Squeezing Performance: Clickhouse@4GB on Kubernetes

ClickHouse is renowned for its high-performance analytics—but can it run on resource-constrained environments? In this talk, we present the findings from our deep-dive benchmark of ClickHouse on Kubernetes with as little as 4GB RAM. We share insights into system bottlenecks, including aggressive page cache accumulation, jemalloc tuning, and the impact of transparent huge pages. You’ll learn how we systematically pushed ClickHouse to its limits, what failed, what worked, and the trade-offs between stability and performance. This session offers practical recommendations for running ClickHouse in memory-constrained environments.