KB3 - stock.adobe.com

Reaping green dividends with data management

Getting data management right has been pivotal for Singapore IoT startup SensorFlow to optimise energy consumption and reduce carbon emissions for its hotelier clients

Air conditioning and heating systems can account for nearly half of a building’s energy usage, so optimising the use of those systems can make a difference to carbon emissions and utility bills.

Using traditional energy management systems to manage heating, ventilation, and air conditioning (HVAC) systems can be expensive, leaving building owners in a bind if they wish to do more to combat climate change.

A smart energy management system with IoT (internet of things) and AI-driven automation capabilities, along with occupancy sensors and smart thermostats that building owners can retrofit in rooms, promises to help building owners save up to 30% HVAC energy costs.

The system, a full hardware and software stack, was developed by Singapore startup SensorFlow, which counts the likes of well-known hoteliers The Ascott, Accor, Marriott and Hyatt as clients.

SensorFlow’s devices are connected to network gateways in customer premises using the Lora wireless modulation technique to transmit data to its infrastructure on the cloud.

“We’re a full-stack IoT company, from gathering raw data at the hardware level, to processing and generating insights from that data and facilitating actions based on the insights,” said Max Pagel, co-founder and chief technology officer of SensorFlow.

Getting a handle on data management and platforms is therefore critical, but not everyone gets it right the first time. Pagel said when SensorFlow first started, it jumped on the serverless computing and NoSQL bandwagons with offerings from Amazon Web Services (AWS).

Lured by half a million dollars’ worth of AWS credits, the company found itself going down a slippery slope, with its developers starting to build things without being mindful of cloud consumption costs that skyrocketed to as high as $45,000 a month.

That led to a dramatic move to rebuild its software architecture not only to avoid cloud bill shocks, but also to make it easier to access data which proved to be difficult to do with NoSQL databases.

“One of the big challenges we faced with NoSQL and serverless was the accessibility of data,” Pagel said. “You have everything in a giant NoSQL infrastructure, but you can’t query it efficiently unless you write code.

“Even then, it doesn’t really have a schema that people understand so developers will just dump stuff in. It leads you down the rabbit hole where you have to pull all that data out and transform it again to put it into a data warehouse,” Pagel said.

Pagel said those issues could have been averted had the company started with SQL databases, which could be connected to business intelligence applications to provide immediate access to all the data and insights.

SensorFlow eventually went with PostgreSQL, which it hosted on its own until it encountered problems with database migration that caused major service disruptions. That prompted it to use a PostgreSQL service from Aiven, a managed service provider of open source databases and data platforms.

To handle the massive amount of time-series data that it collects from IoT sensors and devices, SensorFlow uses the TimeScale plugin, which Pagel claimed would not work with AWS’s SQL database services.

“AWS just doesn’t allow you to use it, because they want to sell their own Timestream database, which performs abysmally compared to most other time series solutions,” he said.

SensorFlow has another database for analytics workloads that generate saving reports and other insights for its clients based on the IoT data it collects. But keeping its analytical database and production PostgreSQL database in sync is hard to do because of the sheer volume of IoT data.

Pagel said his team now runs a daily batch job to synchronise the data, adding: “But there’s obviously a point in future where that’s not going to cut it anymore”, adding that he has not come across a “perfectly elegant solution that’s cost effective as well.”

On how hotels are leveraging insights from SensorFlow, Pagel said by correlating energy consumption with cooling load hours, for example, they are able to automate the cooling of guest rooms and optimise energy usage when guests are out in the day.

For now, SensorFlow is mainly eyeing business from the hospitality industry, but plans are afoot to expand its reach into other sectors such as warehousing. “Our solution shines when you have irregular usage of rooms and for warehouses where usage is quite unpredictable, there can be a huge amount of energy wastage if they are not managed properly,” Pagel said.

Read more about data management in APAC

Read more on Internet of Things (IoT)

Data Center
Data Management