Fork of jkimbo/django-graphql-benchmarks. The original project proved useful for discussions about graphene's performance (e.g. in graphql-python/graphene#268). To provide an up-to-date insight into the matter I revived the project.
The Django app is no longer bound to be deployed on Heroku, but is built as a Docker container. This allows for deployment anywhere (including the very cost-efficient "localhost").
main runs on up-to-date dependencies, Python 3.10 and graphene v3.
main_graphenev2_only is (or should) always the same as main, but only exposes a GraphQL endpoint provided by graphene v2.
Prerequisites for running the benchmarks:
- A local version of the
mainandmain_graphenev2_onlybranches - Docker and Python 3.8+
- poetry
mkdir -p results
You can use the run.sh script to deploy and bench the containers locally, or use it as a blueprint for other deployments.
run.sh requires a list of target Python versions as positional arguments, e.g. run.sh 3.9 3.10
Benchmark ran on an AMD Ryzen 9 5900X with 64GB memory and 8 uWSGI workers.