nareix

Benchmarking Nginx with Go

There are a lot of ways to serve a Go HTTP application. The best choices depend on each use case. Currently nginx looks to be the standard web server for every new project even though there are other great web servers as well. However, how much is the overhead of serving a Go application behind an nginx server? Do we need some nginx features (vhosts, load balancing, cache, etc) or can you serve directly from Go? If you need nginx, what is the fastest connection mechanism? This are the kind of questions I'm intended to answer here. The purpose of this benchmark is not to tell that Go is faster or slower than nginx. That would be stupid.

So, these are the different settings we are going to compare:

Go HTTP standalone (as the control group)
Nginx proxy to Go HTTP
Nginx fastcgi to Go TCP FastCGI
Nginx fastcgi to Go Unix Socket FastCGI

Better local require() paths for Node.js

Problem

When the directory structure of your Node.js application (not library!) has some depth, you end up with a lot of annoying relative paths in your require calls like:

const Article = require('../../../../app/models/article');

Those suck for maintenance and they're ugly.

	var mongoose = require('mongoose');
	mongoose.connect('mongo://localhost/test');
	var conn = mongoose.connection;

	var users = conn.collection('users');
	var channels = conn.collection('channels');
	var articles = conn.collection('articles');

	var insertUsers = Q.nfbind(users.insert.bind(users));
	var insertChannels = Q.nfbind(channels.insert.bind(channels));

	Latency Comparison Numbers (~2012)
	----------------------------------
	L1 cache reference 0.5 ns
	Branch mispredict 5 ns
	L2 cache reference 7 ns 14x L1 cache
	Mutex lock/unlock 25 ns
	Main memory reference 100 ns 20x L2 cache, 200x L1 cache
	Compress 1K bytes with Zippy 3,000 ns 3 us
	Send 1K bytes over 1 Gbps network 10,000 ns 10 us
	Read 4K randomly from SSD* 150,000 ns 150 us ~1GB/sec SSD

	all:
	bison -d newick.y
	flex newick.l
	gcc -Wall -O3 newick.tab.c lex.yy.c
	echo "(A,B,(C,D)E)F;" \| ./a.out

	/*
	* linux 2.6.37-3.x.x x86_64, ~100 LOC
	* gcc-4.6 -O2 semtex.c && ./a.out
	* 2010 [email protected], salut!
	*
	* update may 2013:
	* seems like centos 2.6.32 backported the perf bug, lol.
	* jewgold to 115T6jzGrVMgQ2Nt1Wnua7Ch1EuL9WXT2g if you insist.
	*/

	package com.solarfighter.activities;

	import java.io.IOException;
	import java.net.InetAddress;

	import javax.jmdns.JmDNS;
	import javax.jmdns.ServiceEvent;
	import javax.jmdns.ServiceListener;

	import android.content.Context;

	import os
	import sys
	import re
	import hashlib
	import csv
	import time
	import locale
	import getopt

	#include <pebble.h>
	#include <time.h>

	static const int FACE_DISC_RADIUS = 70;
	static const int CENTER_DISC_RADIUS = 20;
	static const int HOUR_DISC_RADIUS = 10;
	static const int MINUTE_DISC_RADIUS = 5;

	static const int HOUR_ROT_RADIUS = 40;
	static const int MINUTE_ROT_RADIUS = 60;

	package main

	import (
	"fmt"
	"syscall"
	)

	type DiskStatus struct {
	All uint64 `json:"all"`
	Used uint64 `json:"used"`

nareix

Benchmarking Nginx with Go

Better local require() paths for Node.js

Problem

Possible solutions