As a consumer/prosumer/user of some IoT devices, the compute model I prefer is probably most closely aligned with what the author calls the "Fog" model. For devices at my home, I want:
* No data to leave my personal private network, except to third-party APIs for necessary information to fulfill my requests. If I have a smart agent device, for example, and I ask it to tell me the movies playing at my local theater, it is acceptable for an API request to go out to the cloud to fetch showtimes. But precisely that and only that.
* Computation runs on either the edge device itself or my local computing resources. Like many households, I have several "desktop class" PCs which each are more than capable of doing the kind of processing used by today's IoT use-cases. And like many enthusiast households, I have home servers as well, which are even more powerful. Despite several starts and stops in the space, I suspect eventually we will see home servers return as more people demand local processing and data autonomy.
* Data is stored either on the edge device or on my local computing resources (as above).
* Devices are available remotely via my home network's secure tunnel that is available anywhere on the Internet. (No specific requirements for the devices themselves.)
You're view seems too informed by your technical background. I can't imagine a rise in home servers to facilitate IoT as most customers will never have that level of interest in IoT products unless this is a type of product installed and maintained by a trained technician. Most applications don't justify that level of involvement.
This is the primary reason for doing compute and data storage away from the edge device. The edge device has to be dead simple that someone with hardly any technology background can place it and set it up. The customer just doesn't care enough to spend much time with any one IoT device. These class of devices are by and large adding only marginal value to the customer's life. Every detail they have to think about when working with the device becomes an impediment to their using the device at all.
> I can't imagine a rise in home servers to facilitate IoT as most customers will never have that level of interest in IoT products unless this is a type of product installed and maintained by a trained technician.
Why not? With the demise of Moore's law and the rise of zero-moving-part computers, as well as the existence of container technology, it will soon be possible to install a single compute node in a home and have it be dependably useful for over a decade. This node will be able to run all of the computing jobs necessary in a home, from the thermostat to the lighting to the TV to web browsing and games-playing (and even running web and email servers — or social networking services!). It would be easily serviced by a dedicated technician (as e.g. hot-water heaters and furnaces are today).
Not being forced to run cloud services saves device manufacturers money. Not requiring cloud services means that devices will continue to work usefully without Internet connexions. Not using cloud services means that user data is more private and secure.
I've heard that the conjecture is failing as we run into certain physical limits of making computer parts smaller, but I think what you've written indicates the opposite?
Sandy Lake. In the 6 years between the 2500K and the 7600K they've managed to increase performance by about 20% (total, not annually). Same # of cores, same price.
>The edge device has to be dead simple that someone with hardly any technology background can place it and set it up.
This is the main thing preventing adoption of home servers. They're a royal pain in the ass to set up and use right now.
Smartphones were kind of like this too before the first iPhone. There's nothing except that really preventing their widespread adoption. It's just that nobody's designed a really great one yet.
>most customers will never have that level of interest in IoT products
This was also the prevailing attitude about smartphones before the first iPhone.
> This is the main thing preventing adoption of home servers.
I don't agree. There'd be someone at least even tackling the problem, if there was actually a value proposition in it. The closest I can think of is Synology NASes, I suppose you roll the edge gateway into one too and that's your home server?
Thinking about what I use my home server for: media library, and backups. Both of those have been (mostly) supplanted by streaming and online services respectively, so I'm struggling to justify the home server even for the electricity to run it... but it is shortly to become an "IoT" gateway too (running Homebridge) since I agree with the top comment that it's best when as little data as possible leaves my local network.
There's a bit of a chicken/egg problem. If there were a thriving app ecosystem for home servers there would be a lot more demand for them and vice versa.
There are tons of apps which would simply be better on home servers because the advertising imperative is gone so once the ecosystem was kickstarted it would gain its own momentum. E.g. a locally run open source 'facebook' could have its feed tailored by the users and it could have privacy settings that are designed to put the users in the driving seat rather than features designed to foment addiction, oversharing and selling eyeballs.
For people who take a lot of digital photos and videos, local backup/data sync is simply always going to be faster than equivalent online services, so there's an 'in' to the chicken/egg problem.
I think the fact that this hasn't happened yet is because hardware companies tend to suck at software, software companies tend to suck at hardware and hardware companies that don't suck at design are pretty rare.
> There's a bit of a chicken/egg problem. If there were a thriving app ecosystem for home servers there would be a lot more demand for them and vice versa.
Absolutely. Where we disagree is that there's actually applications that would significantly benefit from being hosted on a home server as opposed to online servers or on-device.
> There are tons of apps which would simply be better on home servers because the advertising imperative is gone so once the ecosystem was kickstarted it would gain its own momentum. E.g. a locally run open source 'facebook' could have its feed tailored by the users and it could have privacy settings that are designed to put the users in the driving seat rather than features designed to foment addiction, oversharing and selling eyeballs.
There's been a number of attempts to make such a project, and they've all failed to gain any momentum. The difficult part of this problem is indeed "kickstarting the ecosystem", not any of that other stuff.
> For people who take a lot of digital photos and videos, local backup/data sync is simply always going to be faster than equivalent online services, so there's an 'in' to the chicken/egg problem.
Yes, but for 90% of people they do their photo work on a single computer, and plug in an external USB drive to do backups. That's quite simple to do, so the difficulty is demonstrating a real benefit to a server-based system that makes up for the extra cost. Apple's AirPort Extreme and Time Capsule products actually delivered this solution, but were niche enough sales that they couldn't be bothered continuing with them.
Your dream is not too far away from reality. Devices like Mozilla's Things Gateway (https://iot.mozilla.org/gateway/) represent some of the most ideal fog nodes as they are inexpensive and fairly powerful. Not too long ago there was even a research project called Paradrop (https://paradrop.org/) which made a fog node out of a consumer-grade router. My company is working on modeling software right now for Fog deployments and hope to use the knowledge we gain to build our own fog computing solution. Our first application will likely be either a fog-based filesystem (already done by researchers within Princeton's Edge Lab) or a Fog-based backup / NAS solution. It'll be a while before we get to this stage, but it's emerging from the mist so to speak.
>Ha. You'll get a proprietary black box that steals all of your data, eats all of your bandwidth, offers no public API and stores everything in the cloud, and you'll like it.
It took me far too long to work out that "fog" is a "cloud" that's right next to you, and hence "fog computing" is a local cloud. Haha. Of course, that made "mist", a thin fog (?) make sense too.
* No data to leave my personal private network, except to third-party APIs for necessary information to fulfill my requests. If I have a smart agent device, for example, and I ask it to tell me the movies playing at my local theater, it is acceptable for an API request to go out to the cloud to fetch showtimes. But precisely that and only that.
* Computation runs on either the edge device itself or my local computing resources. Like many households, I have several "desktop class" PCs which each are more than capable of doing the kind of processing used by today's IoT use-cases. And like many enthusiast households, I have home servers as well, which are even more powerful. Despite several starts and stops in the space, I suspect eventually we will see home servers return as more people demand local processing and data autonomy.
* Data is stored either on the edge device or on my local computing resources (as above).
* Devices are available remotely via my home network's secure tunnel that is available anywhere on the Internet. (No specific requirements for the devices themselves.)