dosse91 f432f84e3a Added more troubleshooting

2017-06-27 16:04:19 +02:00

11 KiB

Raw Blame History

HTML5 Speedtest

by Federico Dossena
Version 4.2.7, June 22 2017
https://github.com/adolfintel/speedtest/

Introduction

In this document, we will introduce an XHR based HTML5 Speedtest and see how to use it. This test measures download speed, upload speed, ping and jitter.

First of all, the requirements to run this test:

The browser have to support XHR Level 2 and Web Workers and Javascript must be enabled.
- Internet Explorer 11
- Microsoft Edge 12+
- Mozilla Firefox 12+
- Google Chrome / Chromium 31+
- Apple Safari 7.1+
- Opera 18+
Client side, the test can use up to 500 megabytes of RAM
Server side, you'll need a fast connection (at least 100 Mbps recommended), and the web server must accept large POST requests (up to 20 megabytes). Apache2 and PHP are recommended, but not mandatory.

If this looks good, let's proceed and see how to use the test.

Quick installation videos

Installation

To install the test on your server, upload the following files:

speedtest_worker.min.js
garbage.php
getIP.php
empty.php

You may also want to upload one of the examples to test it.
Later we'll see how to use the test without PHP.

Important: keep all the files together; all paths are relative to the js file

Usage

To run the test, you need to do 3 things:

Create the worker
Write some code that handles the responses coming from the worker
Start the test

Creating the worker

var w = new Worker("speedtest_worker.min.js")

Important: use the minified version, it's smaller!

Response handler

First, we set up a timer that fetches the status of the worker continuously:

var timer = setInterval(function () {
  w.postMessage('status')
}, 100)

Then we write a response handler that receives the status and updates the page. Later we'll see the details of the format of the response.

w.onmessage = function (event) {
  var data = event.data.split(';')
  var testState = data[0]
  var dlStatus = data[1]
  var ulStatus = data[2]
  var pingStatus = data[3]
  var jitterStatus = data[5]
  var clientIp = data[4]
  if (testState >= 4) {
    clearInterval(timer) // test is finished or aborted
  }
  // .. update your page here ..
}

Response format

The response from the worker is composed of values separated by ; (semicolon) in this format:

testState;dlStatus;ulStatus;pingStatus;clientIp;jitterStatus

testState is an integer 0-5
- 0 = Test starting
- 1 = Download test in progress
- 2 = Ping + Jitter test in progress
- 3 = Upload test in progress
- 4 = Test finished
- 5 = Test aborted
dlStatus is either
- Empty string (not started or aborted)
- Download speed in Megabit/s as a number with 2 decimals
- The string "Fail" (test failed)
ulStatus is either
- Empty string (not started or aborted)
- Upload speed in Megabit/s as a number with 2 decimals
- The string "Fail" (test failed)
pingStatus is either
- Empty string (not started or aborted)
- Estimated ping in milliseconds as a number with 2 decimals
- The string "Fail" (test failed)
clientIp is either
- Empty string (not fetched yet or failed)
- The client's IP address as a string
jitterStatus is either
- Empty string (not started or aborted)
- Estimated jitter in milliseconds as a number with 2 decimals (lower = stable connection)
- The string "Fail" (test failed)

Starting the test

To start the test with the default settings, which is usually the best choice, send the start command to the worker:

w.postMessage('start')

If you want, you can change these settings and pass them to the worker as JSON when you start it, like this:

w.postMessage('start {"param1": "value1", "param2": "value2", ...}')

Test parameters

time_dl: How long the download test should be in seconds
- Default: 15
- Recommended: >=5
time_ul: How long the upload test should be in seconds
- Default: 15
- Recommended: >=10
count_ping: How many pings to perform in the ping test
- Default: 35
- Recommended: >=20
url_dl: path to garbage.php or a large file to use for the download test
- Default: garbage.php
- Important: path is relative to js file
url_ul: path to an empty file or empty.php to use for the upload test
- Default: empty.php
- Important: path is relative to js file
url_ping: path to an empty file or empty.php to use for the ping test
- Default: empty.php
- Important: path is relative to js file
url_getIp: path to getIP.php or replacement
- Default: getIP.php
- Important: path is relative to js file

