📊Liquid Clustering Outperforms ZOrder for Most Lakehouse Workloads
Most teams should switch to Liquid Clustering
TL;DR
New benchmarks show Liquid Clustering outperforms ZOrder in most cases, delivering lower latency despite scanning more data. Ideal for diverse or evolving query patterns.
New benchmarks reveal Liquid Clustering outperforms ZOrder in most lakehouse workloads, delivering lower latency despite scanning more raw data. Teams with diverse or evolving query patterns should switch to Liquid Clustering for better performance. Liquid Clustering wins on three out of four workload types, achieving the lowest latency at 1.003 seconds, 49% faster than the partitioned baseline, despite scanning more raw data than ZOrder.
Key Points
Liquid Clustering delivers the lowest latency on point lookups despite scanning more raw data than ZOrder.
ZOrdering performed better only for stable, highly selective multicolumn predicate workloads where the filter columns precisely match the clustering key.
Liquid Clustering achieved the lowest latency at 1.003 seconds, 49% faster than the partitioned baseline, despite scanning more raw data than ZOrder.
A synthetic transactional dataset with 50 million records was used to simulate a retail purchase environment.
The results show that Liquid Clustering's Hilbert curve organization colocates related customer_id values within files.
Why It Matters
If your lakehouse queries filter on different columns or patterns evolve, Liquid Clustering is the way to go. It delivers lower latency and removes the operational overhead of manual OPTIMIZE scheduling. For stable, highly selective multicolumn workloads, ZOrder remains the better choice.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 3,566 builders reading daily.