AI Briefing#kakao
The latest AI and developer news about #kakao, with the original source and a short summary.
The latest AI and developer news about #kakao, with the original source and a short summary.
Kakao revealed the architecture and operational technologies of its real-time recommendation system 'Rubix', highlighting 99.998% availability and the ability to handle 200,000 QPS during peak times.
Developed since 2013, MobiL supports automated builds based on Jenkins and MacPro, along with deployment via an internal-only app center.
Kakao has released 'hbase-tools', an open-source tool that automates metric monitoring, region management, and snapshot management for HBase clusters.
This DNS TTL 0-based Redis HA approach restores connections by immediately switching the domain IP to a slave upon master failure.
Kakao has released 'HBase Region Inspector', an open-source tool that visualizes region distribution and request status in HBase clusters.
The tool operates stably with 2-3% CPU usage in a 30,000 QPS environment and its source code is available on GitHub.
The KakaoStory web client team shared their experience in resolving bottlenecks during the adoption of code reviews by implementing a review master, preview server, and convention automation.
This technology processes 200 million users and 10 billion relationship data points via real-time breadth-first search.
To resolve traffic inflow issues during deployment caused by L4 device session table retention, a procedure was presented that involves waiting for HealthCheck failure before stopping the web server process to clear sessions.
Leveraged the Quagga open source to enhance the stability of DNS and Contents Cache services without dedicated hardware.
It introduced DSCP-based configuration and health check methods to overcome the physical limitations of L2DSR and enable load balancing across different IDCs.