Software-defined networks are key cloud ingredients enabling the Internet of Things, says Joe Byrne of Freescale.
Alphabet soup would be boring if it contained only three letters. Fortunately, two of today's hottest three-letter acronyms belong in the same broth: IoT and SDN.
The Internet of Things (IoT) conjures images of water heaters talking to power meters in the home and robots talking to conveyor belts in the factory, but it transcends simple machine-to-machine communication (add M2M to the soup). Services differentiate the IoT from M2M, with many of them being delivered or orchestrated from the cloud.
With the cloud joining the discussion, software-defined networking (SDN) can't be far behind. It has mainly been viewed as a technology for managing datacenter and enterprise networks, but it has broader relevance. Because SDN provides a framework for centralizing control of network systems, it is ideal for service providers seeking to manage boxes dispersed widely -- exactly the case in the IoT.
Speaking simply, the topology of an IoT deployment reaches from local sensors and actuators to the cloud. The sensors and actuators attach to a local controller. Controllers link to gateways at the customer premises, which connect to the service-provider's access and edge networks. These networks connect to the cloud. The cloud can analyze data and make decisions.
Alternatively, the cloud can delegate the analysis and decision making to gateways, access systems, or edge routers. Performing everything in the cloud makes the most sense when there is a flood of data, linkages among multiple data sources, and relaxed time constraints. Localizing analysis and decision making (known as fog computing, because it disperses cloud technology) is most sensible when there are constraints on time or other resources.
SDN provides a mechanism by which commands can flow from control systems to action systems, such as from the cloud to a gateway. This can reduce the cost of the gateway by transferring some of its intelligence to the cloud. For example, the controller network may use TCP/IP and thus assign IP addresses and route packets for its domain. Instead of putting the DHCP server and router that does this in the gateway, these functions can be moved to the cloud.
Consider also another case in which two things connect to separate gateways but need to talk to each other. If the service provider implements a strict hierarchy, communication must flow high up the chain, perhaps all the way north to the cloud. With SDN, the cloud can instead broker a link between the gateways, giving them a policy to forward east-west traffic between the things without routing the packets to a central node. This example is similar to how SDN is used in the datacenter to set optimal paths between nodes. It's a technique that can be applied throughout the network.
In summary, while the IoT is distinct from SDN, to be most effective, it must use SDN. Conventionally thought of as being for use only in big switches and massive datacenters, SDN is also particularly useful to service providers managing far-flung networks composed of equipment ranging from rack-scale servers to palm-sized gateways.
This change in equipment management must lead to a change in perspective. The industry must look at these systems not as discrete technologies, but as a single system of different components working together. What sits before us is not separate bowls of alphabet soup, but a single large one.