Not mine, thank God, but interesting to see one in test mode. In spite of all the concerns about Zigbee and RF interference its impressive that the architecture easily supports this capacity (and this outfit routinely deploys networks with 500-800 devices per coordinator). Makes Z-Wave's 232 max devices and 4 hop routing limit look strictly consumer grade in comparison.
What the hall where he is going to apply all that?
Though the coordinator supports it they are all in close proximity. How will it be in a mesh is the big question
Large Zigbee networks are not that uncommon in industrial applications; but I think the key to their success in that arena is that they don't rely on the ZHA profile (the dumbed down flavor of Zigbee that most retail consumer devices now use). The industrial versions can support features like end device configuration, frequency agility (so devices can change channels dynamically in response to RF conditions), etc.
The nice aspect of Z-Wave is they seem to have nailed down how the features are implemented as it has evolved ; in contrast the Zigbee specification is kind of wishy-washy ("a device can do x.. " vs. " a device must do x... "). The home automation profile was an attempt to move in the direction of greater interoperability but it doesn't specify all the capabilities that an industrial strength application would require. Hence some combinations of consumer Zigbee routers and end devices seem to perform better or worse than others in terms of self healing and fault tolerance in different environments.
I also think a significant factor is how chatty those 600 devices are. How often are they reporting temperature, power, battery, binary status, etc. That's a lot of potential traffic if the devices are typical talkative HA devices.
I agree, it may be fun to do what they did there and test the theoretical limits of the network but this is not really a good test, it is way too controlled with all devices sitting next to each other. In real life with all the environmental variables, noise, interference, much higher variance in manufacturers, firmware versions, protocol versions, device types, old devices misbehaving, etc. such a deployment would be a huge PITA to run and maintain...
Not likely that is a home automation profile they are running and I'm pretty sure that networks with this many devices are not theoretical in industrial environments in 2019. The outfit that produced that video has a blog Sun May Sky: 660 Zigbee devices in the same Zigbee network! One of the comments from Cimconlighting.com mentions that they routinely deploy Zigbee networks with 500-800 nodes per coordinator. They specialize in outdoor lighting management in urban environments so its likely that they are using some of the non-2.4Ghz frequency bands that the Zigbee Pro spec provides which could provide a transmission range up to 1km.
My first reaction when I saw that setup was that it's not representative of any real world deployment in terms of RF environment and that is certainly true. But I don't think that is what it is intended to test. Its more likely a setup like that is to validate the protocol stack. With hundreds of sensor nodes, something as basic as routing table communication for the 'many to one' scenario becomes a major chokepoint. What works for a network with a few dozen devices may die when there are hundreds of devices issuing route request broadcasts; setups like these are useful to validate that the coordinator is capable of managing its routing table. Back in 2014, when a 400 node network was considered a barrier, TI had Zigbee Pro (and its routing enhancements) to address that issue: http://www.ti.com/lit/an/swra427c/swra427c.pdf
Still impressive that an architecture spec'd out in 2004 (Pro in 2007) has has this extensibility. Apparently Zigbee Pro and non-Zigbee devices can be intermixed, but because of routing differences a Pro device can only function as a non-router end device in a non-Pro network (and vice-versa).