What is edge computing?
Head of NG Cloud Lab, Atos Research and Innovation and member of the Scientific Community
Distinguished Expert & Cloud Domain lead
Posted on: 10 July 2019
Edge computing locates computing and storage resources at the edge of the network, with the intention of getting data and analytics to the right place, at the right time. It decentralises data processing and avoids non-essential data transmission.
There are many definitions of edge computing currently used in the market, we believe the following is the most accurate from the Linux Foundation Open Edge Computing Glossary which defines the Edge Cloud as:
“Cloud-like capabilities located at the infrastructure edge, including from the user perspective access to elastically-allocated compute, data storage and network resources. Often operated as a seamless extension of a centralized public or private cloud, constructed from micro data centers deployed at the infrastructure edge.”
A common topology for an edge computing installation is composed of three-layers. From bottom to top these are:
- IoT devices: IoT devices are connected to an edge device. IoT devices communicate via diverse communication protocols with the edge environment acting as data sources.
- Edge nodes: they enable data processing close to sources of data through near real-time data analytics and model execution. It offers diverse communication and messaging protocols for data acquisition from near-by IoT devices and acts as temporal data storage.
- Cloud services: they develop management functionalities for both edge and IoT devices and they perform long-term data storage and analytics. Additionally, they provide the point for integration with other enterprise systems.
In some advanced scenarios, such as autonomous vehicles, the limits among these three layers blur, the car being a rich environment of IoT devices and edge capacity all integrated in a single device.
The recognized advantages of edge computing are:
- Privacy: to avoid sending all data to be stored and processed on cloud servers.
- Connectivity: to avoid and reduce costs associated with streaming/sending all raw data to cloud services.
- Latency and reliability: to improve responsiveness and reliability by maximizing processing at the edge and thus minimizing dependence on cloud connection.
- Wide physical distribution: in contrast to centralized cloud, the services and applications in edge computing are decentralized and distributed by nature.
- Need for near real-time interactions: edge computing aims to enable IoT actuation scenarios by bringing computation close to data sources. In doing so, there is a strong requirement for immediate response times.
- Autonomy: edge computing installations are not subject to data center boundaries and well-known fault tolerance practices.
- Heterogeneity: Heterogeneity is a key characteristic of edge computing which is present both in the variety of devices which form and connect to edge computing instances, but also in the diversity of network and security protocols they are exposed to.
- Interoperability and federation: typically, an edge installation works together with a federated cloud offering. For this, both edge and cloud environments interoperate for operations such as edge device management or long-term data storage.
You can read more about edge computing technology, use cases and market developments in our edge computing white paper.
From September 10th to 12th, over 400 people including Atos technical experts, partners and start-ups will gather during our 2nd Expert convention in Madridto build the future of technology and to instil and drive technological changes that will help our clients master these challenges in a sustainable and ethical way.