Kakao Reveals Development Story of 'Work Environment as a Service' Platform AAAP for Applied Analytics Team

Kakao's Applied Analytics Team developed a work environment service called AAAP (Application Analysis Access Point) to solve the complex development environment configuration issues faced by data analysts. Going beyond simply providing container images, this service is a form of Work Environment as a Service that includes an interface managing the entire lifecycle from virtual environment creation to removal.

Existing Problems and Solution Direction

The Applied Analytics Team uses various tools such as Hadoop, Spark, Python, and R, requiring different computing resources by job role. This led to frequent issues of environment mixing, including dependency conflicts, tangled environment variables, and non-standardized environment configurations per individual. AAAP systematizes such implicit know-how, providing standardized virtual environments that allow team members to immediately engage in work without separate configuration.

Internal Platform Integration Architecture

AAAP was built utilizing Kakao's internal platforms.

• DKOS: Automated instance creation and management through DKOS, which is Kubernetes-based.
• D2Hub: Deployed reliable container images using D2Hub, a Docker image registry.
• Tenth2: Integrated Tenth2, a distributed storage system, for persistent data storage.
• WatchTower: Utilized WatchTower, an internal messaging bot, for idle resource detection and notification features.

User Experience Improvements

Initially, a CLI-based AAAP CLI was provided, but it transitioned to a web-based AAAP Web to improve accessibility. This allowed non-developer roles to create and manage virtual environments via a web browser without complex commands. Additionally, a feature that automatically detects idle virtual environments and sends notifications prevents resource waste.