Articles like this remind me that people have a really difficult time wrapping their head around just how mind-bogglingly different the nature of the Internet of Things architectures actually are. Traditional technical solutions to most problems completely break down for this use case. Some specific points:
- A lot of IoT data is passively collected and reconstructed external to a device you own for applications that run on the device. Decentralizing the applications does not decentralize access to the data in practice. This makes perfect engineering sense: it saves a lot of battery and bandwidth on the device to not have the device involved in phoning home even if it is effectively "phoning home". (Few people grok how sophisticated this type of data reconstruction is.)
- IoT data coming off consumer devices is higher velocity and higher volume than anyone imagines unless you work with it. Billions of records per second continuously, petabytes per day, from single data sources. See above: the data your device effectively generates is not limited by the bandwidth of the device. Most applications of this data joins several of these data sources, often in real-time.
- The fundamental operation done on IoT data that makes it uniquely useful for consumer and other applications is the spatial join. If you think you are going to do that on a decentralized peer-to-peer network then you don't understand spatial joins. Doubly so considering the aforementioned bandwidth requirements.
Having physical control of a device will allow some control of where data goes but the architectural requirements of IoT will greatly constrain the extent to which this is possible in practice.
Those are lots of buzzwords but none of this happens or makes sense in practice as we see it today (for consumer, not industrial, IoT).
- decentralization is fine; no one is complaining that a thermometer doesn't store temperature log (though it could have some benefits) - it's fine to store it on a server, but people are complaining that they can't store that data on their own servers, they have to use insert-random-toilet-paper-startup's servers. Especially that those third parties do not provide any value (even though they could, in principle), they just farm you for data.
- collecting a lot of data doesn't mean learning a lot of useful things; quite often it's the opposite. What really matters is what you do with this data. If your Hardware-as-a-Service smart thermostat records temperature twice a second and the only thing you do with this data is logging it and looking at pretty graphs, it could collect readings twice an hour for more-less the same value given to you.
- spatial join - could you elaborate on that? From cursory Googling, it seems to be a term from GIS that mixes some data with information about its underlying database implementation...
- A lot of IoT data is passively collected and reconstructed external to a device you own for applications that run on the device. Decentralizing the applications does not decentralize access to the data in practice. This makes perfect engineering sense: it saves a lot of battery and bandwidth on the device to not have the device involved in phoning home even if it is effectively "phoning home". (Few people grok how sophisticated this type of data reconstruction is.)
- IoT data coming off consumer devices is higher velocity and higher volume than anyone imagines unless you work with it. Billions of records per second continuously, petabytes per day, from single data sources. See above: the data your device effectively generates is not limited by the bandwidth of the device. Most applications of this data joins several of these data sources, often in real-time.
- The fundamental operation done on IoT data that makes it uniquely useful for consumer and other applications is the spatial join. If you think you are going to do that on a decentralized peer-to-peer network then you don't understand spatial joins. Doubly so considering the aforementioned bandwidth requirements.
Having physical control of a device will allow some control of where data goes but the architectural requirements of IoT will greatly constrain the extent to which this is possible in practice.