Advanced test parameters

enable_quirks: enables browser-specific optimizations. These optimizations override some of the default settings. They do not override settings that are explicitly set.
- Default: true
garbagePhp_chunkSize: size of chunks sent by garbage.php in megabytes
- Default: 20
- Recommended: >=10
- Default override: 5 on Safari if enable_quirks is true
xhr_dlMultistream: how many streams should be opened for the download test
- Default: 10
- Recommended: >=3
- Default override: 3 on Edge if enable_quirks is true
- Default override: 5 on Chromium-based if enable_quirks is true
xhr_ulMultistream: how many streams should be opened for the upload test
- Default: 3
- Recommended: >=1
- Default override: 1 on Firefox if enable_quirks is true
xhr_ignoreErrors: how to react to errors in download/upload streams and the ping test
- 0: Fail test on error (behaviour of previous versions of this test)
- 1: Restart a stream/ping when it fails
- 2: Ignore all errors
- Default: 1
- Recommended: 1
time_dlGraceTime: How long to wait (in seconds) before actually measuring the download speed. This is a good idea because we want to wait for the TCP window to be at its maximum (or close to it)
- Default: 1.5
- Recommended: >=0
time_ulGraceTime: How long to wait (in seconds) before actually measuring the upload speed. This is a good idea because we want to wait for the buffers to be full (avoids the peak at the beginning of the test)
- Default: 3
- Recommended: >=1
overheadCompensationFactor: compensation for HTTP and network overhead. Default value assumes typical MTUs used over the Internet. You might want to change this if you're using this in your internal network with different MTUs, or if you're using IPv6 instead of IPv4.
- Default: 1.13359567567567567568 (1048576/925000) assumes HTTP+TCP+IPv4+ETH with typical MTUs used over the Internet
- 1.0513: HTTP+TCP+IPv6+ETH, over the Internet (empirically tested, not calculated)
- 1.0369: Alternative value for HTTP+TCP+IPv4+ETH, over the Internet (empirically tested, not calculated)
- 1460 / 1514: TCP+IPv4+ETH, ignoring HTTP overhead
- 1440 / 1514: TCP+IPv6+ETH, ignoring HTTP overhead
- 1: ignore overheads. This measures the speed at which you actually download and upload files

Aborting the test prematurely

The test can be aborted at any time by sending an abort command to the worker:

w.postMessage('abort')

This will terminate all network activity and stop the worker.

Important: do not simply kill the worker while it's running, as it may leave pending XHR requests!

Troubleshooting

These are the most common issues reported by users, and how to fix them

Download test gives very low result

Are garbage.php and empty.php (or your replacements) reachable?
Press F12, select network and start the test. Do you see errors? (cancelled requests are not errors)
If a small download starts, open it in a text editor. Does it say it's missing openssl_random_pseudo_bytes()? In this case, install OpenSSL (this is usually included when you install Apache and PHP on most distros).

Upload test is inaccurate, and I see lag spikes

Check your server's maximum POST size, make sure it's at least 20Mbytes, possibly more

All tests are wrong, give extremely high results, browser lags/crashes, ...

You're running the test on localhost, therefore it is trying to measure the speed of your loopback interface. The test is meant to be run over an Internet connection, from a different machine.

Ping test shows double the actual ping

Make sure your server is sending the Connection:keep-alive header

Using the test without PHP

If your server does not support PHP, or you're using something newer like Node.js, you can still use this test by replacing garbage.php, empty.php and getIP.php with equivalents.

Replacements

Replacement for `garbage.php`

A replacement for garbage.php must generate incompressible garbage data.

A large file (10-100 Mbytes) is a possible replacement. You can get one here.

If you're using Node.js or some other server, your replacement should accept the ckSize parameter (via GET) which tells it how many megabytes of garbage to generate. It is important here to turn off compression, and generate incompressible data. A symlink to /dev/urandom is also ok.

Replacement for `empty.php`

Your replacement must simply respond with a HTTP code 200 and send nothing else. You may want to send additional headers to disable caching. The test assumes that Connection:keep-alive is sent by the server.

Replacement for `getIP.php`

Your replacement must simply respond with the client's IP as plaintext. Nothing fancy.

JS

You need to start the test with your replacements like this:

w.postMessage('start {"url_dl": "newGarbageURL", "url_ul": "newEmptyURL", "url_ping": "newEmptyURL", "url_getIp": "newIpURL"}')

Known bugs and limitations

The ping/jitter test is measured by seeing how long it takes for an empty XHR to complete. It is not an acutal ICMP ping
Chrome: high CPU usage from XHR requests with very fast connections (like gigabit). For this reason, the test may report inaccurate results if your CPU is too slow. (Does not affect most computers)
IE11: the upload test is not precise on very fast connections
IE11: the upload test may not work over HTTPS
Safari: works, but needs more testing and tweaking for very fast connections

Making changes

Since this is an open source project, you can modify it.

To make changes to the speedtest itself, edit speedtest_worker.js

To create the minified version, use UglifyJS like this:

uglifyjs -c --screw-ie8 speedtest_worker.js > speedtest_worker.min.js

Pull requests are much appreciated. If you don't use github (or git), simply contact me at dosse91@paranoici.org.

Important: please add your name to modified versions to distinguish them from the main project.

License

This software is under the GNU LGPL license, Version 3 or newer.

To put it short: you are free to use, study, modify, and redistribute this software and modified versions of it, for free or for money. You can also use it in proprietary software but all changes to this software must remain under the same GNU LGPL license.

11 KiB Raw Blame History