Fog, Edge and Cloud Computing

Cloud

In a 2015 article on Medium, Cloud was defined still using the NIST (The National Institute of Standards and Technology) cloud definition expanded over the years to two pages. This is most accurate, long, academic, monotonous, boring definition of cloud computing. 

Cloud Factories

This academic definition has been challenged by Jason Hoffman, the Ericsson's cloud guru at that time. Over the past decade Web-based services like Google, Facebook, Amazon and Microsoft have stopped buying finished computers, storage devices and network components and instead developed their own systems in-house to create massive, low-cost data centers in the cloud to serve billions of users.

Jason Hoffman

He called them cloud factories, because CIO’s are thinking today (2015) how to begin to modernize the end-to-end IT infrastructure, so it can be a cloud factory.

Also there are some 2015 slides  with the main points I extracted.

The Internet of Things (IoT) for business

This is a  white paper from Aeris compiled by  Syed Zaeem Hosain, CTO

The explosion of IoT, envisaged to reach 20 billion by 2020, requires a new look at the cloud

“The cloud” has been coined to describe the systems that allow processing and storage of information and data in extremely large data centers for a fee.

This has transferred the need for entities and corporations to maintain their own physical hardware, data centers, and data networks, etc., to the cloud providers. This eliminates traditional operational burdens of physical site maintenance, electrical power management, environmental conditioning, and system redundancy. "

The fee charges can be high for large-scale applications and large numbers of device deployments. 

Fog Computing

Fog Computing Diagram (see reference)

In general, the process in a cloud is  a general is a “transmit everything and process in the cloud” implementation. 

However, if actions based on the data must be processed in real-time or near-real-time, it may be better to process or filter the data remotely—at the device, or elsewhere hierarchically in the data flow before it gets to the remote storage. This remote processing and filtering has been termed “fog computing” by Cisco.

The diagram is self-explanatory

Cisco White Paper Fog Computing and the Internet of Things: Extend the Cloud to Where the Things Area originated the modern  term of Fog Computing.

Edge Computing and Fog Computing

Cisco baptized the name Fog. but really in my opinion, this is the same as Edge Computing

In a recent interview Jason Hoffman, now CEO MobiledgeX answered the question "Where and what the edge computing is for you? 

"For some reason, there is almost an over obsession of where and what their edge computing is right." In terms of Fog definition, wherever there is a need to (1)  pre-process the data remotely, or  (2) some data are processing and used in real time  we have an edge.

Fog computing then manages the cloud and its' multiple edges, no matter where they are located.

Pros and cons for Fog Computing

This is from IBM

Telecom operated as Cloud and Edge Computing

This is for example the 13 countries operated by Deutsche Telekom today

The goal is to replace the legacy technologies in individual European countries with more centralized cloud systems catering to the entire region. In this way, Deutsche Telekom AG (NYSE: DT) hopes to collapse about 650 different service platforms into just 50. Instead of developing a new service 13 different times for 13 different countries, it would in future build and manage one service for all its operations.

They do not exist yet

The Edge and the Cloud universal platforms do not exist yet, but must be in produccion  Most IoT advertising cloud functionality offer a lock-in that works for their products only.

Jason Hoffman has the idea of an IoT universal platform. He reminds us how Amazon Web Services started with a developer payment API, then to an API called messaging cue system and then a three API called optic store and then you know a compute service that was really just batch compute on the object store. The documentation for the whole thing was a couple of pages.