An Openflow-based Scheme for Service Chaining’s High Availability in Cloud Network

High availability (HA) is one of the important requirements of many end-to-end services built on network function virtualization (NFV) and NFV-enabled service function chaining (SFC) due to hard- and soft-ware failures. Once a virtual network function (VNF) fails, the entire SFC operations are broken down. Existing HA solutions for SFC in NFV environments require a significant delay for fault notification to the NFV-MANO (Management and Orchestration) and restoration notification of NFV-MANO to affected controllers. This leads to a significant load at the NFV-MANO, delays the recovery action of the SFC, and causes packet loss during the recovery time. In addition, in Software-Defined Networking (SDN) architecture, controller-based HA solutions are not efficient because studies have shown that SDN centralized architecture may not be able to achieve failure recovery within a 50 ms interval. To address above issue, this paper presents a proposal and implementation of an efficient controller-independent HA scheme based on OpenFlow. In particular, we exploit the OpenFlow group table to design protection plans and a fast failover mechanism, which enable a quick fault recovery for SFCs without requiring fault notification to the NFV-MANO or controllers. Experimental results using OpenFlow with OpenStack show that our proposed scheme achieves a significant improvement in terms of SFC link throughput and packet loss compared to existing schemes.