The lesser heard challenges of IoT

In mid 2015, US ran out of IPv4 addresses. This is the 32-bit addressing system we’ve relied on since the 80s. And it supports around 4.3 billion IPs.

Mid of 2016, a company in the UK strapped around a dozen pigeons with small backpacks which hosted an air quality monitoring unit, and set them free across the city of London. One could then tweet and get local air quality metrics as a response.  ​

I often refer to it as the 'Internet of Pigeons'.

The two events are completely disconnected. But two takeaways:
  1. We’re adding connected devices at an exponential pace.
  2. Since a while now, it’s been rather easy to put together connected devices. Integration takes work no doubt, but the technology itself is rather mainstream. 
Many imagine IoT to be these discretely sized, yet swarm like, ubiquitous devices. But whether its cars, homes, warehouses or personal assistants, they’re all getting connected. They’re all IoT.

Typically, security is the most discussed concern when it comes to internet connected things. So, I’ve put together a list of some of the lesser heard challenges of IoT.

1. IP Addressing

A system may utilise high data, yet another may not. But irrespective, every connected device must have an address. We’ve already exhausted IPv4 and most organisations aren’t ready to switch to IPv6. The primary issue is that though most systems are capable of handling 128-bit IPv6 addresses, most networks still aren’t. It’s not the end of the internet though. We have several ways to deal with IPv4 exhaustion.

2. Lack of Standards

There are probably more IoT standards bodies and IoT industry consortiums than IoT standards. And as of today, there are no clear winners. For engineering and product ecosystem to evolve, we’ll need standards so this collective knowledge can grow faster.

3. Human Vs Industrial IoT

These two areas of deployment demand very different device and connectivity attributes. While a human IoT product’s lifecycle could depend on what’s in vogue, an industrial IoT product could be used until its dead or obsolete. Not just not when it comes to product lifecycle, but even network topologies, internet accessibility and response-to-failure attributes differ quite vastly.

4. Energy Harvesting

IoT deployments could occur in spaces where its either difficult or prohibitive to provide a reliable source of power. Think deep mines, body implants, ocean beds, hectares of farm lands. In such cases, it becomes imperative for the device to able to harvest and store energy whenever its available. Harvesting could be achieved out of solar, geothermal, or even via salinity gradients or piezo-electrics.

The point being that for industrial IoT at least, power management and harvesting is crucial.

5. Need for contextual computing

A connected device which is always dependant on a cloud server will not only spend more energy in transmissions, but will be prone to delays, could get exposed to higher connectivity failures while also consuming network bandwidth.

We can reduce a device’s dependency on the network, by trying to do more and more processing locally and as close to the end-node as possible. So, based on the context, a device could take decisions at a closest gateway, rather than logging on the internet and depending on a centralized server to make such calls. A server round trip would only need to be made, when a larger decision making context in play.

But, being able to take decisions locally, could also necessitate the deployment of more compute capable MCUs in nodes and gateways.

6. Need for cloud’s evolution

In order to better manage capital and operational expenditures, our current network infrastructure would need to evolve into a more ‘intelligent’ one. Most networks currently act as dumb pipes carrying data packets back and forth between the devices and data centres. So, while data is consumed at one end of the network, all compute and storage capabilities lie at the other end. An evolution of this would be if compute, storage and acceleration capabilities could get spread across the pipe. This would significantly reduce the bandwidth needed for large scale deployments.

And that concludes the list.

• • •