Deployment time

When every minute counts

Project overview

CLIENT

Vocalink

INDUSTRY

Financial

LOCATION

United Kingdom

TIMELINE

10/2023 - 11/2023

Task

The Solution

After the analysis process, we established several facts that significantly impacted the duration of Ci/Cd tasks. First, duplicate sections of code have installed/configured the same service multiple times. Second, pipelines have always installed everything from scratch, which, on the one hand, is not a bug, but it takes time. Quite by accident, it also turned out that the environments created by the pipeline are over-scaled, and several cost optimizations can be made.

 

Technology used

ansible
azure
erlang
aws
java
terraform
gitlab
kubernetes

Elimination of duplicates

Cleaning the code from duplicates and adding tasks verifying whether a given service has already been installed significantly influenced the execution time of the pipelines.

AMI and Docker images

Static dependencies have been enclosed in Docker and AMI images. Thanks to this, it was possible to freeze them in specific versions and not install them from scratch each time.

Replacing RDS with self-hosted

Replacing RDS (Oracle) with a version installed directly on the EC2 instance significantly reduced the costs of the cloud infrastructure.

Cost reduction

Additional cost reductions were achieved through more accurate selection of EC2 instance sizes, changing the volume type to gp3 and optimizing IOPS.

The Result