Internet of Tomatoes: Building a Scalable Cloud Architecture

What is the work you're doing today?

I am the CTO of 30MHz, we're an agriculture platform company. Our users are farmers and we try to gather all of the data from different systems onto one platform to enable them to work with it, to visualize it, learn, advise, communicate about it. One important part of our platform is sensing the data. From the beginning, we have been working with sensors that we connect to the platform. We collect data every minute. So sensor data is a big part of it. And this is data which is coming continuously, every minute for every sensor. These can be sensors for temperature, air humidity, soil humidity, CO2, these kinds of things.

Do you also go into the fields and put down sensors?

We integrate sensors and produce them to connect to our platform. They are wireless. As part of our offer, we provide wireless sensors which you can choose and locate in your greenhouse or even outside. We don't build the sensors ourselves, but we select them and integrate them to our platform by making them wireless and optimizing them to run on battery with very low power usage.

When you say wireless, are we talking about 3G?

I am not planning to talk much about this, but we use a proprietary wireless protocol called XBee. It's pretty efficient and allows us to send data very frequently. From all the options that we have tried, this is the one which works best.

What kinds of technology do you use internally for gathering all this streaming data? Kafka?

We don't use Kafka at the moment. It's something we consider. We started building our platform seven years ago. Our architecture, running on AWS, is based into different components, each with a very specific function. Each of these components is resilient in itself, and they communicate with each other using queues, SQS.

What are the takeaways from your talk?

I'm going to give 10 architectural patterns to build software for scale. Be ready to scale two or three times the current load is one of the things that I would advise. Monitor everything because then you know which parts you have to re-architect. If you architect for scale, your platform can adapt to new functionality.