Internet of Things becomes one of the main trends in development of future networks. It introduces new challenges to network operator, like exponential growth of the traffic, large amount of network nodes to manage or variety of services. Some IoT services include the control of critical processes such as remote manipulation of a robotic arm in tele-surgery. In such processes, keeping very low end-to-end latency in the communication is crucial. The critical traffic generated becomes important and thus it needs special treatment, especially in situations when latency may be affected (e.g. network congestion or network node malfunctioning). In this work we analyzed what are the technical possibilities to manage traffic delay for important data. We selected Software Defined Networking technology to manage the end-to-end IoT traffic. An implementation based on OpenDayLight SDN controller allowed to build a solution that is able to identify in real-time the routing path with minimum delay and to route the important data traffic to this path. In order to monitor the latency, a probe packet is sent over each path and its trip time is measured. Simulations based on changes of the physical delay on links forming paths confirmed the effectiveness of our approach. They also showed the conditions, where the solution can be applied.