Performance degradation from 3.1.2 to 4.2.0

We are attempting to upgrade from redis 3.1.2 to 4.2.0, but in our testing, we observed a throughput degradation with the upgraded version. As I was investigating, I found the benchmarks in this project, but it appears that the tests favor 4.x, as they take advantage of automatic pipelining, while the v3 tests were not executing as a pipeline. When I corrected for this, I was able to reproduce the performance reduction. I also created a separate project to run a number of performance tests, which suffered similar results.

Forked branch containing the changes to the node-redis benchmarks:
https://github.com/jspitzig-cribl/node-redis/tree/benchmark-pipelining

Performance test project:
https://github.com/jspitzig-cribl/redis-perf-comparison

Perhaps there is a mistake or misunderstanding in my tests. If this is the case, please let me know.

Environment:

• Node.js Version: v14.18.3
• Redis Server Version: 6.2.5
• Node Redis Version: 4.2.0
• Platform: Ubuntu 22.04, MacOS 12.3.1

Start by comparing the node-redis benchmarks with the benchmark-pipelining branch, then run the redis-perf-comparison project under the listed Node.js, Redis, and node-redis versions. Done means determining whether the regression comes from the benchmark methodology or the client implementation, with a reproducible finding and an agreed next step.