A tool for benchmarking RabbitMQ on Kubernetes with various hardware and messaging configurations.
The tooling currently supports GKE, AKS, EKS, and Calatrava (Project Pacific) via terraform modules.
To configure the Kubernetes cluster and hardware to deploy RabbitMQ, edit the terraform configuration in
PROVIDER-cluster-config.tfvars, where PROVIDER is one of gke, aks, eks, or calatrava.
To configure RabbitMQ properties, edit the values YAML in RabbitMQ-values.yml.
To configure the benchmark topology, edit topology.json.
To configure the benchmark RabbitMQ policies, edit policy.json.
The topology and policy files are used to configure RabbitTestTool. For more detailed descriptions and configuration options, see the RTT documentation.
To provision an environment and run the benchmark, run the script
benchmark --provider (gke|aks|eks|calatrava)This script will deploy a Kubernetes cluster on the selected provider, deploy the cluster operator, a RabbitMQ cluster, Prometheus, Grafana, and InfluxDB on that Kubernetes cluster, then run the benchmark, exporting the results to the databases.
To use an existing Kubernetes cluster to run the benchmark, run the script
benchmark --skip-terraformThis will deploy the cluster operator, a RabbitMQ cluster, Prometheus, Grafana, and InfluxDB on the targeted Kubernetes cluster, then run the configured benchmark, exporting the results to the databases.
To access the Grafana dashboards, run the command
kubectl -n prom port-forward svc/prom-grafana 3000:80then open a browser window to http://localhost:3000 and login with the credentials admin:admin.
To tear down the infrastructure provisioned by the operator, run
benchmark destroy --provider (gke|aks|eks|calatrava)Note: this operation is destructive and will result in the loss of the benchmark data.
There is a significant amount of performance information captured in a series of blog posts by Jack Vanlightly from 2020.
| Messages/second | Message Size (KB) | Queue Type | Replication Factor | Cluster Size | Cores (per Node) | Memory (GB per Node) | Disk Type |
|---|---|---|---|---|---|---|---|
| 36,000 | 1 | quorum | 3 | 3 | 16 | 32 | SSD |
| 37,000 | 1 | quorum | 3 | 3 | 36 | 72 | SSD |
| 42,000 | 1 | quorum | 3 | 5 | 8 | 16 | SSD |
| 54,000 | 1 | quorum | 3 | 5 | 16 | 32 | SSD |
| 54,000 | 1 | quorum | 3 | 7 | 8 | 16 | SSD |
| 67,000 | 1 | quorum | 3 | 7 | 16 | 16 | SSD |
| 66,000 | 1 | quorum | 3 | 9 | 8 | 16 | SSD |
To add new benchmarking results, please open a PR with your addition.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent