Skip to main content Skip to footer


A bold vision to free up the most important resource—time

Arek Oy streamlines pension calculations by moving to Google Cloud.


Call for change

Arek Oy delivers pension calculation services to all pension providers in Finland. These calculations, based on decades worth of data, are highly regulated and extremely complicated. They also come with stringent compliance and accuracy requirements. Historically, the company calculated the value of every Finnish citizen’s pension via a proprietary mainframe application and MIPS (million instructions per second) architecture. These valuations—drawn from analyses of 22 million lines of code and 7,000 COBOL modules—were then regularly passed along to the pension providers in the form of standardized reports.

Arek Oy needed to make a choice. It could update the existing MIPS architecture to accommodate growing demand, but the static costs associated with running the infrastructure were high and the COBOL skills needed to maintain it were in short supply. The other option was revolutionary. It involved shifting its system of pension calculations from the mainframe to a private cloud solution.

When tech meets human ingenuity

Arek Oy worked closely with Accenture to establish the necessary structure and technical framework to facilitate the conversion to Google Cloud—and, specifically, to the Google Cloud’s Anthos application management platform. Arek Oy selected this platform because of its faster CPUs and the fact that Google operated a Finland-based data center which would meet the country’s strict requirements for keeping sensitive data within its borders.

The team started by identifying the key requirements of the new solution. One of these was that the performance of individual calculations in the Anthos environment needed to match the mainframe performance perfectly. Arek Oy needed to be able to scale the cloud-based application in real time as demand varied. In addition, a new DevOps pipeline needed to be able to support ongoing application development and source code changes on the mainframe and then automatically deploy those changes to the Anthos infrastructure.

A valuable difference

The migration of Arek Oy’s entire pension calculation engine to Anthos was completed in under a year—and generated positive results much sooner than that.

Moving the calculation engine to a Kubernetes cluster managed by Anthos and executing it on an Azul JVM has enabled Arek Oy to improve its services to pension providers and to the millions of pensioners who rely on accurate, fast calculations. The highly automated, cloud-based solution is integrated with pension providers’ systems across Finland—thereby enabling them to meet pensioners’ requirements more efficiently.

Importantly, running the application on a private cloud is generating significant cost savings for Arek Oy. In fact, the company’s cost per transaction is projected to drop by more than 80% as a result of utilizing the Kubernetes infrastructure on Anthos instead of the legacy mainframe. On top of all this, Arek Oy has been able to eliminate the yearly expense of expanding the mainframe system’s computing architecture. In the new cloud environment, Arek Oy can dynamically scale its computing power as needed, in real time and at minimal cost.

Together, the lower cost per transaction and the ability to scale on-demand are having a direct and positive impact on pensioners’ experiences—and the profitability of Arek Oy. Its leaders were bold and forward-thinking when they decided to move the company’s most critical application to the cloud. Their calculated risk has paid off handsomely.


The company’s cost per transaction is projected to drop by more than 80% as a result of utilizing the Kubernetes infrastructure on Anthos instead of the legacy mainframe.


Only two Kubernetes pods were required to outperform the mainframe system by 26%. With four pods, outperformance reached 152%, with an average response time of less than 600 milliseconds.


Re-platforming was about more than just converting its 22 million lines of code, it would require a high-performance testing environment to ensure all pension calculations were converted accurately.