Edge computing technology development trends
The origins of edge computing lie in content delivery networks created in the late 1990s to serve web content from edge servers deployed close to users.
Edge computing involves hardware and software technologies that enable storage, computing, processing and networking close to the device that generates or consumes data.
‘Close’ is a relative term that could include: servers co-located with cell towers, on premise servers, gateways, and operational devices. Read on to discover the latest trends in industrial edge computing.
Event Driven Architecture mimics real life processes and business logic
What is it?
Event driven architecture (EDA) is an application architecture that monitors and broadcasts incoming events within a network. Each system in the network can independently decide what they want to do about events by subscribing to events.
Event driven architecture is especially suitable for edge computing because the events need to be processed near the edge.
- Business Logic: EDA follows business logic and mimics real-life operations.
- Data Usability: EDA enables data collected to be used immediately by all the applications that may need it.
- Event notifications will form 60+% of new digital business solutions. Using event driven architecture will enable higher performing IoT solutions. (Solace)
EDA benefits systems that need to have immediate action triggers when events happen.
- Alert systems for condition monitoring.
- Push notifications for consumers to receive promotions, tips, or suggestions.
Hardware vendors are developing specialized integrated circuits with unique edge computing capabilities for specific tasks, e.g., vision
What is it?
Specialized processors will be developed to optimize performance for each type of computing. Application Specific Integrated Circuits can be optimized to do a single task with maximum efficiency. An extension of ASICs are programmable field programmable gate arrays, which can be reprogrammed for different tasks.
- Higher performance demands with lower resources: As more computing moves to the edge, higher performance is needed at the edge with lesser resources available.
- Requirements for specific tasks: Usually, edge computing devices perform a limited range of tasks repeatedly. It is not efficient to use general-purpose chips as they are not optimized for that single purpose.
- ASIC market will grow 45% CAGR from 2020 to 2026, reaching USD 32 B in 2026. (GM Insights). Chip manufacturers will need to produce more chip types at smaller volumes.
- As use cases mature, the specialization of each device will increase and performance requirements will also increase.
ASIC performs best for specific, repeated actions performed in high frequency.
- Image / facial recognition (visual processing units)
- Training of distributed neural network models in AI (AI chips)
Device to device communication enables swarm intelligence / federated learning instead of centralized, cloud-based machine learning
What is it?
Devices of the same level communicate with each other to transmit information without a control layer. The device can act like a sensor, actuator, and processor
all in one, to reduce reliance on control layers like cloud platforms.
- Hardware feasibility: As field and edge devices get smarter, they can gather data, make decisions, and act on these decisions. Communication with other devices at the same level helps them get the information they need to make these decisions.
- Data volume: Machine learning at the cloud will become increasingly expensive as data grows.
- Robots and analytics applications will be developed to make decisions to aggregate data from surrounding devices and make decisions. The swarm intelligence market is projected to grow at 40.47% CAGR from 2020 to USD 447 M 2030. (Markets & Markets)
New use cases that are not possible today will be enabled with swarm intelligence as the requirement for a reliable control layer decreases. Such use cases usually involve many devices moving alongside each other, such as nanorobots, miniature drones.
- Vehicle to vehicle communication about road status, routing, breakdowns
- Autonomous AGVs or drones to transport goods
Edge containerization enables deployment of lightweight applications on edge devices and simplifies mass deployment
What is it?
Edge containers are decentralized computing resources located as close as possible to the end user. It virtualizes the the entire operating system. Each container comes packaged with its own user space to enable multiple containers to run on a single host hardware. Containers do not include the operating system and can be deployed on any device with container runtime.
- Easy portability of applications across different hardware: Applications running in containers can be deployed easily to multiple different operating systems and hardware.
- Simplified device management: Applications in containers will run the same, regardless of where they are deployed.
- Resource constraints of edge devices: Edge devices have resource limitations. The lightness of edge containers enables more applications to be run on edge devices.
- Future proof: Modular software models can allow individual modules to be upgraded independently as requirements change.
- Application container market will grow at 30.8% CAGR from 2017 to 2022 to reach USD 4.3 B. 48% of enterprise end users are already implementing or plan to implement container technologies. (451 Research)
Architecture that can be used horizontally across different use cases. Best for devices with low processing capabilities, and for applications that would be used across many different devices.
Multi-access edge provided by telcos serves the majority of current edge computing needs well, without significant CAPEX investment
What is it?
Multi-access edge computing allows data to be processed before it goes to the network core.
It is enabled by localized and proximate edge data centers 5-10km from data source.
- Increasing data volume to send to cloud: Although core networks and centralized clouds support the use of smart devices now, they will not be able to handle the amount of data created by IoT devices.
- Pure edge computing is not mature enough: Edge devices lack compute power so localized edge data centers are used to supplement the core cloud offering to support the use of latency-sensitive applications for end users.
- Product familiarity: It is also easy for end users to accept and understand this product as it is more like a private cloud with similar pricing models.
- The global MEC market size is anticipated to reach USD 15 B by 2027 (Grand View Research) driven mostly by telco operators (Business Insider).
- Edge data center demand will grow along with MEC growth.
MEC is most useful to enable “Intranet” type use cases, where the devices within the same zone are primarily communicating with each other.
- Smart campus / smart building / smart factory / smart retail stores
- Content delivery / streaming services