Brotli Compression Algorithm Motivation

Total Page：16

File Type：pdf, Size：1020Kb

Brotli Compression Algorithm motivation Main differences with deflate Areas ● entropy code reuse ● window size ● context modeling ● literal count coding ● distance cache ● match lengths ● block end codes Entropy code reuse Entropy code requires some bytes to be described. Deflate: the whole Brotli: a block may entropy code has to re-use past entropy be described before codes (within the the data. same metablock). Window size Deflate: deflate can Brotli: brotli can refer remember past 32 back to past 16 MB. kB. TBD: freeze a resource sweet spot for the specification, perhaps just 4 MB window. Context modeling Deflate: each literal Brotli: literal bytes byte is coded can have an entropy independently. code that depends on one, two or three last bytes. Literal count coding Deflate: every symbol Brotli: the number of pure has a probability for literals is codified, next literals, backward copy the literals with a literal length, or end code. only entropy code. These are entropy coded individually for every symbol. Distance cache Deflate: each distance Brotli: backward codified separately. references can be described in the form of past four distances ...abcdeX12345... ...abcdeY12345... Match lengths Deflate: match lengths of Brotli: match lengths of 3–258. Match lengths 2–enough. Match lengths codified in the same codified in joint entropy entropy code as literals. code with literal lengths. Block end code Deflate: each symbol can Brotli: block length is be the block end code, codified in the beginning stealing some efficiency of the block. from the entropy when unique symbol count is small..

Copy Link

Recommended publications

Data Compression: Dictionary-Based Coding 2 / 37 Dictionary-Based Coding Dictionary-Based Coding
Dictionary-based Coding already coded not yet coded search buffer look-ahead buffer cursor (N symbols) (L symbols) We know the past but cannot control it. We control the future but... Last Lecture Last Lecture: Predictive Lossless Coding Predictive Lossless Coding Simple and effective way to exploit dependencies between neighboring symbols / samples Optimal predictor: Conditional mean (requires storage of large tables) Affine and Linear Prediction Simple structure, low-complex implementation possible Optimal prediction parameters are given by solution of Yule-Walker equations Works very well for real signals (e.g., audio, images, ...) Efficient Lossless Coding for Real-World Signals Affine/linear prediction (often: block-adaptive choice of prediction parameters) Entropy coding of prediction errors (e.g., arithmetic coding) Using marginal pmf often already yields good results Can be improved by using conditional pmfs (with simple conditions) Heiko Schwarz (Freie Universität Berlin) — Data Compression: Dictionary-based Coding 2 / 37 Dictionary-based Coding Dictionary-Based Coding Coding of Text Files Very high amount of dependencies Affine prediction does not work (requires linear dependencies) Higher-order conditional coding should work well, but is way to complex (memory) Alternative: Do not code single characters, but words or phrases Example: English Texts Oxford English Dictionary lists less than 230 000 words (including obsolete words) On average, a word contains about 6 characters Average codeword length per character would be limited by 1
[Show full text]
Package 'Brotli'
Package ‘brotli’ May 13, 2018 Type Package Title A Compression Format Optimized for the Web Version 1.2 Description A lossless compressed data format that uses a combination of the LZ77 algorithm and Huffman coding. Brotli is similar in speed to deflate (gzip) but offers more dense compression. License MIT + file LICENSE URL https://tools.ietf.org/html/rfc7932 (spec) https://github.com/google/brotli#readme (upstream) http://github.com/jeroen/brotli#read (devel) BugReports http://github.com/jeroen/brotli/issues VignetteBuilder knitr, R.rsp Suggests spelling, knitr, R.rsp, microbenchmark, rmarkdown, ggplot2 RoxygenNote 6.0.1 Language en-US NeedsCompilation yes Author Jeroen Ooms [aut, cre] (<https://orcid.org/0000-0002-4035-0289>), Google, Inc [aut, cph] (Brotli C++ library) Maintainer Jeroen Ooms <[email protected]> Repository CRAN Date/Publication 2018-05-13 20:31:43 UTC R topics documented: brotli . .2 Index 4 1 2 brotli brotli Brotli Compression Description Brotli is a compression algorithm optimized for the web, in particular small text documents. Usage brotli_compress(buf, quality = 11, window = 22) brotli_decompress(buf) Arguments buf raw vector with data to compress/decompress quality value between 0 and 11 window log of window size Details Brotli decompression is at least as fast as for gzip while significantly improving the compression ratio. The price we pay is that compression is much slower than gzip. Brotli is therefore most effective for serving static content such as fonts and html pages. For binary (non-text) data, the compression ratio of Brotli usually does not beat bz2 or xz (lzma), however decompression for these algorithms is too slow for browsers in e.g.
