Glow is an easy-to-use distributed computation system written in Go, an alternative to Hadoop Map Reduce, Spark, Flink, Samza, etc. Currently just started and not feature rich yet, but should be reliable to run most common cases.

go get



Examples are in this repo



Glow is providing a library to easily compute in parallel threads or distributed to clusters of machines.


go get
go get

One minute tutorial

Simple Start

Here is a simple full example:

package main

import (


func main() {

        "/etc/passwd", 3,
    ).Filter(func(line string) bool {
        return !strings.HasPrefix(line, "#")
    }).Map(func(line string, ch chan string) {
        for _, token := range strings.Split(line, ":") {
            ch <- token
    }).Map(func(key string) int {
        return 1
    }).Reduce(func(x int, y int) int {
        return x + y
    }).Map(func(x int) {
        println("count:", x)

Try it.


It will run the input text file, '/etc/passwd', in 3 go routines, filter/map/map, and then reduced to one number in one goroutine (not exactly one goroutine, but let's skip the details for now.) and print it out.

This is useful already, saving lots of idiomatic but repetitive code on channels, sync wait, etc.

However, there is one more thing!

Scale it out

We need to setup the cluster first. We do not need experts on Zookeeper/HDFS/Mesos/YARN etc. Just need to build or download one binary file.

Setup the cluster

  // fetch and install via go, or just download it from somewhere
  go get
  // start a master on one computer
  glow master
  // run one or more agents on computers
  glow agent --dir . --max.executors=16 --memory=2048 --master="localhost:8930" --port 8931

Start the driver program

To leap from one computer to clusters of computers, add this line to the import list:

    _ ""

This will "steroidize" the code to run in cluster mode!

./word_count -glow -glow.leader="localhost:8930"

Glow Hello World Execution Plan

Read More

  1. Wiki page:
  2. Mailing list:!forum/glow-user-discussion
  3. Examples:


Fork it, code it, and send pull requests. Better first discuss about the feature you want on the mailing list.!forum/glow-user-discussion