bench/

README

benchmark

directories

bench/
  benchmark/  # CLI benchmark script
  agent/      # Communication with pubsub server and frontend on the benchmarker

benchmark

run command bin/app. if your will build make build.

options:

• port - choose webapp server port
• host - choose webapp server host
• file - path of webapp application account list file
• agent - used benchmark request user agent

example run command:

bin/app -host=localhost -port=8080 -file=data/param.json -agent=isucon_go

result

Benchmark returns JSON result that include fields valid, request_count, elapsed_time and so.

example benchmark result when succeed:

{
        "valid": true,
        "request_count": 3651,
        "elapsed_time": 0,
        "response": {
                "success": 1452,
                "redirect": 2199,
                "client_error": 0,
                "server_error": 0,
                "exception": 0
        },
        "violations": []
}

when failed benchmark:

{
        "valid": false,
        "request_count": 1,
        "elapsed_time": 0,
        "response": {
                "success": 0,
                "redirect": 0,
                "client_error": 0,
                "server_error": 1,
                "exception": 0
        },
        "violations": [
                {
                        "request_type": "INITIALIZE",
                        "description": "パス '/initialize' からレスポンスが返ってきませんでした",
                        "num": 1
                }
        ]
}

agent

run command agent. if your will build make build.

options:

• interval - polling queues interval
• pubsub - pubsub server URL
• benchmark - benchmark script file path
• param - parameters file path for the benchmark script
• host - webapp hostname
• port - webapp running port

example run command:

agent -pubsub=http://localhost:9000 -benchmark=./bin/app -param=./param.json -host=localhost -port=8080

Queue

Polling for the benchmark request queues, and dispatch a request to the benchmarker. The result benchmark send queue when finished benchmark.

Send response queue message:

Data: {
  // Benchmark result
}
Attributes: {
  "source_msg_id": // Request message queue ID,
  "team_id":       // Team ID from the request message queue,
  "created_at":    // Benchmark finished time,
},