Vlog #53: Help! The Internet killed my toaster!
ฝัง
- เผยแพร่เมื่อ 17 ก.ย. 2024
- www.superhouse.tv/superhouse-vlog-53-help-the-internet-killed-my-toaster/
A major outage at Amazon Web Services this morning left millions of people with broken home automation devices: door locks, heating, lights, ovens, even doorbells.
Don't let your home automation system be dependent on external services!
I always keep two principles (or rules) in mind when building my own systems.
Rule #1: No external dependencies.
Rule #2: Mechanical overrides.
SuperHouse:
- www.superhouse.tv
- superhousetv
- superhousetv/
Jonathan Oxer:
- jonoxer/
- jonoxer
Please support me on Patreon: www.patreon.com/superhouse
I follow alot of smart home related groups on facebook; wink, smartthings, and etc. Opening up facebook today and seeing how this outage has effected pretty much everyone across so many platforms was crazy. Made me happy I went the home assistant route. If it was not for vlogs and blogs like yours I would have been one of them.
Agree. I use openhab, everything locally served and even if my LAN is down, i Have a remote and when that is gone I have switches. Yes, I have some IFTTT things but all not essential
This would be my biggest hate about the IoT is depending on external servers/services, so when the company decides to shut down the servers or just disappears you are just left with very shiny paper weights.
This is something I thought about when figuring out my home automation system. No third party services and I can still operate things with a pushbutton even without network. Thanks for putting this out there for folks!
I have moved all my Sonoffs over to my own firmware communicating via MQTT to a local Raspi... They are also fitted with RF keyfobs, a local push button AND, turning them off the on again defaults to on..
Thanks to this channel for demonstrating how!!
If you're using Theo Arendst's firmware, yep, it's bloody great! I actually did change the firmware to default to on, but living in SE Asia, we have a lot of power outages, even for a split second. This meant in the middle of the night, lights would turn on for no reason and scare the shit out of me and the wife. I've since changed it back to default to off until it can poll openhab for the last state. Since everything else is on a UPS, it's usually only a few seconds before that happens.
Yes that has been a problem for me as well, as some of my items are 3D printers and PC's etc. At least I know when the power has gone out when I come home and everything is on!
I wrote a small client, based on the pubsubclient. Seems a lot faster and more reliable than the big distributions.
github.com - SonoffMqttLite
Most house locks purely mechanical. The NFC locks. Side door is independent and runs on batteries.
Main one has backup power for core functionality(Smart/web stuff runs localy) and still can use a key hidden away(Repinned the locks to make them. And we were going to do controlled lights, but worked out that moving around switches made more sense.
We didn't even bother automating the generator. Everything is on UPSs and if no one is around to hit the start switch on the gen, then clearly everything turning off isn't going to be a problem and they can still do it when they return :)
At least you have a generator! Not many people are that prepared.
After Brisbane's floods a few years back we thought it prudent. :)
And generators are great. But having it start at 3am for a 10 min outage upsets the neighbors. So manual is good choice. Friend's house we are working on 5-800Ah of battery that charges off solar or off peek power for a 'whole house' UPS that will also save money.
But Malcolm promised we'd have fast & stable NBN by now!
I'm really glad that I found your channel. You are creating systems much like the ones I create, only that you do it on a much bigger scale than me in my dorm room! I, too, believe that a local centralized system is the way to go when it comes to home automation, it just makes me feel a lot safer. I'm getting a lot of good ideas here and i'm really enjoying the content you produce, please do keep it up!
I think we've discussed the whole external dependencies thing before. With the nascent home automation hub I'm setting up at the moment, all the IOT devices will be on a separate wi-fi that has firewall rules prohibiting direct internet access. I'll add exceptions as desired. Generally I'm far more interested in the Intranet of Things than the Internet of Things. The latter feels inherently fragile.
That's a good term, the "Intranet of Things". Sensible!
I have soooo many people say "why didn't you just use WiFi?" when they discover that I've hard-wired all the essential control systems in my house. Yes, it took more time and money to do it this way. But a few meters of wire is a hell of a lot safer and more reliable than WiFi if you want something that will work for the next 20 or 30 years without opening you up to drive-by security hacks.
@@SuperHouseTV steve gibson (security now podcast) once used an even better term: internet of threats.
