An Extended OVN Architecture for End-to-End QoS Guarantee in Network Slicing

In recent years, Network Slicing has emerged as a solution to help deploy services with different Quality of Service (QoS) requirements on the same shared physical infrastructure. To facilitate resource allocation in network slices, a Virtual Infrastructure Manager (VIM), e.g. OpenStack or Kubernetes can be used to virtualize and manage the physical infrastructure. Currently, Open Virtual Network (OVN) is one of the main open-source technologies used in the above VIMs to create virtual networks and deploy network slices. However, these VIMs are only capable of managing overlay network infrastructure built on compute nodes that are directly connected through tunneling protocols and Virtual Tunnel Endpoint (VTEP). OVN can therefore only control traffic at the endpoints of the virtual link (VL). As a result, end-to-end QoS for services deployed on virtual networks created by OVN cannot be guaranteed on shared network node infrastructure. This paper proposes a solution based on a traffic tagging mechanism to bridge this gap of OVN and develops an architecture based on SDN/OVN integration to manage QoS for network slices dynamically. Then, the performance of the proposed architecture is evaluated through specific use cases.