Skip to content

ClickHouse for o11y at Kakao / ClickHouse 활용 최적화 여정

Mark Choi / 최영록

Data Engineer at Kakao / Kakao 데이터 엔지니어 Kakao Observability 전환 과정에서 ClickHouse를 핵심 스토리지로 도입하며 많은 시행착오를 겪었습니다. 이 발표에서는 ClickHouse를 Observability Storage로 실제 운영하면서 무엇이 잘 작동했는지, 무엇이 실패했는지, 그리고 그 과정에서 얻은 실무적인 교훈을 공유합니다. 이상적인 아키텍처보다, 현실적인 선택과 판단의 기록을 중심으로 이야기합니다.


During Kakao’s transition to Observability, Mark's team introduced ClickHouse as the core storage and learned through extensive trial and error. This talk shares what worked, what didn’t, and the practical lessons learned from running ClickHouse for observability, focusing on real-world decisions rather than idealized architectures.

Recent videos

YouTube Video: GpRgNjD2Swk

Open House, User stories

Precomputing fraud signals with ClickHouse materialized views at Moniepoint

Johnson Adekunle and Abraham Imohiosen, Engineering Managers at Moniepoint, explain how precomputed customer state with ClickHouse materialized views turns fraud checks into lookups, reducing p90 aggregate lookup latency by 93.1%.

YouTube Video: kcSlDrWd-HU

Open House, User stories

How Twilio handles a petabyte of logs a day with ClickHouse Cloud

Avleen Vig, Site Reliability Architect at Twilio, explains how ClickHouse Cloud supports 30 million log lines per second at peak, a petabyte of raw logs per day, and 50% annual data growth.

YouTube Video: hWOe1H2nW3g

Open House, User stories

Observability for the age of agents: How Vercel uses ClickHouse

Dom Busser and Vincent Voyer explain how Vercel uses ClickHouse Cloud to ingest 100 billion log events a day, serve customer-facing observability, and support agents querying data with SQL.

Follow us

XBlueskySlackGithubTelegramMeetupRSS