SZ3: A Modular Error-bounded Lossy Compression Framework for Scientific Datasets

Major Authors: Sheng Di, Kai Zhao, Xin Liang
Supervisor: Franck Cappello
Other Contributors: Robert Underwood, Sihuan Li, Ali M. Gok

Citations

Kindly note: If you mention SZ in your paper, the most appropriate citation is including these three references (TBD22, ICDE21, Bigdata18), because they cover the design and implementation of the latest version of SZ.

SZ3 Framework: Xin Liang, Kai Zhao, Sheng Di, Sihuan Li, Robert Underwood, Ali M Gok, Jiannan Tian, Junjing Deng, Jon C Calhoun, Dingwen Tao, Zizhong Chen, and Franck Cappello. "SZ3: A modular framework for composing prediction-based error-bounded lossy compressors", IEEE Transactions on Big Data (TBD 22).
SZ3 Algorithm: Kai Zhao, Sheng Di, Maxim Dmitriev, Thierry-Laurent D. Tonellot, Zizhong Chen, and Franck Cappello. "Optimizing Error-Bounded Lossy Compression for Scientiﬁc Data by Dynamic Spline Interpolation" , Proceeding of the 37th IEEE International Conference on Data Engineering (ICDE 21), Chania, Crete, Greece, Apr 19 - 22, 2021.
SZauto: Kai Zhao, Sheng Di, Xin Liang, Sihuan Li, Dingwen Tao, Zizhong Chen, and Franck Cappello. "Significantly Improving Lossy Compression for HPC Datasets with Second-Order Prediction and Parameter Optimization" , Proceedings of the 29th International Symposium on High-Performance Parallel and Distributed Computing (HPDC 20), Stockholm, Sweden, 2020. (code: https://github.com/szcompressor/SZauto/)
SZ 2.0+: Xin Liang, Sheng Di, Dingwen Tao, Zizhong Chen, Franck Cappello, "Error-Controlled Lossy Compression Optimized for High Compression Ratios of Scientific Datasets" , in IEEE International Conference on Big Data (Bigdata 2018), Seattle, WA, USA, 2018.
SZ 1.4.0-1.4.13: Dingwen Tao, Sheng Di, Franck Cappello. "Significantly Improving Lossy Compression for Scientific Data Sets Based on Multidimensional Prediction and Error-Controlled Quantization" , in IEEE International Parallel and Distributed Processing Symposium (IPDPS 2017), Orlando, Florida, USA, 2017.
SZ 0.1-1.0: Sheng Di, Franck Cappello. "Fast Error-bounded Lossy HPC Data Compression with SZ", in IEEE International Parallel and Distributed Processing Symposium (IPDPS 2016), Chicago, IL, USA, 2016.
Point-wise relative error bound mode (i.e., PW_REL): Xin Liang, Sheng Di, Dingwen Tao, Zizhong Chen, Franck Cappello, "An Efficient Transformation Scheme for Lossy Data Compression with Point-wise Relative Error Bound" , in IEEE International Conference on Clustering Computing (CLUSTER 2018), Belfast, UK, 2018. (Best Paper)

3rd party libraries/tools

Zstandard (https://facebook.github.io/zstd/). Zstandard v1.4.5 is included and will be used if libzstd can not be found by pkg-config.

Installation

mkdir build && cd build
cmake -DCMAKE_INSTALL_PREFIX:PATH=[INSTALL_DIR] ..
make
make install

Then, you'll find all the executables in [INSTALL_DIR]/bin and header files in [INSTALL_DIR]/include

Testing Examples

You can use the executable 'sz3' command to do the compression/decompression.

SZ3 simplifies command line arguments in the previous version. If you are a new user, please follow the instructions given by the executable.

Backward Compatibility with SZ2

For backward compatibility, most of the SZ2 command line parameters are supported in SZ3. Exceptions are listed below. Scripts without parameters below should work fine by replacing SZ2 with SZ3.

Parameter	Explanation	SZ3 roadmap
-c	Config file	SZ3 has different config format with SZ2
-p	Print configuration info	Will be supported soon
-T	Tucker Tensor Decomposition	Will be supported later
-P	Point-wise relative error bound	Will be supported later

API

SZ3 C++ API

Located in 'include/SZ3/api/sz.hpp'.
Requiring a modern C++ compiler.
Different with SZ2 API.

SZ3 C API

Located in 'tools/sz3c/include/sz3c.h'
Compatible with SZ2 API

Python API

Located in 'tools/pysz/pysz.py'
Test file provided ('tools/pysz/test.py')
Compatible with both SZ3 and SZ2
Requiring SZ2/3 dynamic library

H5Z-SZ3

Use examples/print_h5repack_args.c to construct the cd_values parameters based on the specified error configuration.

Example:

Compression:

[sdi@localhost build]$ h5repack -f UD=32024,0,5,0,981668463,0,0,0 -i ~/Data/CESM-ATM-tylor/1800x3600/CLDLOW_1_1800_3600.dat.h5 -o ~/Data/CESM-ATM-tylor/1800x3600/CLDLOW_1_1800_3600.dat.sz3.h5

Decompression:

[sdi@localhost build]$ h5repack -f NONE -i ~/Data/CESM-ATM-tylor/1800x3600/CLDLOW_1_1800_3600.dat.sz3.h5 -o ~/Data/CESM-ATM-tylor/1800x3600/CLDLOW_1_1800_3600.dat.sz3.out.h5

Alternatively, the error bound information can also be given through sz3.config (when there are no cd_values for h5repack)

Example (You need to put sz3.config in the current local directory so that it will read sz3.config to get error bounds):

[sdi@localhost build]$ h5repack -f UD=32024,0 -i ~/Data/CESM-ATM-tylor/1800x3600/CLDLOW_1_1800_3600.dat.h5 -o ~/Data/CESM-ATM-tylor/1800x3600/CLDLOW_1_1800_3600.dat.sz3.h5

Version history

Version New features

SZ 3.0.0 SZ3 is the C++ version of SZ with modular and composable design.
SZ 3.0.1 Improve the build process.
SZ 3.1.0 The default algorithm is now interpolation+Lorenzo.
SZ 3.1.1 Add OpenMP support. Works for all algorithms. Please enable it using the config file.
SZ 3.1.2 Support configuration file (INI format). Example can be found in 'tools/sz3/sz3.config'.
SZ 3.1.3 Support more error control mode: PSNR, L2Norm, ABS_AND_REL, ABS_OR_REL. Support INT32 and INT64 datatype.
SZ 3.1.4 Support running on Windows. Please refer to szcompressor#5 (comment) for instructions.
SZ 3.1.5 Support HDF5 by H5Z-SZ3. Please add "-DBUILD_H5Z_FILTER=ON" to enable this function for CMake.
SZ 3.1.6 Support C API and Python API.
SZ 3.1.7 Initial MDZ(https://github.com/szcompressor/SZ3/tree/master/tools/mdz) support.
SZ 3.1.8 namespace changed from SZ to SZ3. H5Z-SZ3 supports configuration file now.

robertu94/SZ3