Tracking Billions of Devices at the Edge

In the exploding field of IoT and Edge Computing, how do Enterprises keep track of, secure, and get rich insights from all the rapidly proliferating smart devices in their ecosystems? This was a subject we discussed recently with Manoj Rajshekar, Co-founder and Head of Product and AI at SmartHub.ai.

Explosion of Connected Devices

We began by asking Manoj to give us a broad picture of the IoT device landscape, both now and looking into the near future: “In terms of broad numbers or statistics, there are 21.7 billion connected devices in 2020, out of which almost 7.7 billion devices are IoT device connections. By 2025, there will be about 30 billion devices and this would comprise close to 75% of the overall data generated. The data that is being generated from IoT devices is way more than the data generated in a data center or anywhere in the cloud. This level of explosion is happening in every vertical we are looking at. All of these devices are very rich, and they do some amazing stuff, but the interoperability and surfacing the right information or the data deluge for human understanding is where the biggest challenge is right now. Besides the rich information the Enterprise Edge provides, Enterprises are also seeking possibilities of reacting at the Intelligent Edge before this data surfaces in the Datacenters or the Cloud.”

We asked Manoj to explain to us the benefits of Smart Hub’s Software Defined Edge Management solution in relation to this explosion in IoT devices: “Our technology heritage started from VMware, and virtualizing the whole ecosystem in a data center. The same aspect came into play in terms of any enterprise or environment where there is more data happening at the edge. The amount of connectivity that is exploding in this ecosystem is mind-boggling and the complexity arises because pretty much every IoT solution out there, or any IoT device per se, has its own application and its own way of interfacing with that particular device. If you take an example of an enterprise, you might have the facilities and Physical security teams who are responsible for the physical security cameras and all other devices on the floor, and amongst the physical security cameras, you have variants of these cameras where every camera vendor will have his own application to manage these devices. They would also be responsible for all the elevators, the badging systems, counting people, fire alarms, the list goes on. And the complex question is how can they get a unified view across whatever is happening in their environment? The Enterprise Edge plays an extremely critical role in the safety and wellbeing of the employees. Now if we enter the healthcare or retail industry, the complexity explodes, because the number of devices at the edge is far more complex and they generate a lot more information and these devices have a direct impact on the revenues or cost optimization for the Enterprises, besides the safety and regulatory needs.”

Providing Visibility to an Edge Environment

Industry research is pointing to devices like fitness trackers, smart watches and even pacemakers as potential vulnerabilities for hacking.

According to Forrester research the challenge for enterprises is significant.  A recent study conducted by them indicates the following:

• 69% of enterprises have more IoT devices on their networks than computers

• 67% of enterprises have experienced an IoT security incident

• Only 16% of enterprise security managers say they have adequate visibility to the IoT devices in their environments

Hostile international actors have already managed to hack into enterprises via security cameras at a Tesla plant, into an enterprise network via a vending machine and there are reports of hacks occurring via a thermostat.

Manoj described how providing visibility into all the complexity occurring in an edge environment is key: “For us, this is where the conversation starts, when people don't have complete visibility of what's there in their environment. This poses a lot of security challenges also because if you look at all the recent malware, or any of the breaches, a large number of them are all happening through the IoT devices. What we do is we bring visibility from the discovery stage all the way to management. We have capabilities where we can remediate when there is any anomaly in behavior or any anomaly in a pattern of usage of a particular device. So we would know how to isolate them, or we would know how to update the switch patches or send instructions that would augment or secure those devices. Besides security, you have other benefits, so as you capture these devices, you're capturing enormous amounts of data. And with that data comes the opportunity for real machine learning, or artificial intelligence or big data type analysis, where you can actually really begin to see things. For example, we come up with our own risk score which is a combination of various aspects such as: what does the community say about these vulnerabilities? What level of exposure does it provide? And in many cases, it might depend on the business use case.”

Connecting Data Silos

Manoj described the importance of connecting siloed data sources: “Another aspect which becomes very key in our Software Defined Edge solution is if I know how to communicate from one data silo to another silo of its own, then I should be able to start looking at orchestrating some of the actions. So for example, if I see something on my camera, can I enable certain aspects on my digital display? And if I see any occupancy in a certain aspect of a building through my people counters or my occupancy sensors, then can I turn on or turn off lights or heat to optimize my energy utilization in that building? A subsequent aspect would be if I see any anomaly, can I send instructions that can schedule certain kinds of tasks on many of these devices? We also provide real time information on the health of these devices and the status of where these things are currently located. You could start looking at what's your ROI on these investments in terms of how much revenue is being made, or what's the utilization of these particular devices.”

Distributing Workloads on the Edge vs the Cloud

Manoj also described the importance of using AI to determine when computations should happen on the edge or in the cloud: “Then there is the AI element and the ability to run some ML models to do data analysis and indicate our pattern identification. A good example would be if there is a huge amount of information which is being generated at the edge, do I need to ship everything to the cloud? Or can I start identifying what are the important aspects of it, and then send some parts of that learning onto the cloud, and the rest can be defined at the edge? The other aspect is that, specifically in manufacturing and various other critical verticals, there are systems where decisions have to be taken in milliseconds. Many times, you don't want to rely upon the connectivity to the cloud and your decisions being formulated over there. In the edge world, we can run any of these lightweight workloads or these ML models which know how to identify patterns. There could be models which will detect certain anomalies or certain behaviors, and there could be models which will detect any anomalies in the traffic patterns for security reasons, for example.

In an Enterprise's critical infrastructure, continual uptime, ability to prevent and diagnose anomalies and failures in real-time is preeminent. Enabling predictive maintenance using AI models from the Edge data is becoming paramount. And with the new norm of the remote workforce in almost all industry verticals, there's a big need to equip operational teams with necessary information from the Edge to prevent downtime and enable maintenance teams and contractors to focus on imminent issues rather than around the clock inspection and manual monitoring.”

We are grateful to Manoj for his time and for sharing this information with us about the many complexities of managing rapidly growing Enterprise Edge environments.