[Show full text]
The Basic Principles of Data Compression
The Basic Principles of Data Compression Author: Conrad Chung, 2BrightSparks Introduction Internet users who download or upload files from/to the web, or use email to send or receive attachments will most likely have encountered files in compressed format. In this topic we will cover how compression works, the advantages and disadvantages of compression, as well as types of compression. What is Compression? Compression is the process of encoding data more efficiently to achieve a reduction in file size. One type of compression available is referred to as lossless compression. This means the compressed file will be restored exactly to its original state with no loss of data during the decompression process. This is essential to data compression as the file would be corrupted and unusable should data be lost. Another compression category which will not be covered in this article is “lossy” compression often used in multimedia files for music and images and where data is discarded. Lossless compression algorithms use statistic modeling techniques to reduce repetitive information in a file. Some of the methods may include removal of spacing characters, representing a string of repeated characters with a single character or replacing recurring characters with smaller bit sequences. Advantages/Disadvantages of Compression Compression of files offer many advantages. When compressed, the quantity of bits used to store the information is reduced. Files that are smaller in size will result in shorter transmission times when they are transferred on the Internet. Compressed files also take up less storage space. File compression can zip up several small files into a single file for more convenient email transmission.
[Show full text]
A Novel Coding Architecture for Multi-Line Lidar Point Clouds
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 1 A Novel Coding Architecture for Multi-Line LiDAR Point Clouds Based on Clustering and Convolutional LSTM Network Xuebin Sun , Sukai Wang , Graduate Student Member, IEEE, and Ming Liu , Senior Member, IEEE Abstract— Light detection and ranging (LiDAR) plays an preservation of historical relics, 3D sensing for smart city, indispensable role in autonomous driving technologies, such as well as autonomous driving. Especially for autonomous as localization, map building, navigation and object avoidance. driving systems, LiDAR sensors play an indispensable role However, due to the vast amount of data, transmission and storage could become an important bottleneck. In this article, in a large number of key techniques, such as simultaneous we propose a novel compression architecture for multi-line localization and mapping (SLAM) [1], path planning [2], LiDAR point cloud sequences based on clustering and convolu- obstacle avoidance [3], and navigation. A point cloud consists tional long short-term memory (LSTM) networks. LiDAR point of a set of individual 3D points, in accordance with one or clouds are structured, which provides an opportunity to convert more attributes (color, reﬂectance, surface normal, etc). For the 3D data to 2D array, represented as range images. Thus, we cast the 3D point clouds compression as a range image instance, the Velodyne HDL-64E LiDAR sensor generates a sequence compression problem. Inspired by the high efﬁciency point cloud of up to 2.2 billion points per second, with a video coding (HEVC) algorithm, we design a novel compression range of up to 120 m.
[Show full text]
Compresso: Efficient Compression of Segmentation Data for Connectomics
Compresso: Efficient Compression of Segmentation Data For Connectomics Brian Matejek, Daniel Haehn, Fritz Lekschas, Michael Mitzenmacher, Hanspeter Pfister Harvard University, Cambridge, MA 02138, USA bmatejek,haehn,lekschas,michaelm,[email protected] Abstract. Recent advances in segmentation methods for connectomics and biomedical imaging produce very large datasets with labels that assign object classes to image pixels. The resulting label volumes are bigger than the raw image data and need compression for efficient stor- age and transfer. General-purpose compression methods are less effective because the label data consists of large low-frequency regions with struc- tured boundaries unlike natural image data. We present Compresso, a new compression scheme for label data that outperforms existing ap- proaches by using a sliding window to exploit redundancy across border regions in 2D and 3D. We compare our method to existing compression schemes and provide a detailed evaluation on eleven biomedical and im- age segmentation datasets. Our method provides a factor of 600-2200x compression for label volumes, with running times suitable for practice. Keywords: compression, encoding, segmentation, connectomics 1 Introduction Connectomics|reconstructing the wiring diagram of a mammalian brain at nanometer resolution|results in datasets at the scale of petabytes [21,8]. Ma- chine learning methods find cell membranes and create cell body labelings for every neuron [18,12,14] (Fig. 1). These segmentations are stored as label volumes that are typically encoded in 32 bits or 64 bits per voxel to support labeling of millions of different nerve cells (neurons). Storing such data is expensive and transferring the data is slow. To cut costs and delays, we need compression methods to reduce data sizes.
