Kakao Reveals Data Processing and Operations Sessions at if(kakaoAI)2024

Kakao has unveiled nine technical sessions themed around 'Data' at the if(kakaoAI)2024 conference. These sessions share large-scale data processing technologies used in Kakao's actual services, along with challenges and solutions encountered during database operations.

Key Session Highlights

Various topics will be presented across DAY1 (October 22) and DAY3 (October 24). The key contents are as follows:

• AI-based Ad Recommendation: Introduces the real-time data processing pipeline architecture and monitoring strategies utilizing Spark Streaming.
• MongoDB Auto-Upgrade: Shares the case study of building the 'MUA' platform, which conveniently upgrades large-scale internal MongoDB instances and reduces operational costs.
• Kafka Kraft Operations: Covers practical operational know-how and deployment steps for Kafka supporting Kafka Raft (Kraft) without Zookeeper.
• Large-Scale Data Synchronization: Introduces the construction of a CDC system that synchronizes billions of records in real-time from MySQL to Iceberg through the 'SyncBus' project.
• Real-time User Activity Data Serving: Shares the case of optimizing performance by storing and serving user activity data, such as clicks and likes, in real-time via the 'ActionBase' system.
• MySQL DB Sharding: Explains the process of introducing PK-based sharding to resolve increasing transaction volumes and performance issues in the billing system.
• Gifting Caching Strategy: Discusses caching strategies to withstand high traffic, such as Hybrid Cache and Auto Cache Warm Up, and the challenges faced during implementation.
• Game Data Analysis: Introduces the statistical project (SMART STATS) aimed at preventing MMORPG user churn and the construction of a game data search system accessible to users unfamiliar with SQL schemas.