Cisco predicts 50 billion interconnected devices by 2020. Intel, more optimistically, predicts 200 billion by that same year. Should we be concerned about the address needs of this large Internet of Things? Briefly, no. Moreover, we shouldn't assume that every “thing” has to have its own IP address.
Let’s start at the beginning. We ran out of IPv4 addresses managed by the Internet Assigned Numbers Authority (IANA) on Feb. 1, 2011. IPv4’s meager 32-bit address space of four billion is not even enough to give each person on earth a unique identifier. IPv6 has come just in time, providing 340,282,366,920,938,463,374,607,432,768,211,456 possible 128-bit addresses. This makes the address needs of the Internet of Things ho-hum. We can all relax -- IoT is a mere blip, with plenty left over.
Now let’s examine the question: Are the “things” in IoT really each going to need a traditional IP address? In today’s Internet, the answer would be yes, because the things are mostly servers and switches, firewalls and routers, laptops, phones and tablets with IP to IP connectivity. Because of network address translation and private network addressing, a unique-to-the-world IP address is not needed; many machines can have the same address.
But when we start talking about refrigerators, clothing, thermostats, light bulbs, and the temperature probe in your steak, we should ask: Do they all need to be directly on the Internet with an IP address, unique or otherwise? I don’t think so.
Cisco’s “fog computing” concept seems to support this. It's a distributed edge computing paradigm. Some machine-to-machine (M2M) computation and storage takes place locally in a “fog” -- say a home or a car -- and then there is periodic communication with the cloud. Local clusters of devices that interact may do so without reference to the Internet. Then, the locally collected data could bridge to a “smarter” device with intelligence, acting as an integrator or gateway to the cloud. In this case, the smarter device would have a traditional IP addresses, while the feeder devices may not.
Two examples of this today are Bluetooth and RFID. Your iPhone has an IP address; the Bluetooth speaker it connects to seldom does, since it is a Bluetooth link rather than an IP-to-IP connection that is needed for you to hear music.
Similarly, your car’s RFID device on the windshield (E-ZPass in 14 northeastern states) interacts with the tool booth to charge you for travel down a toll road. IP addresses are not needed for the toll interaction, yet you will receive a billing statement via email. This is the secondary part, where the device contacting you needs an IP address.
However, this doesn’t mean that RFID devices don't need to have addresses; it's just that the RFID 96-bit domain is different from the Internet address space. Passive RFID tags have limited resources and cannot talk IPv6; a bridge or translator is required. It's only at this interface that a traditional IP address is needed, though hybrid RFID/IPv6 devices are often talked about for the versatility they would offer.
IP is certainly versatile. It works over wireless (WiFi, WiMax, 3G, 4G, LTE, IEEE 802.15.4/LoWPAN, etc.) as well as wired (Ethernet, SONET, etc.) communications. Where power or distance is not an issue, and larger devices are involved, primary Internet access for IoT may well be desirable, but not essential. Heterogeneous islands can have useful local computation and local networking with IP connectivity to the mainland.
Of course, this is a more federated view of IoT. Either way, the thing to not worry about is whether there will be enough addresses.
Discover the benefits of scalable backbone and network services for application and cloud providers.
Network automation is gaining momentum. Here's how you can drive this powerful technology to its full potential.
Subsea cable alternatives can provide a level of comfort for those concerned about disruptions. Realistically, they will continue to play a small, but critical role in the short term.