Great blog, those rules make a lot of sense.
You still depend on your devices being able to connect to your server within your intranet. However, since the server doesn't have a 24x7 sysad team, there will be times when it's down. Also, your network hardware might fail - it's not uncommon for routers to pack up from time to time, needing a good hard reset - and WiFi might be flaky.
I agree that external dependencies must be kept to a minimum. It's interesting to think about reducing even more external dependencies than you've mentioned in this video. For example, if a motion sensor has to turn on a light, it doesn't really need a server to orchestrate this.
What would this architecture look like? Would be it be worth the effort? It'd be interesting to know what you and others think.
I use quite a lot of z-wave in my home and the great thing about these devices is you can control them from a gateway/server but also set up direct connections between devices. So for instance a motion detector can be directly linked to a light switch but at the same time tell the gateway what it's doing. Only downside is that z-wave units are quite expensive.
Thanks. I'm sure z-wave is awesome, but I'm kinda ideologically opposed to it. It uses a licensed protocol, which brings up all sorts of concerns. I'm pretty sure it's these license fees that makes z-wave devices expensive. Ethernet, being unlicensed, patent-free and royalty-free, has a lot of available network hardware and free OSS software, and people have a lot of experience working with the stack. So it seems like it should be in the manufacturers' and users' interest to gravitate towards Ethernet eventually. It'll grant the best 'freedom', and keep prices low, which should encourage uptake.
However, Ethernet is not without it's problems. TCP is connection-oriented, which isn't the best way to transmit sensor data. UDP is packet oriented and can do things like broadcasts, but doesn't have quite the familiarity and network support that TCP enjoys. (UDP feels under-explored in this space though.) Also, WiFi and TCP aren't designed with low-power devices in mind, which imposes the rather severe limitation that all such devices need to be physically wired in to wall power, and it's not very practical to run them off batteries. So, for example, if I want to throw a motion sensor in a corner of a room, I can't practically run it off batteries if it uses WiFi. I have to supply wall power, which drastically increases installation effort, which means I'll simply install fewer of these devices.
To the point of this video, another major problem with Ethernet is it's star topology, which creates a single-point-of-failure - the network switch. Also, Ethernet encourages (though it doesn't require) a client-server architecture, which introduces another single-point-of-failure - the server.
We'll probably have to accept these limitations and work around them, so we can continue to work with a familiar stack. Or we'll have to create a parallel (hopefully compatible) standard for low powered devices which eliminates the single-point-of-failure - something which is more tailored towards secure ad-hoc peer-to-peer device messaging. But this is going to be time-consuming, needs industry buy-in, and has a high likelihood of never working out.
The NRF24 modules solve the low-power peer-to-peer problem well, and they're cheap to boot, but it's a corporate-managed protocol, doesn't have security built in, and is largely unfamiliar and incompatible with existing networks. Solving the incompatibility issue requires gateways and controllers, which again brings in single points of failure.
So, as it stands, it's a 'you can't have it all' situation. But how do we overcome this? Using non-free licensed protocols (like z-wave) isn't a long-term solution, but might be the best bet for now. I'm ideologically opposed to it, so I'm going to keep away from that kind of stack. At the moment, Ethernet is the best network stack for me. UDP might be the best protocol (though I use TCP because of MQTT in my setup). I have to put up with the network switch dying once in a while, preventing me from turning on my lights, I have to put up with the installation effort of requiring wall power, which is a very big deterrent to the idea of just throwing sensors around. And I have to make sure my internal server is up 24x7. Basically, my stack suffers from the same problems that Jonathan's stack suffers from, and I have the additional problems that WiFi brings in, since my devices are based on the ESP8266.
Works for now, but it can obviously be much better. I wonder what others are doing about these problems. Our stack right now doesn't feel as reliable as as the unconnected devices it replaces. We get awesome conveniences most of the time, but we also set ourselves up for frustration when things break. Must find solution.
(This comment turned out to be rambly and loooong. Apologies.)
i agree 100% with those rules. so much so that i wouldn't even think about using services like IFTTT or web-based voice assistants because they come with too many problems. if it can't happen locally, it ain't gonna happen!
exactly why all my IoT development goin slowly. Focusing on "no-external dependency"! and more than that! no wifi depending! Iot IS an additional Plug-in!
