Oodle benchmark finds ClickHouse lags DenaliDB on metrics queries
Start here
Oodle benchmarked ClickHouse's ClickStack against its own DenaliDB metrics engine and concluded ClickHouse slows sharply as active series, time range and concurrency grow. On a single 192 vCPU NVMe node, ClickHouse took 176 s at 1K series with 128 concurrent queries (p90), versus 3.2 s for DenaliDB; at 250K series over 24 hours it took 124 s versus 3.7 s. Oodle says over 93% of its metric queries now come from agents, making concurrent-load performance critical. Oodle says it used AWS on-demand rates for ClickHouse hardware costs and the default ClickStack schema, without testing materialized views. It also says a ClickHouse feature it wanted to test was still in private preview and could not be benchmarked. On that basis, Oodle says it is not switching its metrics engine to ClickHouse at this time.
Generated from reports · updated 50 min ago
Report timeline
Follow the reports to see the event from different sides.
- Hacker News · AIHas ClickHouse Conquered Metrics? Oodle's DenaliDB Benchmark Says Not Yet
Oodle benchmarked ClickHouse's ClickStack against its own DenaliDB metrics engine at 25M active series and found ClickHouse slows sharply as series, time range and concurrency grow. On one 192 vCPU NVMe node, ClickHouse took 176 s at 1K series with 128 concurrent queries (p90) versus 3.2 s for DenaliDB, and 124 s versus 3.7 s at 250K series over 24 hours. Oodle says over 93% of its metric queries now come from agents, making concurrent-load performance critical.
Heat for this event
Not enough continuous observations yet to draw a trend.