Cloud Infrastructure Development for
a Finance and Credit Broker with AWS RDS Services

Advascale’s Client is one of Europe’s fastest-growing property finance and specialist lending companies. They had about 17 AWS accounts and exceptionally large bills and an “on-premises over cloud” infrastructure without autoscaling. Advascale is re-designing the infrastructure by cutting unnecessary costs

The Client inherited a set of AWS accounts that cost 17000 USD monthly. There were too many unused instances, EIPs, and other resources.

They had a totally “on-premises over cloud” environment, no automation, no spots, and no reservations.

All infrastructure needed to be re-designed almost completely.

Advascale first performed an audit to provide a plan for reducing costs, resource resizing, resource sizes, and reservations. The resulting document provided the guide to cleaning up the resources/cutting the expenses, and such information as project period, all costs, and expected results.

After that, it was decided to:

  • ​Move existent applications to high availability and scalable architecture
  • ​Use Kubernetes – by using Ocean
  • Cost-optimized with ECO
  • ​Make two environments, prod, and non-prod, in separate accounts
  • ​Send VPC flow logs and CloudTrail logs to separate accounts
  • ​Implement CI/CD by using Rolling updates

The application was deployed to the Kubernetes cluster managed by Ocean. Production and non-production environments were deployed to different AWS accounts using Terraform and Terragrunt.

Both environments have the same architecture but differ in resource sizes. As the database Aurora RDS for MySQL 5.7 with a separate reader for the production environment was used. For the session handler and caches, a Redis Replica Set was created.

CloudTrail and VPC flow logs were sent to S3 on a separate account. EBS encryption was enabled for the application regions.

The result of the project is:

  • ​cost decreased six times;
  • ​real-time visibility into cloud spending achieved;
  • ​high availability and scalable architecture performed;
  • ​all unused instances, EIPs, and other resources optimized;
  • ​new architecture allowed to invest less time in managing services.

The infrastructure is deployed in nineteen accounts within two regions: Ireland and London and some global things in N. Virginia. All services for every account were analyzed to reduce costs and make a plan for resource resizing, resource sizes, and reservations.

The total cost is reduced up to $5800/Mo instead of $17000 monthly.