SK Planet's Planet AD Cuts Costs by 27% Through AWS Infrastructure Optimization

SK Planet's reward advertising platform Planet AD optimized its overall AWS infrastructure for service stability and cost reduction. Through this, it achieved a result of reducing AWS costs by approximately 27% compared to the launch point.

EKS Stability and Autoscaling Improvements

By changing the Warm Pool settings of EKS Nodes, IP waste was eliminated, and KEDA (Kubernetes Event-driven Autoscaling) was introduced to implement Pod scaling tailored to traffic patterns. Additionally, CPA (Cluster Proportional Autoscaler) was applied to resolve CoreDNS's DNS lookup latency issues.

Database and Security Cost Optimization

By modifying DynamoDB's data model to include up to 100 segments in a way that changed from the previous allocation of RCU/WCU per single segment, Read/Write Capacity usage was significantly reduced. RDS Proxy was applied to handle Connection management and fail-over for the Golang application, and the introduction of Datadog ASM lowered costs to about 1/3 compared to the existing AWS WAF.

Instance and Storage Efficiency

Graviton instances were applied to RDS and Elasticache, reducing costs by up to 20%, and Spot by NetApp was utilized to manage Spot instance costs. EBS Volumes were changed to GP3, and fixed costs were minimized through RI (Reserved Instance) contracts.