Back
Key Takeaway
Infrastructure optimization based on AI Ops reduces operational costs and accelerates development speed
By applying AIR DevOps's AI Ops to cloud infrastructure operations and management (EKS, Aurora RDS), we analyzed the causes of Cloudwatch cost increases and achieved efficiency improvements. Additionally, through proactive technical support, we shortened the development and deployment lead time of the core event system.
F&B (S Company)
Client :F&B (S Company)
Industry :Retail / Software
Service Area :Applications & DevOps / Managed Services / Data & AI
Applied Solution :AIR
1. Overview (Project Background)
S Company's seasonal promotions are representative high-traffic services where hundreds of thousands of customers access simultaneously each season. In the existing IDC-based Jeus/WebtoB + JSP environment, it was difficult to flexibly respond to traffic surges at 7 AM opening times, and scalability limitations due to legacy architecture were ongoing challenges.
MegazoneCloud built a cloud-native environment capable of stably handling traffic spikes through container architecture based on AWS EKS and modernization based on React/TypeScript.
2. Challenge (Problem Definition)
- Extreme traffic concentration — Users surge within approximately 5 minutes at the gift reservation opening time (7 AM), making stable service processing essential during this time period.
- Scalability limitations of legacy environment — The existing On-Premise-based SSR (Server-Side Rendering) approach made it difficult to respond flexibly to traffic changes.
- Short-term transition requirements — There was a scheduling challenge to rapidly rebuild JSP source code containing existing frameworks to the latest stack.
3. Solution (Resolution Approach)
MegazoneCloud successfully executed the project by combining architecture optimization with AI-based development automation.
- Scalable container architecture design
: We redesigned the structure to remove direct database connections from the frontend and have client terminals call APIs directly. This secured a flexible structure where processing capacity increases linearly when containers are scaled up in response to traffic increases.
- Traffic distribution through CDN transition
: We transitioned the existing IDC CDN to AWS CloudFront and separated static resources such as CSS/Image/Font to minimize the load on the origin server. We pre-calculated expected traffic bandwidth and secured bandwidth to build a stable service operation environment.
- Frontend modernization
: We transitioned the existing SSR (Server-Side Rendering) JSP environment to a CSR (Client-Side Rendering) approach based on React/TypeScript/Next.js. This improved frontend performance and enhanced user experience.
- Development acceleration using AI Code Assistant
: We utilized Cursor and Claude to analyze existing source structures and automatically convert JSP code to React/TypeScript. We achieved groundbreaking productivity improvements compared to manually converting vast legacy sources.
4. Result (Achievements)
- 200% improvement in development productivity — By utilizing AI for automatic code conversion, we reduced the work that previously required 6M/M to 2M/M.
- Analysis phase: 1M/M → 0.5M/M (Source structure analysis using Cursor/Claude)
- Design and conversion phase: 5M/M → 1.5M/M (Automatic conversion of vast sources)
- 100% stable handling of peak traffic — Through pre-calculated expected traffic bandwidth and horizontal container scaling design, we stably processed traffic surges at opening times.
- Maximized infrastructure operation efficiency — By transitioning the legacy IDC environment to AWS EKS-based cloud-native and implementing CloudFront CDN and auto-scaling, we minimized idle resource waste and secured operational cost efficiency.
