Recently, I gave an interview to a reporter who was asking about the Internet of Things and its impact on networks. The Internet of Things (IoT) has become a huge topic in networking. The IoT is a catchall concept for the connection of things rather than people to the internet. Whereas a browser on a computer or a game on a phone will have a human controlling the action, things are always connected and decide for themselves when and what information to collect and transmit. These things range from very simple things like FitBit bands that upload information to a server, to large things like harvesters on farmland using GPS and downloaded maps to optimize land yields. In general, the IoT is more associated with the former than the latter, as many people have FitBits and few have harvesters. The IoT also covers smart refrigerators, DVD players and other entertainment devices, smart thermostats, and the many other smart devices that are now offered. The number of things in the IoT is exploding.

The reporter’s questions, which are in line with questions and concerns by many in the industry, was whether or not the IoT explosion would have a detrimental impact on networks that are already in place. He (and others) was concerned that the large number of devices being added to the network would swamp the limited bandwidth available on the public internet and cause diminished services for the Internet of People. This is a legitimate concern, but one that is very misunderstood and often overstated.

First, one has to understand what is actually carried on the public internet. A survey of internet traffic done a few years ago (before the IoT was a thing) found that the largest amount of traffic used on the internet was from routers talking to each other. ARP (address resolution protocol) accounted for the largest single percentage of traffic measured, followed by other machine-to-machine protocols. This may have changed in recent years, as some reports state that Netflix is responsible for over 35% of downstream traffic. However, as those reports only compare applications and not protocols, and as those reports only talk of downstream (to customers) and not all traffic, it is hard to know for sure. In any case, it is a safe bet to say that machine-to-machine traffic was and is a huge part of the internet without noticeable impact to the end users. Network operators have had to scale for that traffic and they did.

Now, with the IoT, we are adding thousands or millions of new machines and one might suspect that these new machines would measurably increase the machine-to-machine traffic on the internet. In some cases they might, but the impact is limited because networks today are built around the concept that keeping traffic local keeps the costs down. The movies that you watch from Netflix aren’t streaming from Netflix headquarters in California. They are streaming from a local cache that is probably within a few miles of your home (perhaps a bit further in some rural areas). Netflix has agreements with local network operators to put small data centers on the network operators’ networks so that requests for movies – and the associated traffic – don’t have to travel over the internet very far if at all. The concept of keeping all traffic local means that even Netflix traffic doesn’t impact the public internet to the level that might be supposed from reports.

The reporter asked specifically about traffic from smart Samsung refrigerators being carried over the public internet to Korea, which could be a problem as thousands of refrigerators come on line. The undersea bandwidth between the US and Korea is a limited resource, and one would not like to see it consumed by refrigerators. However, similarly to Netflix, Samsung has very likely set up servers in multiple locations around the globe to collect their statistics. Companies like Amazon offer web and computing hosting around the globe and are able to move the compute and collect power wherever it needs to be for companies that use their services. Even if Samsung has not built their own infrastructure, they could leverage a service like Amazon’s to achieve the same result. Thousands of Samsung refrigerators are most likely not trying to call home to Korea – they are sending information to a local center that is much closer.

The other reason not to panic yet is that machines don’t usually need a lot of bandwidth to talk to one another. The amount of information transmitted by a refrigerator or a FitBit would easily fit into an old dial-up modem connection. There are exceptions like file backup that could take more bandwidth, but most of the things on the IoT are not that chatty. Machines don’t communicate with video and don’t need to be entertained with music.

What could be a legitimate concern with the IoT is the number of connections now required. In 2013, I did a survey of the number of connected devices in my home, and found that I already had 15 devices connected to my network – none of them yet refrigerators. As more of these types of devices start to connect o public end points like cell stations, there is a worry that they will overwhelm the ability of those end points to respond. This is one of the pushes behind microcells and femptocells – those little cell tower extenders that can be put into a home or business and take traffic off of the cell tower. It is also a reason that the move to IP version 6 is important, as the number of IP version 4 addresses for devices has been exhausted.

The bottom line for the reporter was that the IoT is a very good thing, something that should be encouraged (although I’m still skeptical about connected refrigerators) and one that will impact networks in some ways. But not by consuming all of the bandwidth required to keep the Internet of People happy.