https://github.com/jorgebay/polar-benchmark-tool

A tool that makes it easy to benchmark PolarStreams
https://github.com/jorgebay/polar-benchmark-tool

Last synced: 7 days ago
JSON representation

A tool that makes it easy to benchmark PolarStreams

• Host: GitHub
• URL: https://github.com/jorgebay/polar-benchmark-tool
• Owner: jorgebay
• License: apache-2.0
• Created: 2022-10-10T08:51:28.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2023-04-14T09:08:38.000Z (almost 2 years ago)
• Last Synced: 2024-11-17T23:25:00.315Z (2 months ago)
• Language: Go
• Homepage:
• Size: 50.8 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE.txt

Awesome Lists containing this project

README

        
# PolarStreams Benchmarking Tool

Uses HTTP/2 clients to benchmark PolarStreams's producer API.

The load tool works in a similar way as [h2load] but it has the ability to generate a pseudo random payload to

try to mimic production values.

## Building

Use go 1.19+ to build the tool.

```shell

go build .

```

## Usage

```shell

./polar-benchmark-tool -h

Usage of ./polar-benchmark-tool:

  -c int

    	Number of clients (default 1)

  -ch int

    	Connections per host. (default 16)

  -h2

    	For workloads that allow it, use HTTP/2

  -hosts string

    	The host addresses of one of brokers in the cluster, the client will discover the rest of the brokers

  -m int

    	Max concurrent requests to issue per client. (default 32)

  -mr int

    	Number of messages per request in the workload (when supported) (default 16)

  -n int

    	Number of  requests across all  clients (default 100)

  -w string

    	The name of the workload (binary, get, http) (default "binary")

```

### Example

Starting a pseudo random workload with 2 million of messages with 6 clients targeting all hosts in the cluster.

```shell

./polar-benchmark-tool -w binary -hosts 10.0.0.100 -c 6 -n 2000000 -m 1024 -ch 1

Finished. Total responses 2000000 in 1224ms.

Latency in ms p50: 1.8; p999: 6.2 max: 6.2.

Throughput 1633986 messages/s (1633986 req/s)

```

[h2load]: https://nghttp2.org/documentation/h2load-howto.html