[ceph.git] / ceph / src / zstd / README.md

 __Zstandard__, or `zstd` as short version, is a fast lossless compression algorithm,
 targeting real-time compression scenarios at zlib-level and better compression ratios.

It is provided as an open-source BSD-licensed **C** library,
and a command line utility producing and decoding `.zst` and `.gz` files.
For other programming languages,
you can consult a list of known ports on [Zstandard homepage](http://www.zstd.net/#other-languages).

| dev branch status |
|-------------------|
| [![Build Status][travisDevBadge]][travisLink]   [![Build status][AppveyorDevBadge]][AppveyorLink]   [![Build status][CircleDevBadge]][CircleLink]

[travisDevBadge]: https://travis-ci.org/facebook/zstd.svg?branch=dev "Continuous Integration test suite"
[travisLink]: https://travis-ci.org/facebook/zstd
[AppveyorDevBadge]: https://ci.appveyor.com/api/projects/status/xt38wbdxjk5mrbem/branch/dev?svg=true "Windows test suite"
[AppveyorLink]: https://ci.appveyor.com/project/YannCollet/zstd-p0yf0
[CircleDevBadge]: https://circleci.com/gh/facebook/zstd/tree/dev.svg?style=shield "Short test suite"
[CircleLink]: https://circleci.com/gh/facebook/zstd


As a reference, several fast compression algorithms were tested and compared
on a server running Linux Debian (`Linux version 4.8.0-1-amd64`),
with a Core i7-6700K CPU @ 4.0GHz,
using [lzbench], an open-source in-memory benchmark by @inikep
compiled with GCC 6.3.0,
on the [Silesia compression corpus].

[lzbench]: https://github.com/inikep/lzbench
[Silesia compression corpus]: http://sun.aei.polsl.pl/~sdeor/index.php?page=silesia

| Compressor name         | Ratio | Compression| Decompress.|
| ---------------         | ------| -----------| ---------- |
| **zstd 1.1.3 -1**       | 2.877 |   430 MB/s |  1110 MB/s |
| zlib 1.2.8 -1           | 2.743 |   110 MB/s |   400 MB/s |
| brotli 0.5.2 -0         | 2.708 |   400 MB/s |   430 MB/s |
| quicklz 1.5.0 -1        | 2.238 |   550 MB/s |   710 MB/s |
| lzo1x 2.09 -1           | 2.108 |   650 MB/s |   830 MB/s |
| lz4 1.7.5               | 2.101 |   720 MB/s |  3600 MB/s |
| snappy 1.1.3            | 2.091 |   500 MB/s |  1650 MB/s |
| lzf 3.6 -1              | 2.077 |   400 MB/s |   860 MB/s |

[zlib]:http://www.zlib.net/
[LZ4]: http://www.lz4.org/

Zstd can also offer stronger compression ratios at the cost of compression speed.
Speed vs Compression trade-off is configurable by small increments. Decompression speed is preserved and remains roughly the same at all settings, a property shared by most LZ compression algorithms, such as [zlib] or lzma.

The following tests were run
on a server running Linux Debian (`Linux version 4.8.0-1-amd64`)
with a Core i7-6700K CPU @ 4.0GHz,
using [lzbench], an open-source in-memory benchmark by @inikep
compiled with GCC 6.3.0,
on the [Silesia compression corpus].

Compression Speed vs Ratio | Decompression Speed
---------------------------|--------------------
![Compression Speed vs Ratio](doc/images/Cspeed4.png "Compression Speed vs Ratio") | ![Decompression Speed](doc/images/Dspeed4.png "Decompression Speed")

Several algorithms can produce higher compression ratios, but at slower speeds, falling outside of the graph.
For a larger picture including very slow modes, [click on this link](doc/images/DCspeed5.png) .


### The case for Small Data compression

Previous charts provide results applicable to typical file and stream scenarios (several MB). Small data comes with different perspectives.

The smaller the amount of data to compress, the more difficult it is to compress. This problem is common to all compression algorithms, and reason is, compression algorithms learn from past data how to compress future data. But at the beginning of a new data set, there is no "past" to build upon.

