AWS Summit Seoul 2025 Presentation Recap - Cloud Scaling Automation for Our Precious Time

Oliveyoung experiences traffic spikes more than 10 times the usual level due to large-scale sales held 4 times a year and monthly marketing activities. In particular, when traffic comes in from advertisements, it shows a pattern of doubling in just 2 minutes.

Existing Auto Scaling takes about 3~5 minutes from the moment conditions are met after a traffic surge until instances are ready, which means it has the limitation of potentially causing about 10 minutes of actual outage. To solve this, Oliveyoung introduced a systematic automation strategy rather than a simple reactive response.

The first step for effective scaling is traffic categorization. Oliveyoung classifies and manages traffic into 5 levels from Low to Ultra High based on core business KPIs such as payments per minute.

Based on this classification, it selects resources to be scaled that match the nature of the traffic, and by combining business-metric-centered monitoring with SRE principles, it builds an automated response system that minimizes operator intervention.