Wide area network architectures are rapidly evolving to accommodate the internet of things (IoT), with everything from advanced fiber-based data center interconnects (DCIs) to software-defined WANs (SD-WANs) on the table.
But in addition to the macro-architectural issues, the IoT will require new forms of packet and data traffic management. Most IoT-connected devices, after all, will continuously stream very small data packets into the web, not the large-volume transfers that the WAN is used to carrying. And these packets will require not just the point-to-point carriage of traditional workloads, but fabric-style multipoint-to-multipoint networking across a global footprint.
Only recently, however, has this aspect of IoT networking gained significant attention, which means that the enterprise industry is under the gun to establish new, preferably universal, protocols and management solutions before IoT traffic starts to achieve significant scale – which will most likely happen within the next two years as 5G networking becomes available.
One of the key problems is devising a sequencing pattern so that millions – nay billions – of packets do not get jammed up in buffering queues waiting their turn for a particular network resource. Latency is a killer on the IoT, especially when it involves life-critical applications like autonomous cars and medical devices.
Researchers at Osaka University and the University of Tsukuba, however, think they have come up with an elegant solution by listening to tree frogs. After analyzing the interplay between individual croaks, they say they have identified a novel pattern of vocalization and silence that could provide the basis for high-speed sensor networking. The frogs, it turns out, are able to manage their temporal overlap in ways that are both consistent and deterministic so that all individuals can be heard. The team is now working on turning this pattern into a phased-based mathematical model that can be applied to multi-sensor, multi-hop environments.
Precise packet coordination is also a key consideration for the industrial IoT, which is aimed at manufacturing and materials processing. As IEEE Spectrum’s Stacy Higginbotham noted recently, syncing data on the factory floor is usually accomplished with quartz timers on each device and a wealth of proprietary protocols to keep it all connected. This gets rather expensive as the number of data points scales up, so the industry is instead looking to devise a universal time-sensitive networking protocol to the Ethernet standard. Not only will this boost data interconnectivity to gigabit speeds, it also has the ability to differentiate between machine traffic and secondary data, such as software updates or file exchanges.
Meanwhile, the Linux Foundation is working on a number of fronts to boost connectivity and interoperability across both the industrial and consumer-facing IoT. One such effort centers on Dell EMC’s EdgeX Foundry solution, which is designed to foster a platform-agnostic software framework that acts as a bridge between sensors and the cloud. Meanwhile, Samsung’s Home Edge Project is looking to do the same for devices in the home. The idea is to streamline the deployment and provisioning process of both IoT hardware and software, and make sure that data traffic can easily traverse the multi-vendor networking paths it will likely encounter from the device, to the edge, to the cloud, to the data center.
But simply moving packets from point to point is not the only challenge in the IoT. Semiconductor Engineering’s Susan Rambo pointed out recently that packet inspection and conditioning is equally important. While most sensors are highly reliable and accurate, they are not infallible. When signals go awry, the enterprise will need ways to determine if the fault lies with the sensor, the network, the algorithm or somewhere else. Likewise, even accurate data is not necessarily useful for each and every application, so in any given use case data must be examined and cleaned either in the data center, at the application or in transit.
IoT networking is a good example of the old adage “It’s not what you do, it’s how you do it.” Faster throughput, wider bandwidth and automated management are all valuable tools for IoT connectivity, but eventually the work has to focus on the nitty gritty of how data gets from here to there.
Packet delivery and traffic management are just as important for the IoT’s success – or failure – as anything that can be done with architecture or infrastructure.