Edge Computing(AI/ML): The Fog rises and envelops the Cloud
Edge computing worth $28.84 Billion By 2025
Written by Arvind Tiwary and Sayan Chakraborty
Sections
- Edge is the seat of Ambient Intelligence
- Cloud
- Fog 2 Edge from Cloud
- Market Size
- Usage Scenarios
- AI/ML on the Edge
- AI Chips
- Cloud Providers embrace Edge
Edge is the seat of Ambient Intelligence
What does a wind turbine, a bulldozer, an automobile, a typical urban home, and a millennial teenager have in common? The answer is all of them are dependent on data generated and consumed using some form of a computer close to them in order to function.
This phenomenon of having computation capability available and providing value from data in devices close to the data sources or data consumers is called edge computing. The rise of edge computing and its benefits have to be understood in context of the historical place and role of computing in the consumer and industrial fields
Gartner estimates, 75 percent of enterprise-generated data will be processed at the edge (versus the cloud) within the next four years, up from less than 10 percent today. The move to the edge will be driven not only by the vast increase in data, but also the need for higher fidelity analysis, lower latency requirements, security issues and huge cost advantages.
The Linux Foundation has a LF Edge organization to create a common framework for hardware and software standards and best practices critical to sustaining current and future generations of IoT and edge devices.
Cloud
Cloud computing has now become mainstream. Jeff Bezos and the AWS team may have or may not have known the crown jewel they created from the 2004 pre public to public launch in 2006 but Cloud or Infrastructure as a Service (IaaS) has now taken over as mainstream for Information technology development as well as production use for service supply. The US $10B Pentagon Cloud Contract awarded in October 2019 is a ringing endorsement of Cloud based IT services as the preferred approach. On premises are no longer the first choice.
SaaS, PaaS, and IaaS are simply three ways to describe how you can use the cloud for your business.
- IaaS: cloud-based services, pay-as-you-go for services such as storage, networking, and virtualization.
- PaaS: hardware and software tools available over the internet.
- SaaS: software that’s available via a third-party over the internet.
- On-premise: software that’s installed in the same building as your business.
Shared services 24X7 availability like Wikipedia have altered expectations of service and usage. The big providers like AWS, Azure, Google Cloud Platform, Alibaba Cloud are complemented by scores of developer oriented like GitHub, Heroku, Digital Ocean, PythonAnywhere etc.
IoT was initially proposed by most consultants and writers as a cloud extension. This was the thin client architecture where low cost billions of devices act like slaves and send data to cloud based master and act on instructions received.
Fog 2 Edge from Cloud
Fog computing was coined by CISCO and envisages multiple compute points in an IoT network.The Open Fog Consortium now merged into Industrial Internet Consortium had a working definition and architecture.
We prefer the IoTWiki operating difference between endpoint as edge computing. Edge computing is when the endpoint which maybe a limited function device with small amount of compute capacity ( even a MCU , not a Raspberry PI) and small amount of memory and may actually run on battery does significant work.
Type of work done by the device can be limited for example in terms of RFID reader or could be a little more functional like limited vibration recognition for industrial devices. It may involve more significant process running inference with a miniature or lite AI/ML algorithm.
Market Size
CB Insights market sizing tool estimates Edge computing to be $6.72B by 2022
Gartner in “What Edge computing means…”
blog estimates that currently, around 10% of enterprise-generated data is created and processed outside a traditional centralized data center or cloud. By 2022, Gartner predicts this figure will reach 50 percent.
IDC FutureScape predicts the IT spend on edge infrastructure will reach up to 18% of the total spend on IoT infrastructure.
