Representative End User Clients
Representative Automation Clients
Representative Software Clients
Cisco’s Fog Computing concept is under-appreciated. The thinking behind it came out of real-world IIoT applications where it became obvious that the solution needed to include local resources (these being computation, storage, and networking). The upstream oil & gas industry is one area where this need has been felt. Another is electric power distribution, which will become far more complex as more homes install distributed energy resources such as rooftop solar PV systems, or storage systems like the recently announced Powerwall.
The concept of fog computing is that of the cloud model, but residing “near the ground”, or nearer to the network edge. One reason for the fog model is that when you reach the network edge in some industries a data backhaul to the cloud becomes difficult and very expensive. Natural resources tend to be found in areas without cellular coverage – a persuasive reason to use the fog model.
But other aspects of cloud computing are net positives when applied to the network edge. Ones that comes to mind strongest for me for me are virtualization and systems management. A large data center provides services at a level that is very far removed from the “bare metal” of the physical servers and networks that it houses and operates. This higher level of abstraction enables a data center to operate and be managed during maintenance periods where parts of the center are being replaced or upgraded. THAT behavior is one that the typical industrial edge device does not have, but could certainly use.
In far too much of the industrial automation world there is only very limited abstraction from the physical systems and servers that perform measurement and automation tasks. Embedded systems have studiously avoided unnecessary abstraction layers because these systems place a very high value on real-time performance metrics. Automation systems built around older single core processors clearly aren’t capable of supporting such advanced services without compromising their real-time performance. But this simply won’t be true in the future, as multi-core processors and large amounts of storage capacity respond to the relentless cost erosion of Moore’s Law.
As data centers scaled up to vast sizes, they had to fully automate their system management processes. Every process required automation because any manual task could not be repeated on a center with thousands of servers. That kind of experience and thinking will prove useful in the future for the IIoT. Managing 1,000 or 10,000 remote sites with today’s embedded system technology is challenging and costly. But add the virtualization and management capabilities of today’s data centers, and this challenge would be dramatically reduced.
The concept of fog computing is that of the cloud model, but residing “near the ground”, or nearer to the network edge. One reason for the fog model is that when you reach the network edge in some industries a data backhaul to the cloud becomes difficult and very expensive. Natural resources tend to be found in areas without cellular coverage – a persuasive reason to use the fog model.
But other aspects of cloud computing are net positives when applied to the network edge. Ones that comes to mind strongest for me for me are virtualization and systems management. A large data center provides services at a level that is very far removed from the “bare metal” of the physical servers and networks that it houses and operates. This higher level of abstraction enables a data center to operate and be managed during maintenance periods where parts of the center are being replaced or upgraded. THAT behavior is one that the typical industrial edge device does not have, but could certainly use.
In far too much of the industrial automation world there is only very limited abstraction from the physical systems and servers that perform measurement and automation tasks. Embedded systems have studiously avoided unnecessary abstraction layers because these systems place a very high value on real-time performance metrics. Automation systems built around older single core processors clearly aren’t capable of supporting such advanced services without compromising their real-time performance. But this simply won’t be true in the future, as multi-core processors and large amounts of storage capacity respond to the relentless cost erosion of Moore’s Law.
As data centers scaled up to vast sizes, they had to fully automate their system management processes. Every process required automation because any manual task could not be repeated on a center with thousands of servers. That kind of experience and thinking will prove useful in the future for the IIoT. Managing 1,000 or 10,000 remote sites with today’s embedded system technology is challenging and costly. But add the virtualization and management capabilities of today’s data centers, and this challenge would be dramatically reduced.