"ClickHouse can't do joins" — is that actually true? 🏎️ A no-tuning, same-query benchmark across three platforms makes for a fun read. Title: Join me if you can: ClickHouse vs. Databricks vs. Snowflake — Part 1 URL: 🏎️ Overview A benchmark comparing ClickHouse, Databricks, and Snowflake on join-heavy SQL workloads using identical queries and data, from 721M to 7.2B rows — testing the conventional wisdom about join performance. ❓ Challenges Solved ClickHouse is known as a fast analytics DB, yet there's a persistent belief that "it can't do joins." ・That belief often becomes a reason to drop ClickHouse when choosing an analytics stack ・This post checks whether it's true via a same-conditions three-way comparison 💡 Methodology ・Reproduces an existing "coffee shop" benchmark that originally compared Databricks and Snowflake ・Runs the 17 join-heavy queries on ClickHouse Cloud with minimal changes ・Key detail: no tuning at all — not the queries, not the ClickHouse side ・Each query run 5 times with the fastest reported, across 2-16 node configs (AWS) 📊 Experimental Results ・721M rows: most queries finish in under 1 second — 3-5x faster than alternatives at lower cost ・1.4B rows: one query joins and processes 1.7 billion rows in just 0.5 seconds (competitors need 5-13s) ・7.2B rows: even complex queries finish in seconds, not minutes It shows ClickHouse handles large multi-table joins efficiently without special configuration. #ClickHouse# #Database#

When a partitioning change to our petabyte-scale ClickHouse cluster caused critical billing jobs to stall, standard metrics showed no obvious errors. Here's how we identified severe lock contention in ClickHouse's query planner and built upstream patches to fix it.

Today I'm writing a weekend guide on how to do DD when shorting $NBIS: First, you look at hyperscaler earnings for AI cloud read through: > $GOOGL: reports record AI cloud demand + backlog + margin increases from earnings > $AMZN: reports record AI Cloud demand + backlog + margin increases from earnings > $META: reports higher than expected prices for available capacity from earnings. Now, time to look at Nebius: -> $NBIS: Growing hundreds of percent to $7-9B ARR by Q4. Growing margins, and guided 4GW+ contracted power. -> Sees Uber/Waymo splitting, putting more focus on Avride -> Sees Clickhouse growing rapidly every quarter. Okay looks bad! But next, you need a hedge? -> Wow! A $NIKE brand executive, after the stock dropped 75% over the past 5 years, went to $LULU to save that brand next? Lululemon seems good. Conclusion: Short Nebius and go long on $LULU