Add Cloud.ru Managed PostgreSQL results

[belo4ya]

This PR adds benchmark results for Cloud.ru Managed PostgreSQL.

Included results

• 4vCPU/16GB, SF=6849
• 16vCPU/64GB, SF=6849
• 16vCPU/64GB, SF=34247

Tested setup

• Single-node clusters
• Instance types: gen-4-16 (4vCPU/16GB) and gen-16-64 (16vCPU/64GB)
• Storage: 1050GB - SSD NVMe (50K IOPS)
• PostgreSQL version: 18.3 (Debian 18.3-1.pgdg13+1)
• Direct database connection, without a pooler
• Mostly default configuration, with only max_connections increased from 100 to 1000 to support 256 pgbench clients

Client machine

16vCPU/64GB VM located in the same availability zone as the tested database instance.

pgbench --version && psql --version
pgbench (PostgreSQL) 18.3 (Ubuntu 18.3-1.pgdg24.04+1)
psql (PostgreSQL) 18.3 (Ubuntu 18.3-1.pgdg24.04+1)

Reproduction

Example environment:

export PGHOST="<host>"
export PGPORT=5432
export PGUSER="dbadmin"
export PGPASSWORD="<password>"
export PGDATABASE="postgresbench"
export PGSSLMODE=prefer

export SYSTEM_NAME="Cloud.ru"
export INSTANCE_TYPE="gen-4-16" # or gen-16-64
export VCPUS=4 # or 16
export RAM_GB=16 # or 64

export INSTANCE_STORAGE=""
export PRIMARY_STORAGE="1050 GB - SSD NVMe (50K IOPS)"

export CLUSTER_SIZE=1
export TUNED="no"

export SCALE_FACTOR=6849 # or 34247
export CLIENTS=256
export THREADS=16
export DURATION=600

export OUT_JSON="cloudru/results/cloudru_4cpu_16gb_6849.json"
# or:
# cloudru/results/cloudru_16cpu_64gb_6849.json
# cloudru/results/cloudru_16cpu_64gb_34247.json

[vercel]

@belo4ya is attempting to deploy a commit to the ClickHouse Team on Vercel.

A member of the Team first needs to authorize it.

[CLAassistant]

All committers have signed the CLA.

[lio-p]

@belo4ya Thanks for your contribution. Looking at your result I can see they look similar across configuration, I suspect there is a bottleneck somewhere in your environment. Did you make sure to run the client and database in the same region?

[belo4ya]

Yes, the tests were conducted in one region, or rather in one availability zone.

You are right, there is indeed a bottleneck, but it is on the implementation side and the user cannot affect it. Currently, this solution uses a proxy-type NLB with an inefficient configuration. However, I hope that it will improve in the near future.

---

btw, I've conducted several more tests on different platforms, and one of them, which is quite young, managed to come quite close to the results of the ClickHouse solution: https://belo4ya.github.io/PostgresBench/, https://github.com/belo4ya/PostgresBench/tree/ru-bench