• 08/19/2014
    8:00 AM
  • Rating: 
    0 votes
    Vote up!
    Vote down!

IoT: Out Of The Cloud & Into The Fog

Cloud computing architectures won't be able to handle the communication demands of the Internet of Things. The future is in fog computing.

By now, most IT organizations have embraced the concept of cloud computing and are using it in some capacity. But if grand predictions regarding the Internet of Things (IoT) turn out to be true, even the most advanced, distributed cloud architectures aren't going to be able to handle the IoT's data and communications needs.

That's where the idea of "fog computing" comes into play. It's a term coined by Cisco, but most major IT vendors are developing architectures that describe how the IoT will work by bringing the cloud closer to the end user -- similar to how fog is nothing more than a cloud that surrounds us on the ground.

The problem that IoT forward thinkers see with the current cloud architecture is that it's heavily reliant on distributed processing and available bandwidth from the edge device to the backend server. Most data in a cloud environment is sent to the cloud to be processed, leaving our edge devices as dumb portals into the cloud.

Though this architecture works well today, it falls apart when we're talking about adding billions of devices and microdata transactions that are incredibly latency sensitive. Instead of forcing all processing to backend clouds and forcing all IoT device intercommunication through a cloud intermediary, fog computing proposes that devices have the opportunity to talk directly to one another when possible and handle much of their own computational tasks.

This evolutionary shift from the cloud to the fog makes complete sense to me. The original cloud boom began when mobile devices like smartphones and tablets were becoming all the rage. Back then, these devices were weak on computing power, and mobile networks were both slow and unreliable. Therefore, it made complete sense to use a hub-and-spoke cloud architecture for all communications.

But now that most of us are blanketed in reliable 4G technologies, and mobile devices now rival many PCs in terms of computational power, it makes sense to move from a hub-and-spoke model to one that resembles a mesh or edge computing data architecture. Doing so eliminates bandwidth bottlenecks and latency issues that will undoubtedly cripple the IoT movement in the long run.

So if you thought that cloud computing was the pinnacle of infrastructure designs for the foreseeable future, think again. If we're talking billions of devices and instant communication, current cloud models won't be able to handle the load. Fortunately, advances in mobile processing power and wireless bandwidth have allowed many to design a far more capable architecture that brings us out of the clouds and into the fog.



Fog computing enables the flexibility that some types of applications, devices and services require to either/both, be economical or possible. Out of the 50 billion devices that are expected to be connected to the internet by 2020, some just require the internet (connectivity), others require the Cloud and some absolutely require the Fog.

Re: Flexibility

On one level this makes perfect sense, and I have seen Cisco demo the technology and it seems reasonable. But when it comes to the details of securing and managing the data that's actually produced in the fog, I am not so confident we can accomplish that yet. After all, we haven't really mastered cloud computing yet, even though the vendors act like that's a done deal. Fog sounds like a good idea, but I suspect it will be a lot more complex to actually make it work.

Re: Flexibility

That's a great point Susan. With traditional cloud architectures, we have the classic "choke points" in and out where we can harden the infrastructure with typical firewalls, IPS and other protections. But with fog computing's mesh design, devices could potentially communicate directly without traditional controls in place.

Re: Flexibility

Whatever end device wounds up being used and however the infrastructure is maintained, what many professionals fail to acknowledge is that the real weakness in many systems begins and ends with the router.  Routers have become notorious for being easy to hack; professionals estimate that about 1 in 5 of every router has a significant backdoor present.  I recall seeing in news on NSA operations/the Snowden papers that the NSA specifically targets routers for these reasons.

Re: Flexibility

I was not familiar with this concept. It looks interesting but, as Susan points out, the security has to be sound first.

There are so many connecting standards for IoT now that is difficult to see where the market is going.

Re: Flexibility

@Susan, I agree, selecting the best (Cloud, Fog and end-device) location to house computational power is extremely complex. Cloud offers centralization and economies of scale -- making huge amounts of computational power cheap for the consumer. The Cloud approach also makes the end-device relatively simple and easy to produce.

Re: Flexibility

Brian, yes cost is definitely another big factor. Patrick Hubbard brought that up in his recent blog, Internet of Things = Internet of Gateways. Consumers are accustomed to paying low prices for goods,so they won't tolerate a big price jump to make their coffee pots and toothbrushes IoT frieindly. Instead, Patrick says, we'll need a series of gateways to consolidate and handle traffic. 

Re: Flexibility

@Susan, thank you for the link. You make a great point -- the traditional subscription model of the service provider will not work with the IoT.

$30 a month might seem like a small price to pay for a device connected to the internet that provides value through analytics. However, continue to add devices to the automated home or office, and $30 can easily end up as a $6,000 per year cost.

This is one reason why I feel that things like software defined mesh networks that enable distributed gateways is a major step forward.

Re: Flexibility

Brian, I'm kind of confused about exactly who will own and manage the mesh. What do you think about that?

Re: Flexibility

@Susan, great point, I am also confused on whether the owner will also be the manager of the mesh. I feel it is important to make this distinction because the owner will be the firm that makes the initial investment of capital and would require a major share of the profits to cover their risks.

I can't think of any traditional service providers taking the lead, but a hypothetical situation involving a newer service provider and a utility company does come to mind. For example, an electric company would deploy a gateway on their transformer, the smart electric meters that are in range of the gateway would communicate directly to the gateway and the electric meters out of range would communicate through electric meters (using it as gateways) that were in range -- forming the mesh. Next, the electric company would sell communication to the Water Company and gas company and so forth, for their smart meters. The newer service provider would be providing a service to the electric company that owns the communication infrastructure.


Re: Flexibility

Brian, yes, thanks for explaining that. The scenario you describe is similar to what I was thinking, but I couldn't quite articulate it. Seeing it all written out, I don't feel so bad for being confused -- it's pretty complicated!

NCs ahead of their time.

It's interesting to me that hard drive-less Network Computers -- "NCs" (as opposed to PCs) -- never caught on in the late '90s.  Some evangelists thought they would -- that we would eschew local storage in exchange for cost savings --but they were wrong.  The NC was an idea ahead of its time, because the demand wasn't there yet.

Now, they're making a comeback in a way because of this mobile revolution -- and the new way of making money will be finding ways to make that more efficient with "fog" solutions and even local, hybrid "cloud-in-a-box" solutions.