The Internet of Things (IoT) and the cloud are impossible to separate. Only about a third of the data collected by the growing army of sensors is analysed at source, but as the IoT grows, that is going to need to change.

“Sensors and the data they create are the next big thing,” says Scott Gnau, CTO of data management platform firm Hortonworks. “Compute can now happen at the sensor level, and not solely in a centralised data centre or cloud footprint, and it's sometimes referred to as fog computing.”

Scott Gnau, CTO of Hortonworks

What is fog computing?

Fog computing and edge computing are basically the same thing. “Fog computing is the process of computing data that is not in the cloud or at the branch, but at the extreme edge of the network, enabling analytics of that data close to its source,” says Sarah Eccleston, Director of Architecture Sales, Cisco UKI. It’s Cisco itself that has pushed the term ‘fog computing’, but others understand it slightly differently.

“We see ‘fog’ computing as synonymous with ‘edge’ computing, and the terms are often used interchangeably,” says Neil Postlethwaite, Director, IBM Watson IoT Platform and Device Ecosystem. IBM’s Watson IoT Platform helps IoT customers transition work from the cloud to the edge. “It’s reflective of operations performed in the network layer i.e. ‘at the edge’, with provision of compute, data and storage capabilities,” he says.

IBM’s Watson IoT Platform helps IoT data move from cloud to edge

Combining edge, fog and cloud

If the IoT is primarily about cost-saving for industry, then fog computing is part of that – it's about conducting analytics in the most efficient way possible.

“Businesses must work out the best place in the system to perform the computation that is needed to deliver the required outcome and value,” says Graeme Wright, CTO for Manufacturing, Utilities, and Services at Fujitsu UK.

This is about combining edge, fog and cloud computing. “Edge computing may be used to control the device that is being monitored by a sensor, and only send data back when something changes,” says Wright, who offers an IoT example. “This could then be complemented by fog computing, to alert other sensors or devices of the status change, and take appropriate action.”

The cloud can then be used to perform analytics on the system as a whole, alerting staff about maintenance issues. “This setup can not only provide real-time analysis of the data, but also lower data storage, and more importantly improve efficiency,” observes Wright.

Fog computing allows decisions to be made closer to IoT devices

Cloud to edge in real-time

Moving some of the data storage and processing to the edge of a network means using edge gateways. “Ideally, you should be able to seamlessly move computing from cloud to edge as and when workload dictates,” says Postlethwaite, who thinks that having intelligence at the edge means decisions can be made closer to the actual IoT sensors and devices.

One example is image processing: visual analytics close to a manufacturing line to check quality – so ‘at the edge’ – saves sending large amounts of data to the cloud for processing. 

Top Image Credit: Siemens

The benefits of fog computing

“Using edge or fog computing helps reduce both your latency and your costs, and insulates you against any network outage or glitches,” says Postlethwaite. “Edge is about latency, autonomy, data security – such as near-source data anonymisation – local analytics and rules.”

Other advantages of processing at the edge include saving money on data transfers (since most IoT cloud platforms charge by the amount of data they accept). “It results in a reduction of transportation costs that are often associated with the movement of data to a physically-distant cloud-compute resource,” says Eccleston. This is about slashing data's travel expenses. 

However, aside from the technical advantages of fog computing, it also taps into the existing desire to process data away from the cloud. “There is a reticence amongst industrial and manufacturing clients to expose their facilities to external communications for security reasons or data privacy reasons,” says Postlethwaite.

Fog or edge analytics cuts data’s travel expenses

Fog gateways and the ‘jagged edge’

Fog gateways are an excellent way of dealing with the different systems and equipment – and often outdated technology – that exists in industry and manufacturing.

“Edge gateways can talk to a myriad of devices and translate between their proprietary protocols and those understood by cloud environments,” says Postlethwaite.

That’s helpful in IoT projects, but there are other problems with handling data away from corporate firewalls. “When it comes to fog computing generated by IoT, the perimeter of data processing is outside the traditional corporate firewall, and can be very jagged,” says Gnau, who thinks that transmitted IoT data needs to be secured, but it also needs to be received correctly, quickly, and in a way it can be audited.

“Keeping track of one-way streams of data is a tall order, but IoT requires bi-directional and point-to-point communication, and optimisation of bandwidth, all within a jagged edge construct,” says Gnau.

Cloud robotics will benefit from 5G

How will 5G affect the fog?

By processing data at source, one of the major advantages of fog computing is a reduction in latency. That’s something that the advent of 5G networks will theoretically remove altogether. However, that doesn’t mean that fog or edge computing will suffer because 5G will mean the IoT can grow massively.

“5G will enable a greater connectivity of ‘things’ to the internet, which in turn creates more applications and data that requires compute and analytics,” says Eccleston. She thinks that businesses are under pressure to quickly scale and extract the true value from all the data collected by IoT devices, and that fog computing is a valuable tool. “Fog computing meets the real-time needs of many IoT-based applications and ensures the capabilities and benefits made possible by 5G are realised,” she says. 

However, 5G will come at a high financial cost, and aside from latency-critical uses such as cloud robotics, will not bring discernible advantages to most IoT projects. “For use-cases with large bandwidth requirements, 5G will make a key difference,” says Postlethwaite. “But for many IoT scenarios, existing communication technology combined with a good edge strategy will go a long way.”

  • 5G is coming soon – but what’s in it for businesses?