Building a cellular tracker: Part4 – Which cloud-service to chose for storing the data

As described earlier, i want to create a cellular tracker – below i am going to describe where to store the data…

Tracking sensor-data only makes sense if the data is being stored in a central place for further action. Usually you would want to analyze and visualize the data, drive further actions from the data and probably even perform some machine learning tasks to get deeper insights.

Of course this can be done locally with your own server and file-system which most probably would be something hadoop-based in these days with a stack of spark-services on top. If you really want to build something significant and do very specific stuff i would recommend exactly that – but: it will be a huge effort to build a backend-service like this and operate it on a daily basis.

Fortunately a lot of cloud-based IoT services have come to life recently. There are so many that it is even hard to create a list of them because new services are going live every weak – just recently sparkfun and adafruit launched their own tracking-services for instance…

So i cannot offer a full-blown evaluation of all services available – i was looking for something stable, with easy ways of accessing the data both for storing and retrieving, with good tutorials and code-examples and a nice interfaces.

I tried to use xively, but after weeks i still seem to be on a waiting list to get an account. I checked thingspeak and was impressed how fast and easy i was able to generate the first data-points (in fact it took not more than 2 minutes from account registration to tracking first data). Thingspeak is interesting in many dimensions – the cloud-platform is well documented and has some nice visualization features + a very straight and easy to understand/implement tracking api. But it has a lot of restrictions if you look closer. There is a rate limit of one update per channel per 15seconds for example, every channel can have only 8 fields for storing data etc. – this all makes sense as thingspeak.com is a free service. But it can quickly be too restrictive if you plan to do something serious with the data and i don’t want to imagine having deployed a lot of sensors in the wild and then figuring out that the storage platform is reaching it’s limits…

Interestingly thingspeak is not only a cloud-service but also an Open-Source Software Stack that easily can be hosted on your own (cloud-based) servers. In this case the limitations obviously can be removed, but i read that other things like visualization options etc. are harder to implement with a local installation…

I decided not to use thingspeak in my project, but i think it deserves a close look if you are considering a project or even a business in the IoT field – especially because you could prototype with the cloud-service and then establish your own backend based on the Open-Source stack – that can be a very promising approach.

Instead i fell in love with ubidots, another cloud-based IoT Service that is a little bit more business focussed than thingspeak but in the same time less arrogant than xively…

In fact ubidots is a very nice and reliable platform with a well documented API, nice tutorials and really amazing features for inspecting and visualising the data. There are different plans starting with a free-service that gives enough room and features to really start something + 3 levels of paid services starting with 5$/month. For 100$ per month businesses can get a customizable service under their own domain-name + advanced features. Their service is competent and available and the dashboard they offer is probably best in class. Just recently they for instance offered a GPS-trace app (GPS location was already available) that gives you a map with a route based on the latest GPS-coordinates you tracked. Pretty amazing.

The downside: ubidots is closed source and you are locked into their service with your data (as far as i can see only csv-export is offered on a per variable base – not really something you would consider to move a whole account to another service…).