[Show full text]
Data Latency and Compression
Data Latency and Compression Joseph M. Steim1, Edelvays N. Spassov2 1Quanterra Inc., 2Kinemetrics, Inc. Abstract DIscussion More Compression Because of interest in the capability of digital seismic data systems to provide low-latency data for “Early Compression and Latency related to Origin Queuing[5] Lossless Lempel-Ziv (LZ) compression methods (e.g. gzip, 7zip, and more recently “brotli” introduced by Warning” applications, we have examined the effect of data compression on the ability of systems to Plausible data compression methods reduce the volume (number of bits) required to represent each data Google for web-page compression) are often used for lossless compression of binary and text files. How deliver data with low latency, and on the efficiency of data storage and telemetry systems. Quanterra Q330 sample. Compression of data therefore always reduces the minimum required bandwidth (bits/sec) for does the best example of modern LZ methods compare with FDSN Level 1 and 2 for waveform data? systems are widely used in telemetered networks, and are considered in particular. transmission compared with transmission of uncompressed data. The volume required to transmit Segments of data representing up to one month of various levels of compressibility were compressed using compressed data is variable, depending on characteristics of the data. Higher levels of compression require Levels 1,2 and 3, and brotli using “quality” 1 and 6. The compression ratio (relative to the original 32 bits) Q330 data acquisition systems transmit data compressed in Steim2-format packets. Some studies have less minimum bandwidth and volume, and maximize the available unused (“free”) bandwidth for shows that in all cases the compression using L1,L2, or L3 is significantly greater than brotli.
[Show full text]
Parquet Data Format Performance
Parquet data format performance Jim Pivarski Princeton University { DIANA-HEP February 21, 2018 1 / 22 What is Parquet? 1974 HBOOK tabular rowwise FORTRAN first ntuples in HEP 1983 ZEBRA hierarchical rowwise FORTRAN event records in HEP 1989 PAW CWN tabular columnar FORTRAN faster ntuples in HEP 1995 ROOT hierarchical columnar C++ object persistence in HEP 2001 ProtoBuf hierarchical rowwise many Google's RPC protocol 2002 MonetDB tabular columnar database “first” columnar database 2005 C-Store tabular columnar database also early, became HP's Vertica 2007 Thrift hierarchical rowwise many Facebook's RPC protocol 2009 Avro hierarchical rowwise many Hadoop's object permanance and interchange format 2010 Dremel hierarchical columnar C++, Java Google's nested-object database (closed source), became BigQuery 2013 Parquet hierarchical columnar many open source object persistence, based on Google's Dremel paper 2016 Arrow hierarchical columnar many shared-memory object exchange 2 / 22 What is Parquet? 1974 HBOOK tabular rowwise FORTRAN first ntuples in HEP 1983 ZEBRA hierarchical rowwise FORTRAN event records in HEP 1989 PAW CWN tabular columnar FORTRAN faster ntuples in HEP 1995 ROOT hierarchical columnar C++ object persistence in HEP 2001 ProtoBuf hierarchical rowwise many Google's RPC protocol 2002 MonetDB tabular columnar database “first” columnar database 2005 C-Store tabular columnar database also early, became HP's Vertica 2007 Thrift hierarchical rowwise many Facebook's RPC protocol 2009 Avro hierarchical rowwise many Hadoop's object permanance and interchange format 2010 Dremel hierarchical columnar C++, Java Google's nested-object database (closed source), became BigQuery 2013 Parquet hierarchical columnar many open source object persistence, based on Google's Dremel paper 2016 Arrow hierarchical columnar many shared-memory object exchange 2 / 22 Developed independently to do the same thing Google Dremel authors claimed to be unaware of any precedents, so this is an example of convergent evolution.
