Starcounter.MultiModelBenchmark

This is an implementation of the open source ArangoDB's benchmark using Starcounter 3.0.

• NoSQL Performance Benchmark 2018 – MongoDB, PostgreSQL, OrientDB, Neo4j and ArangoDB.

Setup

Starcounter.MultiModelBenchmark requires Starcounter 3.0 Alpha-20190930.

• Clone the Starcounter.MultiModelBenchmark repository.
• Create artifacts folder on the same level as Starcounter.MultiModelBenchmark folder.
• Download the latest available Starcounter 3.0 from starcounter.io and unzip it into the artifacts folder.
• Build Starcounter.MultiModelBenchmark with dotnet CLI or with Visual Studio 2019.
• Download and unzip test data into a folder.
• Start Starcounter.MultiModelBenchmark project with dotnet CLI or with Visual Studio 2019.
• Create required database indexes using create-indexes end point.
• Import test data using the import-profiles and import-relations end points.

See Starcounter.MultiModelBenchmark.postman_collection.json PostMan collection for the detailed REST API description.

Test data

• Pokec social network.
• Profiles - 1632803 vertices.
• Relations - 30622564 edges.

Benchmark

Starcounter.MultiModelBenchmark offers two benchmarking options.

1. Using Starcounter native C# language binding

In this mode, client is a Starcounter native application with direct database access, which means that database (server) and application (client) reside on the same physical or virtual server. Starcounter native applications have significant performance advantage over the traditional client-server network communication.

The benchmark payload is located in the data/data.zip archive, and has to be manually unzipped into the same folder prior to running any tests.

Use benchmark entry point to start required test. See the PostMan collection above.

2. Using HTTP REST API

In this mode, client is a classical REST API consumer.

Use ArangoDB NodeJs benchmarking suite to run the tests and collect results.

Starcounter REST API NodeJs implementation is available in the scripts/description.js file. There is also an sh script to perform all the benchmarks using ArangoDB's NodeJs benchmarker - scripts/.

Note: NodeJs consumes a lot of RAM when performing multiple parallel HTTP requests, it might be required to manually increase available amount of RAM for the NodeJs process.