Fog computing utlizes computing resources closer to the users, i.e edge devices.

The terminology used for Fog and Cloud is analogous to actual cloud and fog in atmosphere. Cloud is much higher in the sky and fog is much more close to the earth surface and therefore much dense and covers bigger area.

Fog computing allows cloud resources to shed some of their load to edge devices and therefore fog complements the cloud.

For example, the sensors can collect the raw data, and instead of sending all the collected data to the cloud for further processing, fog computing can be utilized to perform some processing such as filtering, prediction, classification, etc etc. Then the final concise information can then be sent to the cloud for storage and further data analysis. Obviously any algorithms can be applied at the edges including ML.

What is Fog Computing?

Fog Computing is the capability to compute at the distributed area / network closer to the IoT sensors & actuators e.g. at the IoT gateways. You can also refer to the following links:

https://info.opto22.com/fog-vs-edge-computing

https://www.automationworld.com/fog-computing-vs-edge-computing-whats-difference

Can it be link with Machine learning?

Personally I think fog computing can link to machine learning in order to make fog computing more intelligent & versatile. However, at the moment fog computing as its name suggests - might have limited computing power & storage capacity vs cloud computing in which the latter is located in the back end at large DCs. Think we need to strike a balance on what use cases / scenarios that require machine learning at the edge nearer to all the IoT sensors & actuators e.g. short-term event, event require low latency etc.

Fog is cloud near customers. In other words cloud companies attempt to move or create data centers near users. One of the most advantages of this platform is the computational resources that iot and smatphone users may use to overcome storage and computational issues.

It can be connected easily with machine learning for prediction, estimation and classification....

Dear Shivam Vatshayan,

Fog computing (FC) can be seen as complementary to cloud computing (CC). The importance of FC for the Internet of Things (IoT) illustrates my figure. FC represents an extremely significant development of CC.

Typically, the FC is implemented near the Internet Edge (using a type of micro-cloud). These "micro-clouds" are installed near IoT access gateways and are often referred to as "Fog Nodes". To provide a widely geographically distributed Near Real-Time Services, e.g. to realize Smart City or Smart Healthcare, multiple Fog Nodes often need to be spatially "distributed" and coordinated from a central cloud.

FC is a distributed system architecture, especially for data processing, storage and control. These FC functions are brought closer to IoT devices using cloud nodes along the Internet continuum between clouds and things. The FC architectures across the Internet Edge help bring the "intelligence" from the clouds closer to the real "things" that "intelligence" needs.

The FC Intelligence - as Artificial Intelligence (AI) - is provided by Fog Nodes in cooperation with Clouds. In Fog Nodes various types of AI learning can be realized - such as e.g. Machine Learning (ML), Deep Learning (DL). Thanks to the intelligence of Fog Nodes, for example, we speak of Intelligent IoT.

The IEEE standard ROOF defines the principles for both roof computing and interoperability of IoT systems (for example, at home, factory, workplace, etc.) so they can cooperate in real time.

Best regards

Anatol Badach