To solve this situation, Zstd offers a __training mode__, which can be used to tune the algorithm for a selected type of data.
Training Zstandard is achieved by providing it with a few samples (one file per sample). The result of this training is stored in a file called "dictionary", which must be loaded before compression and decompression.
Using this dictionary, the compression ratio achievable on small data improves dramatically.

The following example uses the `github-users` [sample set](https://github.com/facebook/zstd/releases/tag/v1.1.3), created from [github public API](https://developer.github.com/v3/users/#get-all-users).
It consists of roughly 10K records weighing about 1KB each.

Compression Ratio | Compression Speed | Decompression Speed
------------------|-------------------|--------------------
![Compression Ratio](doc/images/dict-cr.png "Compression Ratio") | ![Compression Speed](doc/images/dict-cs.png "Compression Speed") | ![Decompression Speed](doc/images/dict-ds.png "Decompression Speed")


These compression gains are achieved while simultaneously providing _faster_ compression and decompression speeds.

Training works if there is some correlation in a family of small data samples. The more data-specific a dictionary is, the more efficient it is (there is no _universal dictionary_).
Hence, deploying one dictionary per type of data will provide the greatest benefits.
Dictionary gains are mostly effective in the first few KB. Then, the compression algorithm will gradually use previously decoded content to better compress the rest of the file.

#### Dictionary compression How To :

1) Create the dictionary

`zstd --train FullPathToTrainingSet/* -o dictionaryName`

2) Compress with dictionary

`zstd -D dictionaryName FILE`

3) Decompress with dictionary

`zstd -D dictionaryName --decompress FILE.zst`


### Build

Once you have the repository cloned, there are multiple ways provided to build Zstandard.

#### Makefile

If your system is compatible with a standard `make` (or `gmake`) binary generator,
you can simply run it at the root directory.
It will generate `zstd` within root directory.

Other available options include :
- `make install` : create and install zstd binary, library and man page
- `make test` : create and run `zstd` and test tools on local platform

#### cmake

A `cmake` project generator is provided within `build/cmake`.
It can generate Makefiles or other build scripts
to create `zstd` binary, and `libzstd` dynamic and static libraries.

#### Meson

A Meson project is provided within `contrib/meson`.

#### Visual Studio (Windows)

Going into `build` directory, you will find additional possibilities :
- Projects for Visual Studio 2005, 2008 and 2010
  + VS2010 project is compatible with VS2012, VS2013 and VS2015
- Automated build scripts for Visual compiler by @KrzysFR , in `build/VS_scripts`,
  which will build `zstd` cli and `libzstd` library without any need to open Visual Studio solution.


### Status

Zstandard is currently deployed within Facebook. It is used continuously to compress large amounts of data in multiple formats and use cases.
Zstandard is considered safe for production environments.

### License

Zstandard is dual-licensed under [BSD](LICENSE) and [GPLv2](COPYING).

### Contributing

The "dev" branch is the one where all contributions will be merged before reaching "master".
If you plan to propose a patch, please commit into the "dev" branch or its own feature branch.
Direct commit to "master" are not permitted.
For more information, please read [CONTRIBUTING](CONTRIBUTING.md).

