Sun Apr 21 11:02:05 2024

Benchmark Report

The results below were generating using the IMDBench benchmarking suite: github.com/edgedb/imdbench.

Targets

• edgedb 4.5
• neon drizzle
• supabase drizzle
• planetscale drizzle
• neon prisma
• supabase prisma
• planetscale prisma

Test environment

Linux 6.2.0-1014-aws (Ubuntu 22.04, x86_64) on AMD EPYC 9R14

Duration (per benchmark)

10 seconds

Concurrency

20, 20, 20, 20, 20, 20, 20 clients

Simulated client-to-database latency

~0ms

Overall Results

The charts below show the geometric average of all benchmark results.

Throughput (more is better)

Latency (less is better)

Individual benchmarks

POST /movie (existing cast)

Create a new movie record, linking existing directors and cast.

Throughput (more is better)

Latency (less is better)

Sample Outputs

GET /movie/:id

Get information about a given movie: title, year, directors, cast, recent reviews, and average review rating.

Throughput (more is better)

Latency (less is better)

Sample Outputs

GET /user/:id

Get information about a given user: name and a sample of the latest movie reviews this user authored.

Throughput (more is better)

Latency (less is better)

Sample Outputs