IoT Trends: AI/ML and Edge Computing in IoT
IoT Trends: AI/ML and Edge Computing in IoT
There is something magical about the “edge.” For example, in environmental science, one studies “habitat borders” where certain varieties of plants grow heavily at the edge but no further. Similarly, in astronomy, we have seen dramatic phenomena happening at the edge of the universe.
It seems no different with human societies as a new revolution is brewing with high computational powers moving to the edge in what is increasingly called – Edge Computing.
Industry 4.0 and IoT
At its heart, IoT encompasses collecting and analyzing data, insights, and automation of processes involving machines, people, things, and places. IoT is, therefore, a combination of sensors, actuators, connectivity, cloud and edge computing and storage, AI and ML intelligence, and security.
Industry 4.0, also known as the fourth industrial revolution, relies heavily on IoT technology, redefining automotive, transport, healthcare, public mobility, retail, and commerce. Since AI and ML drive all 5G and IoT innovations, their coming together is a natural phenomenon.
5G Supporting Strong IoT
One of the most significant innovations of 5G is providing strong support for IoT, including support for low-cost, long-battery life sensors.
As data storage and computing become a ‘fabric continuum.’ That fabric touches all sectors, and a slew of AI, ML, and Edge Compute solutions will eventually find their optimum zones to reside and roles to play depending on the use case.
Hyperscalers
A new breed of cloud service providers, such as Google, Amazon, and Microsoft, introduced an entirely new scale of IaaS (Infrastructure as a Service), with manageability, transparency, and analytics at competitive prices.
Connectivity
A plethora of options from LoRAWAN, NB-IoT, and rapidly proliferating 5G made offering advanced IoT solutions easier and more feasible.
Edge Computing
Massive advances in computing silicon have made edge devices powerful – allowing localized data assessment and real-time decision making, totally avoiding the round trip to the server.
Some of the other transformations are beginning to be seen, more visibly in terms of engagements of players. For example, telecom operators with Hyperscalers (mentioned above) and transition of Operations Technology (OT such as machine control systems moving into the IT domain).
An increasing number of IoT applications are being conceived and built in the hope they will be consumed and paid for by enterprises and consumers. However, for various reasons, it is clear it will be best delivered from a location closest to the customer, and that zone is a hotbed of activity.
The space closet to the enterprise and consumers is the space that all these players want to dominate.
This space will essentially mean a shift in workloads from central data centers to near-premise or on-premises dedicated edge compute cloud. The central data centers shift could sometimes mean moving the on-premises data center to an on-premises dedicated edge compute cloud.
This suggests a shift from the Capex model to an Opex model, impacting enterprise financial models.
Edge Cloud in an IoT world
In some ways, this increases a new direction in the value chain. The cloud is moving toward the edge, in what is now called the industry as the Edge Cloud.
A Straightforward Explanation of Edge Cloud
The Edge Cloud is simply the combination of smart edge devices (including sensors, nodes, and gateways) with software (algorithms, security stacks, connectivity modules, sensing and actuating components, a processor in a full stack) to handle hundreds of sensors per gateway.
Thousands of such nodes come together automatically with advanced clustering and nodal reconfiguration, re-routing in case of failure in some sort of a super-intelligent IoT network.
Where Does the Tech for Cloud Come From?
The various technologies are already in place, and scalable use cases are becoming common. They even have standardized protocols to offload data to the cloud for non-critical data processing, but retain real-time data processing at the Edge Cloud itself.
Can We Stop AI or ML?
As Silicon volumes build and prices drop, we expect millions of IoT devices to communicate via Edge Clouds and aid human life. The movement of AI/ML toward the edge is an irreversible process and, in fact, a necessity.
The following benefits are clearly visible to the industries that deploy these technologies:
- RT processing ensures low latency responses (improves safety, reduces defect rate)
- Localized security of data
- Data sorting, filtering, and pre-processing before it hits the cloud (eases cloud load)
- The combination of licensed and unlicensed spectrum creates more efficient transport mechanisms.
AI image processing, object detection, and audio/video recognition capabilities have also improved significantly. They are now available as additional capabilities with certain Silicon vendors. We expect these to become even more prevalent and improved pricing as deployments grow.
Is this Edge Cloud for real?
Is there a overhype around Edge Computing? Several assumptions are being made that fuel this extraordinary boom in Edge cloud theory:
- It is still unclear if the Edge Cloud market will be as large as it is made out to be. There is evidence that it can be a high-growth sector. However, the ownership elements of this service are still being defined; telco versus hyperscaler versus a specialist enterprise system integrator versus a Cloud Edge specialist.
- Edge Cloud is a cure-all versus use-case-specific solution. What applications require Edge Cloud capabilities (perhaps those in manufacturing RPA and healthcare versus retail applications) is unclear. Companies are still evaluating the business cases, and the cost of models will play out over time.
- Is Edge Computing a solution looking for an application? Is the round trip delay of a few micro-seconds making the consumer less ready to spend more? Are we mistaking edge storage for edge computing and treating all those applications the same?
4.
What about the sunk cost of centralized data centers? Will they now be under-utilized or end up storing only non-critical data? If everything Edge Computing is about Real-Time, then 5G slicing could be a vital element of the solution. If that is the case, will a sustainable business model build-out between telcos and hyperscalers.
Conclusion
These revelations will come to light soon. As the space evolves further, it will become more apparent how AI/ML technologies will realign in the new world order. This will possibly define new positions of the service providers, their alliances with solution providers, and combined business models.
The combination of AI/ML will eventually settle down on a unique landscape that will drive consumers and users in the next decade.
The post IoT Trends: AI/ML and Edge Computing in IoT appeared first on ReadWrite.
(48)
