Postgres Benchmark

[kiwicopple]

Chore

As part of our move from Alpha to Beta, we will want to be clear about the limitations and performance of our Postgres offering (free tier).

In this benchmark just the Postgres database.

We need to do 3 things:

• [ ] Design a benchmark (if it isn't already designed)
• [ ] Perform the benchmark
• [ ] Create a blog post or documentation detailing the results

[kiwicopple]

From Angelico:

Online resources (eg.) seem to point towards using pgbench and its resultant metric transactions per second or tps.

Should be a helpful decider between t3a.nano & t3a.micro for the DB instance

[steve-chavez]

Edit: As noted below the previous methodology here was wrong. This is now corrected.

pgbench tests for some of the m5a instances(tuned according to pgtune).

# steps taken from https://www.cloudbees.com/blog/tuning-postgresql-with-pgbench

createdb example
pgbench example -i -s 50 --foreign-keys

# Ran the following command 3 times on all the instance types
# This was run on a different machine with 8 cores(t3a.2xlarge)
pgbench -c 100 -j 8 -M prepared -T 30 example 

type	run1	run2	run3
m5a.large	1509.766608	1685.471195	1791.559195
m5a.xlarge	3081.890468	3799.897694	3836.188151
m5a.2xlarge	4753.743995	7775.855491	7836.620631
m5a.4xlarge	5702.618662	11688.693390	11828.751758
m5a.8xlarge	5854.134383	12403.998820	12550.028498

[darora]

(Just a note for any observers; there are some flaws with the methodology used for the above numbers, we'll be updating them soon)