an annotation with (potentially context dependent) reliability requirements. Next to the embedding, the developed embedding algorithms will then also schedule one or more coexistent reliability mechanisms, such as e.g. redundancy of nodes or links or improved restart capability by (partial) storage of the function state in the network.
Iteratively running the algorithm with current and predicted contexts will contribute to MAKI’s focus on predictive, model-based transition planning for communication mechanisms. Formulated as an optimization problem, previous research on distributed and robust optimization in the energy domain can be transferred to communication networks. All results will be motivated and demonstrated with existing applications in MAKI and a new application, namely communication systems in energy automation.
The project was funded from 01.10.2018 – 31.12.2020.
Subproject leader X2
- Prof. Dr.-Ing. Florian Steinke