[Show full text]
The Deep Learning Solutions on Lossless Compression Methods for Alleviating Data Load on Iot Nodes in Smart Cities
sensors Article The Deep Learning Solutions on Lossless Compression Methods for Alleviating Data Load on IoT Nodes in Smart Cities Ammar Nasif *, Zulaiha Ali Othman and Nor Samsiah Sani Center for Artificial Intelligence Technology (CAIT), Faculty of Information Science & Technology, University Kebangsaan Malaysia, Bangi 43600, Malaysia; [email protected] (Z.A.O.); [email protected] (N.S.S.) * Correspondence: [email protected] Abstract: Networking is crucial for smart city projects nowadays, as it offers an environment where people and things are connected. This paper presents a chronology of factors on the development of smart cities, including IoT technologies as network infrastructure. Increasing IoT nodes leads to increasing data flow, which is a potential source of failure for IoT networks. The biggest challenge of IoT networks is that the IoT may have insufficient memory to handle all transaction data within the IoT network. We aim in this paper to propose a potential compression method for reducing IoT network data traffic. Therefore, we investigate various lossless compression algorithms, such as entropy or dictionary-based algorithms, and general compression methods to determine which algorithm or method adheres to the IoT specifications. Furthermore, this study conducts compression experiments using entropy (Huffman, Adaptive Huffman) and Dictionary (LZ77, LZ78) as well as five different types of datasets of the IoT data traffic. Though the above algorithms can alleviate the IoT data traffic, adaptive Huffman gave the best compression algorithm. Therefore, in this paper, Citation: Nasif, A.; Othman, Z.A.; we aim to propose a conceptual compression method for IoT data traffic by improving an adaptive Sani, N.S.
[Show full text]
Key-Based Self-Driven Compression in Columnar Binary JSON
Otto von Guericke University of Magdeburg Department of Computer Science Master's Thesis Key-Based Self-Driven Compression in Columnar Binary JSON Author: Oskar Kirmis November 4, 2019 Advisors: Prof. Dr. rer. nat. habil. Gunter Saake M. Sc. Marcus Pinnecke Institute for Technical and Business Information Systems / Database Research Group Kirmis, Oskar: Key-Based Self-Driven Compression in Columnar Binary JSON Master's Thesis, Otto von Guericke University of Magdeburg, 2019 Abstract A large part of the data that is available today in organizations or publicly is provided in semi-structured form. To perform analytical tasks on these { mostly read-only { semi-structured datasets, Carbon archives were developed as a column-oriented storage format. Its main focus is to allow cache-efficient access to fields across records. As many semi-structured datasets mainly consist of string data and the denormalization introduces redundancy, a lot of storage space is required. However, in Carbon archives { besides a deduplication of strings { there is currently no compression implemented. The goal of this thesis is to discuss, implement and evaluate suitable compression tech- niques to reduce the amount of storage required and to speed up analytical queries on Carbon archives. Therefore, a compressor is implemented that can be configured to apply a combination of up to three different compression algorithms to the string data of Carbon archives. This compressor can be applied with a different configuration per column (per JSON object key). To find suitable combinations of compression algo- rithms for each column, one manual and two self-driven approaches are implemented and evaluated. On a set of ten publicly available semi-structured datasets of different kinds and sizes, the string data can be compressed down to about 53% on average, reducing the whole datasets' size by 20%.
[Show full text]
Session Overhead Reduction in Adaptive Streaming
Session Overhead Reduction in Adaptive Streaming A Technical Paper prepared for SCTE•ISBE by Alexander Giladi Fellow Comcast 1899 Wynkoop St., Denver CO +1 (215) 581-7118 [email protected] © 2020 SCTE•ISBE and NCTA. All rights reserved. 1 Table of Contents Title Page Number 1. Introduction .......................................................................................................................................... 3 2. Session overhead and HTTP compression ........................................................................................ 4 3. Reducing traffic overhead ................................................................................................................... 5 3.1. MPD patch .............................................................................................................................. 6 3.2. Segment Gap Signaling ......................................................................................................... 7 3.3. Efficient multi-DRM signaling ................................................................................................. 8 4. Reducing the number of HTTP requests ............................................................................................ 9 4.1. Asynchronous MPD updates.................................................................................................. 9 4.2. Predictive templates ............................................................................................................. 10 4.3. Timeline extension ..............................................................................................................