Grand View Research estimates Edge Computing Market Size Worth $28.84 Billion By 2025 | CAGR 54%
Usage Scenarios
There are a number of scenarios for the advantages of latency, reliability and lower cost of edge computing. IoTForum has seen several startups in India exploiting edge computing. Usage areas cover inspection, compliance and process controls in [ see here ,here ,here and MINT] some articles from IoTForIndia :
Hubitat makes a strong case for the cloud-less smart home hub
Texas Startup Edgetensor Develops AI-based Driver Monitoring System for Autonomous Vehicles
- Smart City
- Health Care
- Transportation
- Smart Farms
- Utility Electric, Water, Gas Grid Control
- Autonomous Vehicles:
- Industrial Automation:
- Connected Homes/Offices:
- Retail:
AI/ML on the Edge
There is a Cambrian explosion of chipsets, software and Methodologies for Edge and Fog moving the momentum of innovation for IoT away from the Cloud. Headlines like Linux Foundation exec believes edge computing will be more important than cloud computing are now common. Toolset and methodologies to shrink large memory and compute hungry AI / ML solution to a Lite version with acceptable accuracy running on small power and memory and compute capacity of edge end points are being piloted.
The traditional ML tools especially neural net have needed significant memory and compute capacity as well as watts to KW of power to run . There was a strong business case of more local or edge based inference and has seen the development and promotion of edge AI/ML co-processors .Single board computers like Intel Movidus Neural compute stick, Google Coral, Nvidia Jetson Nano , SparkFun TensorFlowLite edge , ST micro MobileNet are examples shrinking the compute, memory and power consumption for AI/ML inferencing use.
AI Chips
Future advances with new AI chips being designed around FPGA or neuromorphic approaches and over 50 new AI chips are surging.
compiled list by Pallav Aggarwal is here.
Cloud Providers embrace Edge
Microsoft Azure IoT Edge is a fully managed service built on Azure IoT Hub. Deploy your cloud workloads — artificial intelligence, Azure and third-party services or your own business logic — to run on Internet of Things (IoT) edge devices via standard containers. By moving certain workloads to the edge of the network, your devices spend less time communicating with the cloud, react more quickly to local changes and operate reliably even in extended offline periods. Microsoft acquires Express Logic.
Amazon AWS IoT Greengrass, connected devices can run AWS Lambda functions, execute predictions based on machine learning models, keep device data in sync, and communicate with other devices securely — even when not connected to the Internet.
Google cloud IoT SDK: as an open source collection of Embedded C libraries that enables developers to “connect, provision, and manage” devices with Cloud IoT Core service. It specifically targets energy-, compute-, and size-constrained apps — like battery-powered cellular devices and Wi-Fi smart home devices. The Cloud IoT Device SDK works on hardware with as little as 25KB of flash memory (or 80KB with a TLS software solution) and uses non-blocking sockets to decrease power consumption.
Microchip has pioneered with simplifying the security setup for edge devices. They began with a pre installed AWS approach and now have a generic Industry’s First Pre-Provisioned Solutions for Deployments of Any Size
Cost is also reasonable for small batches going from 1.2–1.4 $ for lot of 10 to 0.77–0.88 $ for lot of 2000.
Azure Intelligent Gateway: Create highly-secured design with Microchip’s Microsoft Azure Certified intelligent edge solution based on the SAMA5D2 micrprocessor (MPU). This secured solution environment allows the Edge gateway device to be authenticated as part of the Azure cloud. This combination of hardware and software security provided by Microchip and Sequitur Labs enables the deployment of secured gateways certified for the Microsoft Azure ecosystem to prevent potential theft or loss of data.
The AWS-ECC508 is designed to provide end-to-end security between the IoT device and the cloud infrastructure. It does this by leveraging Amazon’s mutual authentication system, which verifies the identity of the cloud service and the device before any data or commands are accepted. The identities are based on cryptographic keys. Until now, creating such cryptographic identities relied on the original manufacturer — typically a contract manufacturer working for a device company — securely generating the keys and then passing the keys securely along the manufacturing chain. Instead, the AWS-ECC508 can generate its own keys that Amazon will accept as authentic.
IoTNext 2019: Open Source Edge AI /ML
At the upcoming IoTNext we will explore this at depth.
