The Finnish parliamentary elections of 2023 culminated in the vote count on election night, with hundreds of thousands of visitors flocking to media websites. This is how MTV and Cirit addressed the data processing challenges in MTV’s election results application (mtvuutiset.fi/vaalitulokset), and how AWS managed services delivered a 50% improvement in data processing speed.
About MTV
MTV Oy is Finland’s leading commercial TV broadcasting and video content company — the home of phenomena, discussion topics and stars. Its media group includes free TV channels, the MTV Katsomo streaming service and MTV News (mtvuutiset.fi), which shares the latest news across multiple channels. MTV also organises SuomiAreena, Finland’s biggest annual discussion forum, together with the city of Pori.
The election results application
MTV provides accurate and timely information to its users, and in election events — parliamentary and presidential alike — that instant transfer of data is what counts. MTV has invested in an election results service of its own, integrated into the main news site. During the previous elections in 2021 and 2022 there were problems with data processing speed: the application took over four minutes to process the voting data, which delayed the updates users were waiting for. Left unaddressed, that would send users to other platforms on election night.
There were 2,424 candidates in the 2023 parliamentary elections.
Solution architecture
Cirit, an AWS Select level partner, was enlisted to tackle the problem. The solution was built on AWS managed services: CloudFront, API Gateway, Aurora Postgres Serverless, DynamoDB, Lambda, CloudWatch, Amazon EventBridge, Secrets Manager and S3.
Every five minutes, Lambda functions fetched the data files from a remote server and stored them on S3. That triggered parallel Lambdas to unzip the files and batch-load the CSV data into an Aurora Postgres Serverless database, and the CloudFront and API Gateway caches were then updated with the latest data on request.
Boosting performance
- Data processing speed improved by 50% — from over four minutes to around two.
- At peak hours the application handled 45,000 to 95,000 requests, with an error rate between 0.5% and 1.75%.
- Optimising the database functions brought the biggest gains, and keeping Aurora at 384 ACUs from the start of the vote count played a crucial part.
Harri Lammi presented the project at AWS Community Day Nordics 2023: Boosting Performance with AWS Serverless in Elections.