FYI, this is by far the best TPC-C implementation for Postgres: github.com/petergeoghegan. The OpenSCG repo disappeared so I set up a mirror. Postgres 12 has decent TPC-C results. Getting a valid result is expensive due to spec's cap on xacts per minute per warehouse.
Conversation
TPC-C is required to be I/O bound, which is why it costs so much to run correctly. Many newSQL systems have unofficial TPC-C results with unrealistic TPS for the available warehouses (arguably this is cheating). See cockroachlabs.com/blog/2-dot-0-p.
1
1
Also arguably a more relevant workload? IMO the low xact per warehouse limit makes the workload less realistics given today's workloads?
Still makes it annoying to compare results, of course.
2
1
Right. TPC-E addresses this criticism by being far more read heavy, and more cpu bound. It's a lot easier to run. But most vendors prefer TPC-C because it's far easier to shard.