OT: Amazon's new IoT service

I just got back from the Amazon Web Services re:invent conference, and one of their big announcements was a new IoT ( Internet of Things ) service.

The Rules and Shadows are what makes it a very interesting offering.

Let's use the HeaterMeter as an example: Rather than have a local web server running on the RasPi to make all of the HM config changes and viewing cook data, all that data would be streamed to the IoT service and you could then render the cook graph through a simple html file sitting on S3 storage.

If you wanted to make a config change, you would do it through the IoT service and the HeaterMeter would pick up the config changes, and almost instantly it would be set. With the Shadow functionality, you could make a configuration change on the IoT service and as soon as the HM booted up, it would pick up those changes and be done.

The Rules feature would similar to what the current alarm actions and scripts allow you to do, but it's all managed on the IoT service, not on the HM device.

Now, I'm not suggesting that the HeaterMeter move in this direction, I was just using it as a familiar device to use with the analogy. Personally, I think this type of stuff is pretty exciting and will be very disruptive, in a good way.

This Make: magazine article has a good write up on it as well. Here's part of the Keynote presentation where BMW describes how they're using IoT in the new 7 series.

This sort of thing is great for low-frequency data-- data that doesn't change very often. HeaterMeter pumps out like 10 data points per second and all the cloud storage stuff I've seen has been more about "store all data" or "store no history" instead of the RRD style. You're also limited in how you can interact with the data because you can only use tools that can interact with their data storage. There's also the question of cost. Who pays to store everyone's data?

The biggest question is: what value does it add? You still need a device that can talk to the internet, has some sort of UI to be able to configure the network settings. It isn't like the Pi can't handle the load of dealing with the data and doing notifications. The only advantage I can see is that you can access it remotely without cutting a hole in your firewall which is definitely a bonus.

Not that I don't see the benefits of a cloud storage for IoT stuff, things like power usage, thermostats, humidity, lights, garage doors, stuff like that. But for real-time constantly changing data I think it is better to go right to the device.

Call me skeptical...but if you weren't making Amazon/google/facebook money by giving them your data, they wouldn't be offering this 'service'.

Well you'd pay for the CPU time and storage in AWS I am sure.

After watching the video it looks like they do support moving high volume streams of data with somewhat low latency. I use MQTT for a lot of things here in my house along with Node-RED for event monitoring and notifications, a lot like they talk about with the lambda functions. When I needed to do some limited-time data collection (a few hours per experiment, 12 data points per minute), I used a esp8266 (nodemcu running lua) to do all the data acquisition and publish it into MQTT into Node-RED where I had some influxdb + grafana stuff to allow me to pull the graphs I needed for validation. Pretty much the same thing, except I had to roll it all myself so I can see the benefit of using a high-availability standard service to do that.

Yeah, it's not a free service like Facebook where the rule of "if you aren't paying for the product, you are the product" applies. It takes care of a lot of backend infrastructure that you don't need to worry about so that you can get on with developing a product and interacting with other devices. What's cool about Lambda is that it's reducing the need for you to even care/know about servers. You drop your code into Lambda, something triggers the function and it runs.

I work for Amazon and yes it's a pay model however you have to get to a certain volume. To pay for it you would have to do something like advertisements on the web gui to generate click revenue. You could also brand heatmeters with bbq vendors like weber or kingsford. You could also subsidize heat meters via branding. Yes there is a way to monetize it Amazon will help you store and provide infrastructure for a nominal monthly fee. Its up to the user to figure out how to make money and pay for the service. The idea is your IOT device is the vehicle to bring in money via branding, and web click revenue. Bryan you can always sell out your hobby to try and make money. I understand I fly model helicopters for a hobby, and I have been offered several sponsor ships. I turned them down because once your hobby becomes your job it looses all the fun of a hobby where you don't have to do it every day. The Amazon IOT service is one method of monetizing your IOT device, by providing infrastructure what is unique is that you pay for the volume of usage, and theoretically if your volume is increasing so is your ad revenues.

That's an interesting take on it. I was thinking it could be a less obtrusive model, such as a $5/yr subscription fee to "heatermeter plus", which gets you advanced logging or something similar. You could also just make it easy enough for a user to drop in their own AWS API keys, so that the end user is eating all of the costs.

Anyways, I was just using the HeaterMeter as an example use case. I'm in no way suggesting that things be moved in the direction, and since the RasPi has gobs of storage on it, even with a 4GB card, it's not really solving any problems, other than the fact that you could look at historical cook data beyond 24 hours, even if the device is off.

I'm doing something similar with ThingSpeak, where the RasPi is sending HM data every 30 seconds. I can go back and view the history any time.

Well another thing you could do is create a recipes site, that users could upload their recipies, heat meter settings, and various instructions events and alarms. Then other users could download from the cloud these recipes, and cook them. For example I could upload my ragin' rattlesnake ribs to the cloud and then you could download my recipe, and your heater meter would be all set to go. It could alarm you when it's time to check the ribs, show you images of what the ribs should look like, give you precise cooking instructions, and the pace at which I drank my beers at. You know the sky is the limit with the cloud, AOT is not the only show in town but they do make it easy, and you could link the heater meter with Alexa which is my next project after I'm done putting together my heater meter. She could read the recipe to you and upload the bill of materials to your phone to pickup at the store or have Amazon bring you the stuff from the pantry. Then you can query Alexa as to what temperature the grill is and she will respond back that your grill is at cooking temperature, and it's time to put the meat on her err the grill. Oh yes my mind is wandering but these are all possibilities.

Yeah the Alexa stuff is something that I'd definitely want to integrate as well. I've had an Echo in my "maybe" wishlist for a few months now as well.

I just don't see the benefit of the cloud storage though. It adds a level of complexity, the HeaterMeter then requires an internet connection to function (which you won't have at competitions), and you end up back with what you've got initially which is a webpage with a graph and statistics. I can see some super-advanced users maybe coming up with something to open their garage doors when the food is ready but is that really worth anyone's effort?

I'm not sure I see a use for "Recipes" either because 99% of BBQ is "cook low and slow at 225F for 10-15 hours". :-D

I see the benefits of cloud for devices which don't have sufficient wifi connectivity (intermittent for power reasons), or lack of storage, or the inability to generate rich web interfaces, or interface to other services easily. The original HeaterMeter is a huge candidate for cloud services because it didn't have a 1GHz CPU and a gig of RAM or even the ability to open two sockets at once. With things like the CHIP that bring all those things for less than the cost of a wifi adapter, I can't see the advantage of leveraging a system with more parts.

Who would want to control their garage doors with RasPi?

I'll just leave this here...

https://github.com/shmick/pi-scripts/blob/master/mordoor.sh

I agree there isn't a whole lot of value add that you can't get some where else. Plus like I said who would pay for it, and if it was click adds who wouldn't ***** about them. The Alexa integration has started I will start a thread when it's complete. There will have to be a process that will handle requests. It will have to live on the pi in openwrt.

As far as I see it, the only benefit is being able to set up someone with pairing their device with just a login, like all commercial IoT devices do. Log in with an account on both ends and you're done. This will be more necessary as IPv6 becomes mandatory. No need to learn how to do port forwarding on your router. And maybe passing some usage information back to your IoT device provider so they can improve their services.