Bench

Bench is a golang package to help speed up benchmarking projects. It can be used to do benchmark your database, queue or rest server.

Features

Bench can call a function repeatedly or at a uniform rate with the specificied concurrency. For each call, the called function can captures metrics, which are summarized at the end of the run.

Captured metrics include:

• Timers. Use it to capture latency of functions, I/O bound calls.
• Counters.

A sample run will spew similar output

$ go run simple.go
Rate: 10 calls/sec, Duration: 5.00s, Concurrency: 2, Total runs: 50
  >>Timer: Latency
    5.0th percentile: 0.59ms
    50.0th percentile: 0.67ms
    99.9th percentile: 0.82ms
    100.0th percentile: 0.82ms
  >>Counter: Call Counter
    Value: 50

Bench provides a Context in each function call. Use it to

• Pick different between urls for each call
• Add it as a custom header for your Rest call
• Generate different DB data
• Store run specific values. Example connection pools.

Usage

Bench is easy to integrate with your code. This code benchmarks a http server with a concurrency of 10 for 10 seconds by calling the benchmarkMe function continuously.

func benchmarkMe(c *bench.Context) {
	start := time.Now()
	resp, err := http.Get("http://www.flipkart.com")
	if err != nil {
		fmt.Printf("Error %s", err)
	}
	defer resp.Body.Close()
	c.RecordTime(time.Since(start))
}

func main() {
	b := bench.NewBench(10, time.Second*10, 0, benchmarkMe)
	b.Run()
	fmt.Printf("%s", b)

Here, the same function is called at a rate of 100 per second.

func main() {
	b := bench.NewBench(10, time.Second*10, 100, benchmarkMe)
	b.Run()
	fmt.Printf("%s", b)

Upcoming features

Two features are in the offing.

1. Support for an initial calibration period during which metrics are ignored.
2. Production quality http benchmarks.