SDN based Network Function Parallelism in Cloud
This addresses performance bottlenecks in cloud-based network function virtualization for network operators, though it is incremental as it builds on existing SDN and NFV frameworks.
The paper tackles the problem of inefficient network function deployment in cloud networks, which can cause security violations and performance overhead, by introducing Network Function Parallelism (NFP) for OpenFlow networks, achieving a 1.40-1.90x reduction in latency for service function chaining in an SDN-managed OpenStack cloud.
Network function virtualization (NFV) based service function chaining (SFC) allows the provisioning of various security and traffic engineering applications in a cloud network. Inefficient deployment of network functions can lead to security violations and performance overhead. In an OpenFlow enabled cloud, the key problem with current mechanisms is that several packet field match and flow rule action sets associated with the network functions are non-overlapping and can be parallelized for performance enhancement. We introduce Network Function Parallelism (NFP) SFC-NFP for OpenFlow network. Our solution utilizes network function parallelism over the OpenFlow rules to improve SFC performance in the cloud network. We have utilized the DPDK platform with an OpenFlow switch (OVS) for experimental analysis. Our solution achieves a 1.40-1.90x reduction in latency for SFC in an OpenStack cloud network managed by the SDN framework.