v4 supports (multi-doc) transactions for the 1st time but, leveling off at 14 threads, can’t compete with #Postgres, a traditional #RDMS, according to an independently conducted benchmark commissioned by — #PostgresVision #noSQL
Conversation
Except that published everything: whitepaper, config, source code, 100% automation to reproduce the results and a bucket with the raw data. And just unproven statements.
Stay tuned, will publish a response blog post in the coming days.
1
4
14
The OnGres trials ignored published best practices for MongoDB, claiming falsely that MongoDB does not require or benefit from tuning, used an experimental driver without connection pooling, and ignored indexes. Those aren’t unproven statements, they’re in the code and report.
2
5
There’s no need for MongoDB to publish anything — the OnGres repo is itself sufficient for any tester familiar with `explain` to see the problems with it.
3
1
3
There may well have been flaws; benchmarking is hard. But the counterclaim about "Query A" being ~240x faster on MongoDB is absolutely extraordinary. This difference seems more qualitative than quantitative. Thought experiment: Why is it not (say) 1000x faster than Postgres?