The Challenge

FINN.no had run its IT infrastructure from its own data center for years. However, the company wanted to move its IT to the cloud and concentrate on developing its strategy and approach rather than managing infrastructure. This involved moving more than 800 applications and 145 database instances into the cloud.

This migration included moving a critical cluster that supports the company’s personalization engine for adverts that runs on Apache Cassandra. This application uses data such as a user’s preferences and advertising history to shape the experience that customers have while they visit the site.

“This personalization engine is essential to us as a company. It helps us to up sell products that might be interesting to customers, it can deliver sponsored products to target customer groups, and it supports other companies in the group with their personalization services too,” explained Espen Amble Kolstad, Senior Developer, FINN.no. “It supports around twenty percent of our traffic as a site, which represents millions of interactions, so it’s an important part of the business.”

The FINN.no IT team had decided to use Google Cloud Platform for its migration and had to consider how to move the existing Apache Cassandra cluster into the cloud. According to Benjamin Weina Lager, Technical Domain Expert - Data Intelligence at FINN.no, this move was something that came at a good time for the Data Intelligence team: “Our team is made up of three engineers and three to four data scientists, and we have responsibility for the data modeling and analytics that powers the personalization engine. We have some Cassandra skills internally, but this migration was a good opportunity to re-evaluate.”

As a database, Apache Cassandra can run in multiple cloud services and across different cloud providers simultaneously. However, the FINN.no team wanted help with the future management of Cassandra.

The Solution

The FINN.no team contacted members of the wider Cassandra community to discuss the challenges and they were recommended to look at DataStax Astra DB. Astra DB is a Database-as-a-Service built on Apache Cassandra that allows developers to quickly deploy massive applications in the cloud. This includes native support for Google Cloud Platform, so FINN.no could include the Cassandra cluster migration in its wider cloud project.

“We ran on the open-source version of Cassandra in the past, but we wanted to get support for running our instances in the future,” said Lager. “DataStax Astra DB represented a way for us to move to a managed Cassandra service that would be part of our company’s overall cloud migration, but it also provided us with more support and expertise over time. For us, Astra DB was a perfect fit during our move to the cloud.”

The Results

FINN.no uses Cassandra alongside its data lake as part of its data science pipeline. The Data Intelligence team takes data from the data lake and builds their models in Python. These models are then loaded into the company’s API framework and used alongside the data held in the Cassandra cluster. The cluster is made up of three clusters with three nodes each.

“Our Cassandra installation takes the most recent data from our user behavior and data models and uses this to personalize the adverts that users can see, so it has to work in real-time as people are browsing,” stated Kolstad. “We use the tools within DataStax to limit the volume of data that is held over time so we can respect customer privacy and we don’t hold on to data that is not useful within our data science models. DataStax provides the speed and availability levels that our personalization engine requires.”

After implementing DataStax Astra DB, FINN.no has worked with DataStax on supporting the performance and availability levels of the personalization engine. This helps the Data Intelligence team deliver the speed of recommendations that the FINN.no team requires, as well as supporting more organizations within the Schibsted Group in making use of personalization and data science in their day-to-day activities.

“Personalization can be hard, but it is essential for our team and for our business. Getting it right takes a mix of data science, infrastructure, and customer value. DataStax Astra DB supports our performance and availability needs from our Cassandra clusters, and helps us deliver recommendations that our customers think are valuable,” stated Kolstad.