Re-designing a container architecture for scalability

Availability for web scaling

Cloud services promise flexibility, scalability and reduced costs. But for these advantages to actually be exploited, the foundation of the corresponding systems must be built in a reliable way.

A company in the telecommunications industry started operating a consent management system designed for web scaling. The system was built to support compliance with GDPR requirements and is therefore essential for concluding contracts with end customers. However, during commissioning, shortcomings in the availability of IT resources were discovered. They were unable to follow the dynamic requirements and generated additional high costs.

A new approach

The client asked Liquid Reply to analyse the given platform and application architecture and to design a new approach ensuring the availability of the applications. Changes to the system resources and their configurations were also planned in order to achieve optimum utilisation.

The client’s goals

• Stabilise the application and its performance for productive use

• Increase high availability and reduce incidents during business peaks

• Optimise automation so that components are effectively updated and security risks are mitigated early

A Kubernetes system

The customer network was highly complex, with exceptionally strict rules for the protection of personal data and no room for failures. Liquid Reply decided to set up a fully managed, service-based Kubernetes system on the Managed Kubernetes Service of Amazon Web Services (AWS EKS).

Migrating during ongoing operations

As the consent management system was already up and running, the team faced the challenge of minimising the impact on the productive environment during the migration of the application to the new platform. Costs were also to be kept low during the migration process.

Liquid Reply migrated the entire application complexity, replacing various components such as the service mesh. This increased stability and security by providing a solution tailored to the specific application. In addition, newly established components ensure that security rules are continuously checked and adhered to.

The project steps in detail

• Migration of the existing, self-managed Kubernetes system to a managed service from AWS (AWS EKS)

• Replacing the Istio service mesh with Linkerd to improve the stability of all network communication

• Fully automating platform provisioning and enabling auto-scaling for all resources

The achievements

By refactoring the entire system design within the first few years, Liquid Reply reduced the number of failures and downtimes of the client’s systems by 100 per cent. As a result, the business-relevant system was continuously available for customer billing. Minor changes in the monitoring stack provided a better view of application performance. Ultimately, Liquid Reply reduced the cost of the client’s infrastructure by 43 per cent.