### Miscellaneous

Zstd entropy stage is provided by [Huff0 and FSE, from Finite State Entropy library](https://github.com/Cyan4973/FiniteStateEntropy).
Commit	Line	Data
7c673cae FG	1	__Zstandard__, or `zstd` as short version, is a fast lossless compression algorithm,
	2	targeting real-time compression scenarios at zlib-level and better compression ratios.
	3
	4	It is provided as an open-source BSD-licensed C library,
11fdf7f2	5	and a command line utility producing and decoding `.zst` and `.gz` files.
7c673cae FG	6	For other programming languages,
	7	you can consult a list of known ports on [Zstandard homepage](http://www.zstd.net/#other-languages).
	8
11fdf7f2 TL	9	\| dev branch status \|
	10	\|-------------------\|
	11	\| [![Build Status][travisDevBadge]][travisLink] [![Build status][AppveyorDevBadge]][AppveyorLink] [![Build status][CircleDevBadge]][CircleLink]
7c673cae	12
11fdf7f2 TL	13	[travisDevBadge]: https://travis-ci.org/facebook/zstd.svg?branch=dev "Continuous Integration test suite"
	14	[travisLink]: https://travis-ci.org/facebook/zstd
	15	[AppveyorDevBadge]: https://ci.appveyor.com/api/projects/status/xt38wbdxjk5mrbem/branch/dev?svg=true "Windows test suite"
	16	[AppveyorLink]: https://ci.appveyor.com/project/YannCollet/zstd-p0yf0
	17	[CircleDevBadge]: https://circleci.com/gh/facebook/zstd/tree/dev.svg?style=shield "Short test suite"
	18	[CircleLink]: https://circleci.com/gh/facebook/zstd
	19
	20
	21	As a reference, several fast compression algorithms were tested and compared
	22	on a server running Linux Debian (`Linux version 4.8.0-1-amd64`),
	23	with a Core i7-6700K CPU @ 4.0GHz,
	24	using [lzbench], an open-source in-memory benchmark by @inikep
	25	compiled with GCC 6.3.0,
	26	on the [Silesia compression corpus].
7c673cae FG	27
	28	[lzbench]: https://github.com/inikep/lzbench
	29	[Silesia compression corpus]: http://sun.aei.polsl.pl/~sdeor/index.php?page=silesia
	30
11fdf7f2 TL	31	\| Compressor name \| Ratio \| Compression\| Decompress.\|
	32	\| --------------- \| ------\| -----------\| ---------- \|
	33	\| zstd 1.1.3 -1 \| 2.877 \| 430 MB/s \| 1110 MB/s \|
	34	\| zlib 1.2.8 -1 \| 2.743 \| 110 MB/s \| 400 MB/s \|
	35	\| brotli 0.5.2 -0 \| 2.708 \| 400 MB/s \| 430 MB/s \|
	36	\| quicklz 1.5.0 -1 \| 2.238 \| 550 MB/s \| 710 MB/s \|
	37	\| lzo1x 2.09 -1 \| 2.108 \| 650 MB/s \| 830 MB/s \|
	38	\| lz4 1.7.5 \| 2.101 \| 720 MB/s \| 3600 MB/s \|
	39	\| snappy 1.1.3 \| 2.091 \| 500 MB/s \| 1650 MB/s \|
	40	\| lzf 3.6 -1 \| 2.077 \| 400 MB/s \| 860 MB/s \|
7c673cae FG	41
	42	[zlib]:http://www.zlib.net/
	43	[LZ4]: http://www.lz4.org/
	44
	45	Zstd can also offer stronger compression ratios at the cost of compression speed.
	46	Speed vs Compression trade-off is configurable by small increments. Decompression speed is preserved and remains roughly the same at all settings, a property shared by most LZ compression algorithms, such as [zlib] or lzma.
	47
11fdf7f2 TL	48	The following tests were run
	49	on a server running Linux Debian (`Linux version 4.8.0-1-amd64`)
	50	with a Core i7-6700K CPU @ 4.0GHz,
	51	using [lzbench], an open-source in-memory benchmark by @inikep
	52	compiled with GCC 6.3.0,
	53	on the [Silesia compression corpus].
7c673cae FG	54
	55	Compression Speed vs Ratio \| Decompression Speed
	56	---------------------------\|--------------------
	57	![Compression Speed vs Ratio](doc/images/Cspeed4.png "Compression Speed vs Ratio") \| ![Decompression Speed](doc/images/Dspeed4.png "Decompression Speed")
	58
	59	Several algorithms can produce higher compression ratios, but at slower speeds, falling outside of the graph.
	60	For a larger picture including very slow modes, [click on this link](doc/images/DCspeed5.png) .
	61
	62
	63	### The case for Small Data compression
	64
11fdf7f2 TL	65	Previous charts provide results applicable to typical file and stream scenarios (several MB). Small data comes with different perspectives.
	66
	67	The smaller the amount of data to compress, the more difficult it is to compress. This problem is common to all compression algorithms, and reason is, compression algorithms learn from past data how to compress future data. But at the beginning of a new data set, there is no "past" to build upon.
7c673cae	68