[Show full text]
Context-Aware Encoding & Delivery in The
Context-Aware Encoding & Delivery in the Web ICWE 2020 Benjamin Wollmer, Wolfram Wingerath, Norbert Ritter Universität Hamburg 9 - 12 June, 2020 Business Impact of Page Speed Business Uplift Speed Speed Downlift Uplift Business Downlift Felix Gessert: Mobile Site Speed and the Impact on E-Commerce, CodeTalks 2019 So Far On Compression… GZip SDCH Deflate Delta Brotli Encoding GZIP/Deflate – The De Facto Standard in the Web Encoding Size None 200 kB Gzip ~36 kB This example text is used to show how LZ77 finds repeating elements in the example[70;14] text ~81.9% saved data J. Alakuijala, E. Kliuchnikov, Z. Szabadka, L. Vandevenne: Comparison of Brotli, Deflate, Zopfli, LZMA, LZHAM and Bzip2 Compression Algorithms, 2015 Delta Encoding – Updating Stale Content Encoding Size None 200 kB Gzip ~36 kB Delta Encoding ~34 kB 83% saved data J. C. Mogul, F. Douglis, A. Feldmann, B. Krishnamurthy: Potential Benefits of Delta Encoding and Data Compression for HTTP, 1997 SDCH – Reusing Dictionaries Encoding Size This is an example None 200 kB Gzip ~36 kB Another example Delta Encoding ~34 kB SDCH ~7 kB Up to 81% better results (compared to gzip) O. Shapira: Shared Dictionary Compression for HTTP at LinkedIn, 2015 Brotli – SDCH for Everyone Encoding Size None 200 kB Gzip ~36 kB Delta Encoding ~34 kB SDCH ~7 kB Brotli ~29 kB ~85.6% saved data J. Alakuijala, E. Kliuchnikov, Z. Szabadka, L. Vandevenne: Comparison of Brotli, Deflate, Zopfli, LZMA, LZHAM and Bzip2 Compression Algorithms, 2015 So Far On Compression… Theory vs. Reality GZip (~80%) SDCH Deflate
[Show full text]
Lempel-Ziv Sliding Window Update with Suffix Arrays
i-ETC: ISEL Academic Journal of Electronics, Telecommunications and Computers CETC2011 Issue, Vol. 2, n. 1 (2013) ID-4 LEMPEL-ZIV SLIDING WINDOW UPDATE WITH SUFFIX ARRAYS Artur Ferreira1,3,4 Arlindo Oliveira2,4 Mario´ Figueiredo3,4 1Instituto Superior de Engenharia de Lisboa (ISEL) 2Instituto de Engenharia de Sistemas e Computadores – Investigaçao˜ e Desenvolvimento (INESC-ID) 3Instituto de Telecomunicaçoes˜ (IT) 4Instituto Superior Tecnico´ (IST), Lisboa, PORTUGAL [email protected] [email protected] [email protected] Keywords: Lempel-Ziv compression, suffix arrays, sliding window update, substring search. Abstract: The sliding window dictionary-based algorithms of the Lempel-Ziv (LZ) 77 family are widely used for uni- versal lossless data compression. The encoding component of these algorithms performs repeated substring search. Data structures, such as hash tables, binary search trees, and suffix trees have been used to speedup these searches, at the expense of memory usage. Previous work has shown how suffix arrays (SA) can be used for dictionary representation and LZ77 decomposition. In this paper, we improve over that work by proposing a new efficient algorithm to update the sliding window each time a token is produced at the output. The pro- posed algorithm toggles between two SA on consecutive tokens. The resulting SA-based encoder requires less memory than the conventional tree-based encoders. In comparing our SA-based technique against tree-based encoders, on a large set of benchmark files, we find that, in some compression settings, our encoder is also faster than tree-based encoders. 1 INTRODUCTION dictionaries [4] and to find repeating sub-sequences for data deduplication [1], among other applications.
[Show full text]