Industrial IoT systems depend on networks that can carry data between distributed sensors, actuators, gateways and applications. As those systems grow, the mix of devices and communication technologies makes resource management more complex. An SDN-based IoT network offers one way to address that challenge: it separates network control decisions from packet forwarding, making traffic management more programmable. In an emulated comparison, this approach performed better than a traditional IoT network across latency, jitter and throughput measures.
Why IoT network management is challenging
IoT networks are heterogeneous by design. Devices may use different protocols and access technologies, including Wi-Fi, Bluetooth and ZigBee, while performing distinct roles in sensing, processing and communication. In industrial and other distributed applications, sensors and actuators can be geographically dispersed and may need to exchange data with services at the network edge or elsewhere.
Traditional network devices commonly combine the control plane (CP), which determines how traffic should be handled, with the data plane (DP), which forwards packets. Devices may also rely on predefined rules and vendor-specific implementations. This can make network-wide changes and the introduction of new services more difficult, particularly when application requirements vary.
Quality of service (QoS) is important because different traffic types can tolerate different levels of delay and delay variation. Latency describes the time taken for data to travel between endpoints. Jitter describes variation in packet delay, which can disrupt a consistent stream of data. Throughput measures how much data is successfully transferred over a period. Congestion, packet loss and limited device resources can all affect network performance.
How an SDN-based IoT network works
Software-defined networking (SDN) changes how network decisions are organized. It decouples the control plane from the data plane: a controller makes routing or forwarding decisions, while switches forward packets according to the rules they receive. A programming interface such as OpenFlow can connect the controller and switches.
In the evaluated design, an OpenFlow-enabled switch checks whether an arriving packet matches an existing flow-table entry. When the first packet of a flow has no match, the switch sends it to the controller. The controller determines a forwarding rule and returns it to the switch, which can then handle later packets in that flow using the stored entry rather than requesting a new decision for every packet.
This distinction matters for network overhead. The initial packet may require controller involvement, but subsequent packets can be forwarded using the installed rule. By reducing repeated communication between forwarding devices and the control logic, the architecture aims to make traffic handling more efficient.
What the performance evaluation found
The comparison used emulated network environments: GNS3 for the traditional network and Mininet for the SDN topology. The test setup included 70 host nodes and 450 ICMP packets, with a packet size of 64 bytes. The reported performance measures were latency, jitter and throughput.
Latency and jitter
The reported average latency was lower for the SDN-based IoT network across the tested node counts. In the traditional setup, packets may require routing decisions as they pass through network devices. In the SDN setup, the first packet triggers a controller decision, while later packets in the same flow can use the switch’s flow-table entry. The study reported a 574% average latency percentage improvement from the traditional network to SDIoT across the nodes considered.
Average jitter was also lower in the SDN results. Because jitter reflects differences in packet arrival timing, reducing variation can help traffic that depends on a more consistent stream. The study reported an average jitter percentage improvement of more than 600% across the nodes considered. These findings are specific to the emulated setup; they should not be treated as a guarantee of the same outcome in a production network.
Throughput
The evaluation also recorded higher overall throughput for SDIoT than for the traditional network across the tested cases. Throughput reflects the amount of data successfully delivered over time, so it provides a complementary view to latency and jitter. A network may have acceptable delay for some traffic while still delivering data inefficiently if congestion or packet loss reduces successful transfer.
Practical considerations for network planners
The results indicate potential QoS benefits, but deploying SDN also introduces design decisions. Controller placement and control architecture matter: a logically centralized control plane can simplify network-wide management, while relying on a single control point can create scalability and reliability concerns as the network grows. The paper notes that controller load may increase with the number of nodes, switches, services and traffic flows.
For industrial automation teams and system integrators, the findings support evaluating SDN where network flexibility and traffic differentiation are priorities. A practical assessment should consider:
- Traffic requirements: Identify which applications are sensitive to latency, jitter or packet loss.
- Flow behavior: Account for the controller interaction associated with a flow’s initial packet and the role of installed flow rules.
- Controller architecture: Assess controller capacity, placement, scalability and the consequences of a control-plane failure.
- Validation conditions: Test with network layouts, traffic patterns and hardware representative of the intended deployment.
Conclusion
Separating control from forwarding gives an SDN-based IoT network a programmable way to manage flows and reduce repeated control-plane interactions. In the reported emulation, SDIoT achieved better average latency, jitter and overall throughput than the traditional network. For plant and operations decision-makers, the practical takeaway is to treat SDN as a promising network architecture to evaluate against specific QoS needs—not as a substitute for deployment-specific testing, capacity planning and controller resilience design.
Access Official Technical PDF