Overloaded Network

SDN and the Internet of Everything

Overloaded NetworkSoftware Defined Networking (SDN) is having an ongoing impact on the design of network equipment and in how users architect their environment.  The abstraction of the control and management planes from the data plane offers obvious benefits, although in practice it has not been quite as clear cut as many had hoped.

Now along comes the Internet of Everything (IoE).  New devices, new “things” are being connected to the internet on an ever-increasing basis: measurement and monitoring devices (e.g. temperature controls, surveillance equipment); production line equipment (e.g. controls and actuators); along with all the consumer stuff, such as home equipment (fridges, heating and lighting controls, entertainment systems, etc.). and now wearables — watches, glasses, health monitoring devices, you name it.

Each of these devices creates its own data, and networks need to be able to deal with this.  The problem is that each device is often not intelligent enough to be a good neighbor on the network. Device size, battery life, and cost of production all have implications on how much technical design can be put in place, and the downstream impact of a poorly-engineered data approach can have far-reaching impacts.  In essence, what we are seeing is a raft of devices that collect data and then spew it out to the network in the hope that something else can deal with it. However, what that “something else” could be is not defined.

Sure, SDN can help here: it can capture data streams and use the management and control planes to decide what actions should be taken on the packets of data before it pushes them on back down to the data plane to be passed on to the next stage of the network.

However, this does involve putting bottlenecks in the network.  These higher-level management and control systems will have to look at far more data than they were possibly designed for: trying to carry out intelligent decisions on what to do with millions (or even billions) of small, noise-level data packets from the IoE could stress SDN to beyond breaking point.

We could move to a hierarchical system — the devices on the IoE remain relatively dumb, with highly intelligent centralized SDN management/control systems.  Between these could sit network appliances that are tuned to deal specifically with IoE traffic, filtering out what is true noise level, only leaving data that is pertinent to go further into the network.

But who makes the decision of what is pertinent? My Hive heating controller could be sending out packets of data saying that the room is at the same temperature as it was before, or my Foscam webcam sending out pictures of an unchanging environment — is that “noise” or important?  If I have set the Hive controller to keep a room a set temperature, then it is probably noise level.  However, if I have set the heating to be off, the fact that the room temperature is not falling could indicate a possible problem.  Without the intermediate appliance understanding how Hive works and how I am using it, a lot of false positives and negatives could occur.

How about something like Google Glass?  This could well be pushing out details of where I am on a regular basis: could a large proportion of this data be dropped, as position could well be estimated based on existing direction and speed from an earlier data point?  Possibly — until I suddenly change direction or come to a sudden halt, and the augmented reality being presented to me through Glass becomes disconnected from the actual reality.

Apple Watch?  All that health data — a lot of it is more of the “no change, no change” nature — probably noise level again.  However, as we see functionality grow, the health aspects of wearables could become a core part of how we interact with health services around us, and what was noise level before could become a matter of life and death.

No, a hierarchical approach of dealing with IoE data is unlikely to work, apart from providing aggregation services for data that can be packaged in a different way so as to have less impact on the global internet itself.

It will become incumbent on device designers to build in more intelligence to each device, ensuring that network traffic is minimized to only that data that is needed at that time.  Sure, it will cost more.  Sure, it will require new chip and battery design to make sure that the devices are still attractive to the prospective buyer.  But, without such contextual data and network awareness, the explosion in IoE devices will bring the global internet to its knees.

About the author
Clive Longbottom