Benchmark of a simple GraphQL server on popular spec implementations.
The benchmark consists in a simple blogging GraphQL server:
-
fetching json (author profiles) from a remote HTTP server
-
reading rows (posts and comments) from a Postgres database
type Post {
author: Author!
title: String!
content: String!
comments: [Comment!]
}
type Comment {
author: Author!
post: Post!
content: String!
}
type Author {
firstName: String!
lastName: String!
bio: String!
posts: [Post!]
comments: [Comment!]
}
type Query {
posts: [Post!]
author(id: Int!): Author
}The backend HTTP service is setup with 10 authors.
The Postgres database is setup with:
-
2 posts per author
-
2 comments per post:
-
1 comment from another author
-
1 comment (reply) from the post author
-
Each framework is tested with different queries.
The server must:
-
accept the
BACKEND_HOSTenvironment variable for configuration -
accept the
POSTGRES_HOSTenvironment variable for configuration
The server may:
-
cache author profiles per-query
-
use HTTP and Postgres pipelining
The server must not:
-
cache all posts and comments
-
cache all author profiles
-
3 quad-core servers Intel Xeon e5-2637v3 128G RAM (injector, server, backend)
-
1 quad-core server Intel Xeon e5-2637v3 256G RAM (postgres)
-
10 GbE switch
You can add an implementation in the implementations directory.
Make sure the name describes the software stack. For example, Java/vertx-graphql-java means the implementation:
-
is written in the Java language
-
uses Vert.x as the web server
-
uses GraphQL-Java engine
Each implementation must provide:
-
a
Dockerfile(build and run) -
a
set-env.shfile that defines theFRAMEWORK_NAMEimplementation (must be unique across implementations)
For testing purposes, you can build and run with Docker: