Main menu


Edge computing towards 'complete autonomy' in both deployment and operation


Edge computing towards 'complete autonomy' in both deployment and operation

Edge computing towards 'complete autonomy' in both deployment and operation

With edge computing quickly becoming a superficial concept, users and vendors alike are now focusing on the next goal of edge technology: fully autonomous deployment and operation.

Automated technology that simplifies the inherent complexity of the edge in deploying applications, networks and security, says Teresa Tung, chief cloud technologist at IT advisory and consulting firm Accenture. It is appearing,” he said.

“If the ability to create and manage containerized applications allows for seamless development and deployment in the cloud, the edge is becoming a more tightly resource-constrained specialty,” says Tung. “Zigbee, Z-Wave ), ISA100.11a, and self-organizing and self-healing wireless mesh communication protocols such as WirelessHART can create networks that allow for on-the-fly deployment and self-configuration of devices.”

Matteo Galina, a senior consultant at ISG, a global technology research and advisory firm, said that decentralization of the IT environment encompassing edge systems also has challenges to overcome. , it must be done outside the traditional management domain, such as various security requirements. “As systems become larger and more distributed, the role of automation in ensuring effectiveness and reliability becomes even more important.”

Automated innovation driven by the open source community

The automation trend in edge environments resembles the journey towards AI in that open source groups, infrastructure manufacturers, and cloud service providers are driving innovation, Tung said. He said that it is leading innovation in areas such as management and setting important standards.

“Infrastructure providers are creating solutions that allow computing to run anywhere and embedded in anything,” Tung said. This includes new hardware capabilities that are ultra-low-power, ultra-fast, connect everywhere, and offer extremely high security and privacy features. 5G opens up new opportunities for network equipment providers and carriers to innovate in private and public networks with built-in edge computing capabilities.”

At the same time, cloud provider innovations make it easy to extend centralized cloud devops and management practices to the edge. “Just as a central cloud facilitates access to services for all developers, the same is happening now with technologies like 5G, robotics, digital twins and IoT,” said Tung.

Ron Howell, network architect at Capgemini, said software-defined integration of multiple network services has emerged as the most important technological approach to automating edge environments. Network security with zero trust deployment methods incorporating SASE edge capabilities can significantly enhance automation and simplify the process of deploying and monitoring edge computing solutions. Additionally, full-stack observability tools and methods that employ AIops will help proactively maintain the availability and reliability of data and edge computing resources.

AI applied at the edge of the network is now recognized as a leading vehicle towards network edge availability. "AIops is a significant improvement when used as a form of full-stack observability," Howell said.

There are already many options available to organizations moving toward edge autonomy. “It starts with the onboarding and management of physical and functional assets, and includes automated software and security updates and automated device testing,” Ganilla said. If a device works with some form of ML or AI capability, it can be located at the device level and on-premises or in a centralized edge system to ensure that local ML models are up-to-date and that the right decisions are made in every situation. All within the backbone ML/AI that exist requires AIOps.

Where physical and digital experiences mix

Tonne uses the term “phygital” to describe the consequences of digital methods applied to physical experiences. For example, autonomous management of edge data centers. “The ultimate goal is to create highly personalized, adaptive physical experiences,” said Tung. “In the physical world, anyone can imagine, build and extend the experience.”

In edge computing environments that integrate digital processes and physical devices, direct network management is significantly reduced or eliminated, network failures and downtime are automatically detected and resolved, and configurations are applied consistently across the infrastructure, making scalability easier and faster. .

Automatic data quality management is another potential benefit. “A combination of sensor data, edge analytics or natural language processing (NLP) is used to control the system and provide data in the field,” Ganilla said. Another area where autonomous edge environments will benefit enterprises is “zero-touch” remote large-scale hardware provisioning. OS and system software are automatically downloaded from the cloud.

“Commercialized edge applications and marketplaces are starting to appear, and the number of open source projects is growing,” Ganilla said.

Vendors are also developing solutions to seamlessly manage virtually any type of edge asset, regardless of the underlying technology. Ganilla said that edge-oriented open source software projects, such as those hosted by the Linux Foundation, could speed up large-scale adoption.

AI-optimized hardware is an emerging edge computing technology, and many products offer interoperability and resiliency, Ganilla said. It is likely to expand exponentially over the next few years.”

Leading Companies in Edge Automation AI Area

There are already many technologies available to businesses considering edge automation, including those from hyperscaler developers and other specialized providers. An example is KubeEdge, which provides Kubernetes, an open source system for automating the deployment, scaling, and management of containerized applications.

Ganilla noted that in 2021, ISG classified systems integrators Atos, Capgemini, Cognizant, Harman, IBM and Siemens as global leaders in edge AI technologies. Major edge computing vendors include Hyperscaler (AWS, Azure, Google) and edge platform providers ClearBlade and IBM. In the telecom operator market, Verizon stands out.

Edge capabilities, autonomy and reliability

Vendors are building digital and physical availability capabilities into their products to increase the autonomy and reliability of edge technologies. Ganilla said the two common ways vendors provide autonomy and reliability are internal sensors and spare hardware components.

For example, built-in sensors can use field monitoring to control the environment, detect and report anomalies, and in some cases combine with failover components to provide the required level of redundancy.

Tung mentioned other methods:

• Physical tamper protection to protect devices from unauthorized access

• Security identifiers embedded in chipset for easy and reliable device authentication

• Self-organizing network protocols based on ad hoc and mesh networks to ensure connectivity wherever possible

• Partitioned boot configuration to apply updates without risk of device downtime due to update installation issues

• Hardware watchdog capabilities to ensure that devices automatically restart if they become unresponsive

• Integrity checks at boot time from a secure root of trust to protect devices from malicious hardware installations • A firewall

with anomaly detection that identifies abnormal behavior and alerts you to failures or unauthorized access.

Self-optimization and AI

Networks require a myriad of configuration settings and fine-tuning for efficient operation. “Wi-Fi networks need to adapt to signal strength, firewalls need to be constantly updated to support new threat vectors, and edge routers need constant configuration changes to fulfill service level agreements (SLAs),” said Patrick Millampy, Juniper Fellow at Juniper Networks. need. “By automating almost all of these tasks, we can reduce human labor and errors.”

He said Tonne operates at the edge and requires self-optimization and AI to determine how to handle changes. For example, what happens if the network goes down, there is a power outage, or the camera is out of sync? And what should be done once the problem is fixed? "The Edge doesn't scale if you need manual intervention every time in these situations," Tung warns. Troubleshooting is possible by implementing rules that detect situations and prioritize application deployments accordingly.


“The edge is not just one technology, but a combination of technologies that work together to support a whole new topology that can seamlessly connect data, AI and work,” said Tung. “The biggest innovation has yet to come true.”

Meanwhile, Howell is complementing this with more and smaller network edge centers located closer to customer demand, and large-scale cloud services that can handle additional workloads that are less critical and sensitive to time and latency. said that the weight is moving toward the One thing has not changed, Howell said: "The first rule of the data center - high quality service always available - has not changed."