Optimization Models for Flexible and Adaptive SDN Network Virtualization Layers

Johannes Zerwas,Andreas Blenk,Wolfgang Kellerer

DOI: https://doi.org/10.48550/arXiv.1611.03307

2017-10-11

Abstract:Network hypervisors provide the network virtualization layer for Software Defined Networking (SDN). They enable virtual network (VN) tenants to bring their SDN controllers to program their logical networks individually according to their demands. In order to make use of the high flexibility of virtual SDN networks and to provide high performance, the deployment of the virtualization layer needs to adapt to changing VN demands. This paper initializes the study of the optimization of dynamic SDN network virtualization layers. Based on the definition of reconfiguration events, we formalized mixed integer programs to analyze the multi-objective problem of adapting virtualization layers. Our initial simulation results demonstrate Pareto frontiers of conflicting objectives, namely control plane latency and hypervisor and control path reconfigurations.

Networking and Internet Architecture

What problem does this paper attempt to address?

This paper attempts to address the problem of how to optimize the dynamic network virtualization layer (NV) in Software Defined Networks (SDN) to accommodate the ever-changing demands of virtual network (VN) tenants. Specifically, the paper focuses on how to dynamically adjust the position of the network virtualization layer in a virtualized environment when user demands change, in order to find the optimal balance between control plane latency and the number of network reconfigurations.

### Background of the Paper
- **SDN Network Virtualization**: Achieving flexible programming and management of virtual networks through SDN controllers.
- **Network Virtualization Layer**: Provided by network virtualizers (such as SDN controllers), enabling virtual network tenants to independently configure their logical networks according to their needs.
- **Dynamic Adjustment Requirements**: As user demands change, it is necessary to dynamically adjust the position of the network virtualization layer to maintain high performance and flexibility.

### Main Issues
- **Control Plane Latency**: Control plane latency in virtual networks affects user experience, especially in scenarios like web requests where every millisecond counts.
- **Network Reconfiguration**: Dynamically adjusting the position of the network virtualization layer can lead to network reconfiguration, which may cause high latency or even network interruptions.

### Research Objectives
- **Multi-Objective Optimization**: Using a mixed integer programming model to study how to balance control plane latency and the number of network reconfigurations.
- **Dynamic Virtualization Layer Deployment**: Proposing a method that allows the network virtualization layer to dynamically adapt to the changing demands of virtual network tenants.

### Methodology
- **Mixed Integer Programming Model**: Defining different reconfiguration events and using a mixed integer programming model to analyze the multi-objective optimization problem.
- **Simulation Study**: Conducting simulation studies to demonstrate the Pareto frontier between control plane latency and the number of network reconfigurations.

### Preliminary Results
- **Trade-Off Relationship**: Preliminary results show that even a small relaxation in latency (e.g., 1%) can significantly reduce the number of network reconfigurations.
- **Impact on Networks of Different Scales**: Studying the impact of different numbers of virtual networks and network virtualizers on control plane latency and the number of network reconfigurations.

### Conclusion
- **Effectiveness of Latency Relaxation**: A 7% latency relaxation can completely avoid network reconfiguration.
- **Model as a Baseline**: These models can serve as a baseline for designing heuristic solutions.
- **Future Work**: Planning to extend the research to consider more objectives and analyze larger physical networks.

Through these studies, the paper provides a theoretical foundation and practical guidance for dynamically adjusting the SDN network virtualization layer, helping to improve network performance and user experience.