freshidea - Fotolia

Manage Learn to apply best practices and optimize your operations.

How Cisco's fog computing approach may fuel IoT

The Internet of Things is led by Amazon, Azure and HTTP-tuned networks. But where does fog computing fit into the picture? Learn how Cisco's fog computing approach may fuel IoT.

Recently, I suggested that the future of the Internet of Things is more or less all about Amazon, Azure and HTTP-tuned networks. The abridged version is that implementing complex network services to handle billions of IoT updates per day is too much for enterprises; therefore, IoT will be the domain of cloud. While I'm still confident that the big win will be for Amazon Web Services (AWS), Azure and device manufactures, there is a case where Cisco enables specialized conduits for IoT: fog computing.

Cisco fog computing: Von Cisco Bewirtet

"Wait a second," you might be thinking. "You just got back from Cisco Live Berlin. They dazzled you with a pile of sexy IOx-compatible, hardened and heat-finned aggregation services routers, and now you're flip-flopping?"

While it's true I did poke around the IOx demos, being in Europe provided an opportunity to spend a lot of time talking to SolarWinds customers, especially municipal government and utility admins about what they plan to do with IoT. From those conversations, it's clear that certain enterprises could benefit by not hammering LAN, edge and WAN in a rush to let every verbose IoT chirp make it to the cloud.

Network fog and smog

The idea behind fog computing is that there's no reason to pump what will become trillions of updates from billions of devices to remote servers.

The idea behind fog computing (I prefer "fogging," but it's just a wee too close to condensation for me) is that there's no reason to pump what will become trillions of updates from billions of IoT devices to remote servers. Instead, fog computing moves the bulk of processing, or at least aggregation, to edge networking devices. The upside is less expensive WAN and potentially more responsive systems. For example, a small number of temperature, humidity, occupancy and sunshine sensors working together to optimize local HVAC.

The downside is significantly more complex deployment and maintenance. Even ignoring remote-deployed OS and application code, just the management and monitoring of additional distributed edge networking gear is overhead avoided by the IoT cloud express route. For most enterprises, this may be enough reason to instead double down on Internet bandwidth and event processing as a service in the cloud. But for customers with limited connectivity options or huge IoT event streams, the fog computing approach offers advantages.

We landlubber IT engineers often forget the challenge of tiny pipes, but if you're a cruise ship line with 20,000 devices per vessel communicating to the network operations center over satellite links, IoT is a serious challenge. Same goes for oil and gas, rural production facilities, and the military. To share device economies of scale with their fat-pipe brethren, these operators will need a way to field commodity IoT infrastructure, but keep the majority of data collection and business logic close to endpoints. For them, Cisco's fog computing pitch of a unified infrastructure to distribute collection and logic, summarizing and routing ops data, and bandwidth offloading will be attractive. This should be especially true for factory floor and other operations where a larger upfront investment in custom control and aggregation code development pays significant benefits over large production runs.

It's all about frequency

Fog computing and edge networking certainly have a place, and their ultimate breadth of adoption comes down largely to frequency of update and criticality of local actuation. If IoT update frequency is high, and either bandwidth or collection costs are high relative to the importance of data, then fog computing is an attractive option. If, on the other hand, remote operations are disconnect-tolerant, or the frequency of update is low, then sending all data to the cloud and greatly simplifying the network may be the best option.

My growing collection of tame RaspberryPis is very happy to continue building out my cloud-based sprinkler controller project powered by an AWS IoT back end. But if I had a retired FANUC robot clearing the dishes at home, a little fog compute at the edge might be a better choice.

Next Steps

Is fog computing a paradigm shift?

SDN in cloud computing

Quality, IoT devices help to push edge computing

This was last published in March 2016

Dig Deeper on Edge computing