Alex Shevlakov

Hi Steve, the TPC reports are here at http://pgrocks.com/tpc/ - the two pdfs available now. Yes, I did manually reshuffle the Qs order in 20 scripts before running them, as well as I changed the numbers and years, etc. to make 20 different sets of queries, so that the times from differents sets would differ based on randomly zeroing output from otherwise long-running queries. Then I'd use "echo 1 > /proc/sys/vm/drop_caches" to clean the buffers before run#1. I'd compare "normal" and "optimized" sets,...

Hi, we passed TPC-C on PG11.1 with HammerDB 3.1 and a setup with Dual Xeon Silver 4114 (20cores, 40threads) /256GBRAM/1TB data NVMe, 800 warehouses, achieving 1,200k tpm/500k nopm for 100 Vusers. As I did this I checked numbers of orders before and after the tests, and [nopm multiplied by test duration] was actually about the number of new records, so that was the first good "ground truth". Now I'm doing TPC-H (SF10 and SF100) and here I'm a bit confused because the script 22 queries from psql console...