You originally designed your network to connect workstations and for data communications. Next came VoIP, which changed some of the network characteristics by demands to reduce delay and jitter, and then video, with higher bandwidth requirements and streaming. Now you have to embrace the Internet of Things (IoT), in all its myriad forms.
All Sorts of Endpoints
In manufacturing, IoT employs industrial and machine controls. Fleet managers look for a very different type of sensor information, while health care organizations worry about safety, security, speed of delivery, and information accuracy. For digital signage, the endpoint can be pretty “dumb.” Some sensors and endpoints will be highly interactive, while others will be passive, only delivering measurements. Even within one company, many different IoT endpoints will be employed to support various departmental missions.
Most IoT endpoints comprise sensors, processors, network software, and applications from disparate vendors. In some cases, the endpoints are controlled by a gateway also composed of products from multiple vendors. Endpoints are designed to deliver specific functions, and they produce distinct types of traffic.
The variety of endpoints will vary by industry and application. No two will be alike. Therefore, no single network design will support all the various endpoints.
Various Traffic Types & Characteristics
The traffic created by IoT devices varies considerably. For example, traffic varies by whether or not it passes through a processing gateway, how much intelligence it contains, and what applications are running on the gateway. Smart gateway intelligence can respond locally to sensors and process most of the data generated in the gateway. Dumb devices require more interaction over the network with the data center or cloud service.
Traffic characteristics can include:
- Continuous flow
- Polling for data collection
- Periodic data with automatic delivery
- Interactive communications to IoT devices with actuators
- Sensor data from passive devices
Delay tolerance is the amount of time a data transfer takes to reach a destination. When IP networks transfer data, delays of hundreds of milliseconds is acceptable. For VoIP and video over IP, 150ms is the upper delay acceptable. Delay tolerance for IoT depends on the application. If you’re measuring the volume of propane or oil in a fuel tank, seconds of delay is tolerable. But if you’re dealing with safety data or manufacturing systems, then 100ms delay may be too long.
Data intensity, which is the amount of data created by a sensor, can vary considerably. Very-low-bandwidth usage will occur when devices such as locks, doors, and light switches report their status. When gateways are used, then large volumes of data may be transferred. Video HD cameras and some industrial machinery also require a large amount of bandwidth.
With edge processing, a gateway at the network edge can help reduce delay and bandwidth demands on the network. Gateway-resident applications are typically specific to a particular function rather than a generic application.
IoT devices aren’t monolithic. The demands on the network will be far more varied than what IP network designers have encountered in the past, including for voice and video over IP. Bandwidth delivery will be difficult to manage, especially when the network is congested. One of the benefits expected with 5G networks is the ability to deliver very low delay for IoT. IoT devices that can work with longer delays may be better suited for wired connections.
The network designer may want to influence the gateway design by requesting that more work be performed at the edge. This can reduce the demands for short delays and high bandwidth.
Five-nines availability becomes a high priority in the network design. The use of the IoT endpoints further integrates the network into the success or failure of the organization. Deploying gateways can offer sensor support even when the network connection fails. Wired gateway connections can be automatically routed through wireless networks when the wired connections fail.
Prioritizing traffic will be problematic. Voice and video traffic is usually given higher priority than data traffic. Some IoT devices may demand priority transfers that will need higher QoS than voice and video. SD-WAN and MPLS may be required for some of the IoT traffic, while more tolerant IoT traffic may be transferred with lower priorities.
When network designers first encountered VoIP, network performance, especially delay and packet jitter, had to be improved. The improvements also benefited video over IP. IoT devices will challenge network designers and those that operate the networks. The tools to monitor the network will have to be improved as well so they can report the performance by IoT device types and applications.