Romeo and Juliet benchmark

Usage:

Run this command with the correct port:

cargo run --release -- http://localhost:8000

This will go through the play defined in romeo_and_juliet.txt and create users for each character and sends one /send request for each line they say.

Results (2022-05):

Reproducible runs (contributed by ShadowJonathan)

While contrib contains some rudimentary documentation, to be able to use it, you must first:

• Have ansible-playbook installed
• Have a DigitalOcean account (with billing info registered, or using a budget)
• Have one of your public keys added to digitalocean (it's also important that your local machine tries to log into servers with this automatically)
• Install the ansible-galaxy collections documented in contrib's readme
• Add the DigitalOcean API token to your environment (also as documented in contrib's readme)

Then, cd into contrib/, and run

ansible-playbook digitalocean_playbook.yml

This'll run the tests in parallel on digitalocean droplets, results will be put into contrib/out/, you can adjust which server versions or revisions are ran/built using contrib/group_vars/server.yml, the committed file is "known to work", and can always be rolled back to when something goes wrong.

Some stages of the playbook could take a while, such as building rjbench, or conduit, or the testing itself. It's advised to leave it alone when it does this, although, if a stage takes more than an hour (as a rule of thumb), or if you're curious what's happening, take one of the IP addresses noted during the "Create droplets" stage, do ssh root@IP_ADDRESS, and look around with htop and journalctl -f -n 100.

Contact me:

Matrix: @timo:koesters.xyz