A distributed architecture for IoT 2.0

In my previous article, I have outlined the issues with the existing client-server approach taken towards current IoT solution. In this article, I would like to outline few of the key requirements for the next generation of IoT architecture.

The architecture of the future IoT system has to abolish the central role played by the cloud. Instead, the next generation of IoT systems will be decentralized and self-managed, with minimum human intervention needed. The devices in the field have the real time information and the relevant context to make a right decision by talking and collaborating with each other. The cloud platform of the future will assist them in this decision-making process, by supplying machine learning intelligence and the global contextual information, to make a better decision.

Here are the few key differentiating attributes for IoT 2.0 system:

A) Fully Distributed

There will be no central component in the next generation of IoT architecture and therefore no single point of failure. Devices in the vicinity will auto-discover and create self-managed groups, which will further synchronize with other groups. The role of the cloud will change from its current authoritarian to assistive role. The boundaries between cloud and edge will cease to exists and the entire system will look like a mesh network of cooperating nodes.

What does this mean? Every feature of the IoT intelligence will be distributed throughout the components of the system. Every node will have storage; every node will do a varying degree of analytics. Every node will publish data, forward from one to other and subscribe to what it needs. Some node will be heavily optimized for one operation (such as storing large data) however; the overall system functionality will not be partitioned to one component.

B) Collaborative

The devices will publish, subscribe and forward data between each other in a mesh network topology. They will form a global namespace, which is organized by topic hierarchies. Certain nodes (such as the existing cloud servers) will have a specialized role and will subscribe to the entire global namespaces to perform system level analytics. These nodes will then publish the relevant machine learning models and weight factors which will be subscribed by the individual devices to improve their decision-making capacity. The accuracy of decision-making process shall increase as more nodes participate and more data is available in the system.

Let's take a practical example of this.

Imagine a gas leak sensor and a gas valve installed in the pipeline. In the current architecture, a central system monitors the leak via the sensor and if a leak happens, it instructs the gas valve to shut down. In the IoT 2.0 system, the leak sensor will be able to directly instruct the valve to shut down. How would the sensor detect if it is a false positive? By subscribing to the historical analysis data from a remote node, which has performed statistical analysis over actual gas leak incidences from last several years. Such a fully independent system will not fail even if a central component is not responsive.

C) Autonomous

By speaking a common language and schema, devices will be able to automatically work with each other and accommodate themselves into self-learning networks. When devices join each other, they will advertise their capabilities and ask for others. The more nodes participate, the better the network becomes. Devices across multiple vendors, types, and protocols shall be able to interoperate in conjunction. How can this happen? By using decentralized methods of authentication and using peer to peer trust based upon reputation models. A node will need to work hard and perform its work efficiently to earn the trust of the network.

A cooperating network of agriculture sensors will automatically analyze soil moisture, sunlight, wind speed and ambient temperature data and determine the perfect time for irrigating the crops. The human intervention will be no longer required for such tasks.

D) Interoperable

The era of IoT platform vendor locking will end with fully standardizes interfaces and standards. No vendor would be able to lock a customer into the system. This means any piece of hardware or software will be interchangeable and interoperable within the system. How would the cloud platform providers operate in this situation? By offering better algorithms for machine learning, offering better latency for communication, cheaper storage for saving large amounts of data, and most importantly by being the vendor of data itself. After all, there is a reason why IBM spent billions of dollars to acquire Weather underground.

Promising technologies for IoT 2.0?

All of the above sounds too good to be true. What are the technologies which will enable such transformation? Here are few of the examples.

IPFS: A fully decentralized peer to peer file distribution system, based on a unique amalgamation of BitTorrent, Git and DAG. IPFS promises to build a new version of the web which breaks away from the traditional client-server model. By introducing concepts like the self-certifying file system, it provides encryption along with authentication of the contents in a truly global namespace.

IOTA: A no-fees alternative to Blockchain offering decentralized micro payment. IOTA solves the problem of trust between nodes and allows resource sharing in a fully distributed way.

Autumn AI: An ultralight portable machine learning framework which could be ported to any edge device or gateway.

6LoWPAN: Based upon IEEE 802.15.4, the 6LoWPAN remains the most promising wireless protocol today due to its native features such as IPv6 and mesh networking support.

Which one of the existing cloud platforms are best suited to the needs of the next generation architecture?

From my current exposure in the IoT ecosystem, I could see companies like Atomiton have been pioneering in taking a decentralized and distributed approach to IoT. The big players like AWS are also moving into the distributed computing model with offering such as Greengrass. Companies such as CISCO have also realized the need for greater edge intelligence.

The future architecture for IoT 2.0 may still be far away from actual implementation but the requirements are getting clear with each new device we are getting into the field.

Your comments are always welcome and a great source of motivation!