This repo maintains configuration to benchmark various GraphQL ORMs/backends (hasura, postgraphile, prisma). The database used is a Postgres variant of Chinook. The benchmarking tool used is graphql-bench.

Running benchmarks

You'll need to have docker and curl to run these benchmarks

1. Clone the repo.

   git clone https://github.com/hasura/graphql-backend-benchmarks.git && cd graphql-backend-benchmarks

2. Setup the graphql servers that you are interested in.

   ./hasura/manage.sh init
   ./prisma/manage.sh mysql init
   ./postgraphile/manage.sh init

3. Define the queries that you would like to benchmark in hasura/queries.graphql, prisma/queries.graphql, postgraphile/queries.graphql. There are few queries that already exist.

4. Define a benchmark (checkout sample.bench.yaml) say bench.yaml.

5. Run the benchmark

   cat bench.yaml | docker run -i --rm -p 8050:8050 -v $(pwd):/graphql-bench/ws hasura/graphql-bench:0.3

   or run a benchmark on a sample query

   cat sample.bench.yaml | docker run -i --rm -p 8050:8050 -v $(pwd):/graphql-bench/ws hasura/graphql-bench:v0.3 --query artistByArtistId

   This opens up a http server at http://127.0.0.1:8050 which displays the results of the benchmark.

6. Tear down the setup

   ./hasura/manage.sh nuke
   ./prisma/manage.sh mysql nuke
   ./postgraphile/manage.sh nuke

Sample plot

When the above benchmark is run, the results on an i7-4710HQ CPU, 8GB RAM, SSD machine are as follows: