Amorepacific is a global beauty company generating 90% of its revenue from cosmetics and 10% from its Daily Beauty (DB) business. Operating across South Korea, the U.S., Japan, Europe, India, and the Middle East, it reaches customers through both offline channels and online platforms such as AmoreMall, Amazon, and TikTok Shop. Embracing digital innovation with AI-based skin diagnostics (“Dr. Amore”) and personalized solutions (“Custom Match”), Amorepacific pursues its vision as an “Asian Beauty Creator” while promoting sustainability, as recognized by its membership in the UN Global Compact and inclusion in the Dow Jones Sustainability Index (DJSI).

Building on its previous experience of successfully migrating 300 services entirely to AWS with MegazoneCloud, Amorepacific undertook this modernization project with the goal of reducing operational complexity and costs while ensuring stable service operations. The initiative focused on standardizing and enhancing deployment and monitoring processes, establishing standardized container image policies, refining CI/CD and monitoring systems, and transferring operational capabilities to the internal team. Over a six-month period, the project targeted the containerization of 60 services, with four main objectives: accelerate deployment speed, improve incident detection, shorten incident response times, and reduce infrastructure costs.

 

During this modernization effort, Amorepacific faced several operational and technical challenges:

 

1. Cost and scale

Achieving a 60% reduction in infrastructure expenses while migrating 60 services to containers within six months.

2. Operational complexity

Non-standard operations caused by heterogeneous deployment methods and varying container/image policies across services.

3. Limited visibility

Low incident detection rates and unintegrated deployment and alerting systems increased operator workload and delayed response.

4. Deployment instability

Issues such as 502 errors, container “zombie” processes, and OOM occurrences hindered zero-downtime deployment.

5. High dependency on the operations team

Build, deployment, and infrastructure information was accessible only through a specific team, slowing issue resolution.

To overcome these challenges, the initiative established standardized container policies, automated CI/CD deployment, improved monitoring and alerts, enhanced operational visibility, and implemented a risk management framework to identify, respond to, and prevent recurring issues.

1. Assessment & Standardization Design

Analyzed services to identify migration complexity, deployment/security requirements, and interface differences (e.g., S3 OIDC, LB health checks, fixed IPs).

Defined standardized container image, deployment, and security policies.

Managed the project in phases, introducing risk assessment matrices and change management principles.

2. Container Standardization & CI/CD

Established standardized container images and operated a container registry.

Automated build and deployment via GitLab pipelines to accelerate deployment speed and ensure configuration consistency.

Automated deployment alerts (dedicated channel) to enhance visibility of deployment events.

3. Monitoring & Operational Visibility

Created container-specific alert channels (deployment/operations) and reset container-specific thresholds.

Integrated with operational tools (e.g., Splunk) to visualize service status, logs, and resource usage in real time.

Enabled immediate tracking of build/deploy status via GitLab pipelines, reducing initial response time.

4. Risk Management (Zero-Downtime & Stabilization)

Zombie containers: Optimized cluster management settings to ensure proper application termination.

Zero-downtime deployment / 502 errors: Adjusted health check intervals and paths, improved Nginx (proxy) configuration.

OOM issues: Conducted comprehensive memory assessment, classified services needing additional resources, and optimized allocations.

Operational stabilization: Strengthened alerts (Teams), optimized cluster autoscale policies, and completed operational training and handover.

*** Architecture Overview

 

1. Container Orchestration

Based on AWS ECS and Terraform (standardized from an operational and deployment perspective).

2. CI/CD

Single-entry deployment through GitLab Runner pipelines.

3. Monitoring & Logging

Operational visibility ensured via Splunk, FluentBit, and dedicated alert channels.

4. Multi-Cloud Readiness

Portability and scalability considered through standardized images and policies.

Amorepacific’s modernization initiative delivered significant improvements in both operations and efficiency. Automated deployment processes and standardized container policies strengthened security and reduced operational complexity, while integrated pipelines and monitoring enhanced visibility and accelerated incident detection and response. Alert quality improvements and dedicated channels reduced operator fatigue, and service-specific adjustments eliminated recurring issues such as zombie processes, 502 errors, and OOM occurrences—ensuring a smooth, stable transition.

 

Key quantitative outcomes include:

 

1. Infrastructure cost reduction

Annual costs decreased from $390K (400 EC2 instances) to $60K (34 EC2 instances including solution licenses), an 84% saving.

2. Deployment speed

Average lead time from source commit to production shortened from 45 minutes to 15 minutes (66% faster).

3. Incident response

Alert-to-acknowledgment time reduced from 5 minutes to 1 minute (80% faster), and recovery time dropped from 20 minutes to 5 minutes (75% faster).

4. Stability

Monthly OOM incidents reduced from 5 to 0 (100% elimination).

5. Operational visibility

Effective alert rate improved from 42% to 78% (85% improvement), while incident detection time was shortened from 10 minutes to 2 minutes (80% faster).

1. Standardization is the foundation of innovation

Simply migrating to the cloud is not enough to accelerate business. By standardizing container images and deployment processes, Amorepacific reduced operational complexity while strengthening security and stability.

2. Automation maximizes operational efficiency

Moving away from manual deployment and monitoring, Amorepacific implemented GitLab-based CI/CD pipelines and real-time monitoring, reducing deployment lead time by 66% and significantly improving incident response speed.

3. Visibility drives agility

Real-time dashboards and integrated alerting enabled faster detection and resolution of incidents, while also reducing operator workload.

4. Incremental risk management ensures stability

Service-specific optimizations—such as memory allocation, health check adjustments, and cluster management improvements—eliminated recurring issues like OOM errors, 502 errors, and zombie containers..

5. Cost savings can fuel innovation

Through operational optimization and efficient resource management, Amorepacific reduced infrastructure costs by 84% while maintaining service stability and speed, freeing up resources to reinvest in new services and enhanced customer experiences.