The chart is honest about what the numbers mean. Higher score = more moving parts. ClickHouse at 71 is essentially a column store with opinions. Spark at 29 is "just" distributed compute, so you gotta wire everything yourself.
Real take though: the tool that looks "simplest" on this metric can absolutely be the wrong choice. We've seen teams pick Spark because it's "simpler" and then spend three quarters building plumbing that Presto would've given them for free. Or pick ClickHouse and get mad because it doesn't do transactions. Each engine is optimized for a different soul-crushing problem.
The missing dimension here: what does your data actually look like? If you're running OLAP queries on append-only events, this comparison matters. If you're doing weird hybrid stuff with deletes and incremental updates and god knows what else, none of these tools are actually designed for you and you'll be unhappy with all of them.
That said: Onehouse's hands-on take matters because they're not selling you one of these. They get to be honest about tradeoffs. Read the actual blog if you care.
ninadpathak•1h ago
Real take though: the tool that looks "simplest" on this metric can absolutely be the wrong choice. We've seen teams pick Spark because it's "simpler" and then spend three quarters building plumbing that Presto would've given them for free. Or pick ClickHouse and get mad because it doesn't do transactions. Each engine is optimized for a different soul-crushing problem.
The missing dimension here: what does your data actually look like? If you're running OLAP queries on append-only events, this comparison matters. If you're doing weird hybrid stuff with deletes and incremental updates and god knows what else, none of these tools are actually designed for you and you'll be unhappy with all of them.
That said: Onehouse's hands-on take matters because they're not selling you one of these. They get to be honest about tradeoffs. Read the actual blog if you care.