I use Home Assistant and DIY devices , my house is safe :D
Good!
Nope they sprayed nanites on every tech Research it
What are you talking about Fug Babylon? A link would be useful not just some rambling on conspiracy BS
That's good advice!
what for library do you use for the headers in autodesk eagle. i want to make my one board but i want to use your library, were can i find it
Don't think I have ever posted on your vids before. Love 'em! You got me interested in arduino with your older videos. My question is, have you done any videos on the mechanical overrides? If not, in the future?
I just love your videos. :) You give lots of knowledge to beginners like me. Thanks a lot and your service does mean a lot to us :)
I have Sonoff and WEMO running currently, how would I run these without external servers controlling them ?
I've thought for a long time now that those systems (alexa, etc.) that require you to connect to a remote server to be able to do stuff were sheer idiocy. I decided back then that I'd never have such a system, until and unless the "server" that ran things were local. Or, apparently in the case of things you're doing here, DIY. :-)
You would think all of this would be obvious... why would you make a toaster that doesn't work without internet?
I think I see the problem here: you're sensible! Not everyone is, unfortunately. Or they learn the hard way.
I guess there's a lot of people who like the idea of automation, but don't understand the principles behind it all. Marketing / sales people aren't going to point out issues like this.
That is exactly it. I am not interested in the logistics of how a parcel service gets a package to my door or the steps involved in the creation of a vaccine, I just expect they work. And that is how "non-tech" people feel in these circumstances. Most people are not interested in the details until it fails.
Hey! Very cool video!
Do you have any idea about when you will post the OpenHab & MQTT videos?
Soon! I have about 5 other videos partly filmed, so right now I'm finishing off one of them (which is nearly done) and then I'm going to do an overall system architecture video. It's important that I do an updated architecture video soon, because it helps put everything else (such as OpenHAB and MQTT) in context.
Okay! Thanks for the answer, I'm pretty excited for these videos! :)
Looking forward to this! OpenHAB reminds me a lot of Nagios in a way in terms of configuration. Once you understand how it all ties todether, you have one of those "Ah ha!" moments.
That's a great analogy. I spent years maintaining Nagios instances (and even started writing a Nagios book at one point, but got distracted) and there's a big period where all you can do is cargo-cult configuration snippets from other places, and everything seems very fragile. OpenHAB still seems like that to me, but I've spent enough time trying to make it do unusual things that I'm becoming more comfortable with it. Sometimes though I still break it, and can't figure out why, and have to just roll back the config and begin working forwards again.
A toaster or oven is out of control? I think this is how skynet starts! On a serious note though the whole idea of external dependencies is bad. It wasn't specifically mentioned but what happens when they decide to shutdown their servers permanently? Well you are SOL and have to buy all new devices/software. Microsoft pulled this crap on content purchased from XBox live. Even though the files are downloaded and stored to the HDD of my XBox 360, I cannot watch any of the videos I bought because they removed the DRM verification for those videos from their server! When I called they explained that the content was low demand and removed it therefore I could no longer watch them. probably $50 worth of content, essentially gone, because of outside dependencies. They offered me 1 month of XBox Live (I asked if they could do more but they claimed they couldn't), a whopping $3 value. Total BS!
I just say use a MQTT broker on a Raspberry Pi for example.
An Icom IC-7000 on the shelf behind you. You are or are going to become an amateur radio operator? :-)
Just curious to gather other people's thoughts on internet based voice recognition. I am a big fan of the Intranet of Things kind of thinking, and really don't want anything leaving the LAN, including my voice, since we now live in a country where that can be recorded without your knowledge and used against you (adjusts tin foil hat for slightly better reception). I don't want ASIO knowing that saying "three slices of pizza" opens my front door. But the problem is Microsoft voice recognition basically sucks, and needs to be yelled at in harsh tones before understanding you. Is there some kind of middle ground to be had?
I hope your going to give us something more interesting than these 'chats' and the gardening episodes you seem to giving us lately. Is your house 'done' and you have nothing new and interesting for us?. Maybe find someone with a house that is crying out to be a 'superhouse' and do it all again with us tagging along.
Unfortunately I don't have time to create any videos at the moment. I'm too busy watching all the educational content that you've been creating and making available for free on TH-cam. Oh, wait...