BeeHero is using data driven algorithms to provide pollination services to growers and beekeepers. We face numerous challenges in building, maintaining and scaling our architecture to support these services and to grow our company - in the upcoming series of posts we will focus on the challenges of scaling our services, and the road we journeyed in doing so.
The BeeHero innovation is simple at its core: we put sensors in bee hives, which track the hive’s environment and report to our cloud platform. Leveraging biological expertise and data science, we are able to produce insights from the raw measurements - which enables BeeHero to report the hive health, strength and growth, and even predict its progress throughout the pollination season. We will focus here on the technical aspects of our solution, but I encourage you to read more about how it all works together, and what our beekeepers and growers are saying about pollinating with BeeHero.
Our high level architecture is pretty straightforward and common for IoT-based applications - as illustrated below:
Data is gathered by our BeeHero sensors and uploaded to our cloud platform periodically. The cloud platform is then responsible for receiving, storing, processing and contextualizing the sensor data. Finally, Web and mobile applications access the platform API in order to visualize the data and the derived insights to our customers - crop growers and beekeepers.
When we’re discussing scale at BeeHero, it is the scale of the IoT devices that challenge us. At the time I’m writing this, BeeHero deployed over 300K monitored hives, which are typically configured to collect samples every few minutes, and upload these samples every few hours. We collect more than 15 million samples a day, reaching as much as 30 million samples during peak season time - indeed, seasonality is one of our scale challenges. BeeHero sensors communicate with our servers over mobile networks, which means we should account for communication breakdown and unavailability at times. This same factor of uncertainty further impacts our cloud processing of the data - we must take into account gaps in sensor data, inconsistent behavior and inaccuracies. Eventually the sensor data scale affects how we store the data and make it accessible for our algorithms and user-facing applications.
In our next post we will deep dive into the details of the BeeHero platform in Feb 2022 and the early challenges of scaling our IoT devices.
BeeHero's journey to scale our solution and meet customer expectations. Part one: an introduction to BeeHero IoT and cloud platform and the challenges we faced early on
