I used the insert benchmark to compare ARM and x86 servers on AWS. For ARM I used c7g and for x86 I used c6i. The goal was to compare ARM vs x86 on si

Small Datum: The insert benchmark on ARM and x86 cloud servers

submited by
Style Pass
2023-01-24 19:30:10

I used the insert benchmark to compare ARM and x86 servers on AWS. For ARM I used c7g and for x86 I used c6i. The goal was to compare ARM vs x86 on similar hardware (OS, vcpu, RAM and EBS setups were the same). 

While I am not trying to start a benchmarketing battle for c7g vs c6i or Postgres vs MySQL, I suppose that might happen. My goal in doing comparisons like this is to find things that can be improved. Also, in-memory workloads aren't the target use case for MyRocks. It doesn't do great on them. It does do great on write and space efficiency, but those don't matter here. Were this an IO-bound benchmark the MyRocks results would look much better in comparisons.

The benchmark was repeated for Postgres 15.1, MySQL 8.0.31 with InnoDB and FB MySQL 8.0.28 with MyRocks. All were compiled from source with gcc (gcc 11.3.0 on Ubuntu 22.04) and I didn't go out of my way to change the compiler flags.

An overview of the insert benchmark is here and an update is here. I used a CPU-bound workload, the database fits in the DBMS cache so there were no reads from storage but many writes.

Leave a Comment