Some Initial PostgreSQL 11.0 Database Benchmarks

[phoronix]

Phoronix: Some Initial PostgreSQL 11.0 Database Benchmarks

Among other software releases, yesterday brought the debut of the PostgreSQL 11.0 database server. Given it has possible performance enhancements and the new (non-default) LLVM-based just-in-time compilation ability, I decided to run some benchmarks on the powerful Dell PowerEdge EPYC 2P server...

http://www.phoronix.com/scan.php?page=news_item&px=PostgreSQL-11-JIT-Benchmarks

[anarki2]

Thanks for fixing the charts on iOS, appreciated.

[Michael]

Thanks for fixing the charts on iOS, appreciated.

Was a simple fix once hearing it and then getting access to an iOS device for a few minutes (the Chrome simulator didn't behave the same).

[alpha_one_x86]

I'm interested about benchmark on low end ARM, jit improve drastically the performance?

[GreenReaper]

JIT won't improve every query, and probably won't even be run on most queries (unless your database architecture sucks). It may help computationally-expensive queries (which the optimizer recognizes as such) which have to perform calculations to return values and/or determine which values to return. Think business analytics or other potentially complicated searches with qualifiers that can't be effectively indexed. On the plus side these can also be some of the longest to run, and provide the greatest perceived end-user delay, so it may provide a benefit disproportionate to the benchmark results. The same applies to queries which can be parallelized; several new operations can be parallel in this release (for some DBAs, parallel CREATE INDEX will be huge).

[gilboa]

Michael,

What's the (RAID) setup of the 20 SSDs?

- Gilboa