11fdf7f2 TL	69	To solve this situation, Zstd offers a __training mode__, which can be used to tune the algorithm for a selected type of data.
	70	Training Zstandard is achieved by providing it with a few samples (one file per sample). The result of this training is stored in a file called "dictionary", which must be loaded before compression and decompression.
	71	Using this dictionary, the compression ratio achievable on small data improves dramatically.
7c673cae	72
11fdf7f2 TL	73	The following example uses the `github-users` [sample set](https://github.com/facebook/zstd/releases/tag/v1.1.3), created from [github public API](https://developer.github.com/v3/users/#get-all-users).
11fdf7f2 TL	74	It consists of roughly 10K records weighing about 1KB each.
7c673cae	75
11fdf7f2 TL	76	Compression Ratio \| Compression Speed \| Decompression Speed
	77	------------------\|-------------------\|--------------------
	78	![Compression Ratio](doc/images/dict-cr.png "Compression Ratio") \| ![Compression Speed](doc/images/dict-cs.png "Compression Speed") \| ![Decompression Speed](doc/images/dict-ds.png "Decompression Speed")
7c673cae	79
7c673cae	80
11fdf7f2 TL	81	These compression gains are achieved while simultaneously providing _faster_ compression and decompression speeds.
	82
	83	Training works if there is some correlation in a family of small data samples. The more data-specific a dictionary is, the more efficient it is (there is no _universal dictionary_).
	84	Hence, deploying one dictionary per type of data will provide the greatest benefits.
	85	Dictionary gains are mostly effective in the first few KB. Then, the compression algorithm will gradually use previously decoded content to better compress the rest of the file.
7c673cae FG	86
	87	#### Dictionary compression How To :
	88
	89	1) Create the dictionary
	90
	91	`zstd --train FullPathToTrainingSet/* -o dictionaryName`
	92
	93	2) Compress with dictionary
	94
11fdf7f2	95	`zstd -D dictionaryName FILE`
7c673cae FG	96
	97	3) Decompress with dictionary
	98
11fdf7f2 TL	99	`zstd -D dictionaryName --decompress FILE.zst`
11fdf7f2 TL	100
7c673cae FG	101
	102	### Build
	103
	104	Once you have the repository cloned, there are multiple ways provided to build Zstandard.
	105
	106	#### Makefile
	107
	108	If your system is compatible with a standard `make` (or `gmake`) binary generator,
	109	you can simply run it at the root directory.
	110	It will generate `zstd` within root directory.
	111
	112	Other available options include :
	113	- `make install` : create and install zstd binary, library and man page
	114	- `make test` : create and run `zstd` and test tools on local platform
	115
	116	#### cmake
	117
	118	A `cmake` project generator is provided within `build/cmake`.
	119	It can generate Makefiles or other build scripts
	120	to create `zstd` binary, and `libzstd` dynamic and static libraries.
	121
11fdf7f2 TL	122	#### Meson
	123
	124	A Meson project is provided within `contrib/meson`.
	125
	126	#### Visual Studio (Windows)
7c673cae FG	127
	128	Going into `build` directory, you will find additional possibilities :
	129	- Projects for Visual Studio 2005, 2008 and 2010
	130	+ VS2010 project is compatible with VS2012, VS2013 and VS2015
	131	- Automated build scripts for Visual compiler by @KrzysFR , in `build/VS_scripts`,
	132	which will build `zstd` cli and `libzstd` library without any need to open Visual Studio solution.
	133
	134
	135	### Status
	136
11fdf7f2	137	Zstandard is currently deployed within Facebook. It is used continuously to compress large amounts of data in multiple formats and use cases.
7c673cae FG	138	Zstandard is considered safe for production environments.
	139
	140	### License
	141
11fdf7f2	142	Zstandard is dual-licensed under [BSD](LICENSE) and [GPLv2](COPYING).
7c673cae FG	143
	144	### Contributing
	145
	146	The "dev" branch is the one where all contributions will be merged before reaching "master".
	147	If you plan to propose a patch, please commit into the "dev" branch or its own feature branch.
	148	Direct commit to "master" are not permitted.
	149	For more information, please read [CONTRIBUTING](CONTRIBUTING.md).
	150
	151	### Miscellaneous
	152
	153	Zstd entropy stage is provided by [Huff0 and FSE, from Finite State Entropy library](https://github.com/Cyan4973/FiniteStateEntropy).