These are benchmarks performed on a 3GHz quad core Intel Core2 with 8GB ram using lrzip v0.42. The first comparison is that of a linux kernel tarball (2.6.31). In all cases the default options were used. 3 other common compression apps were used for comparison, 7z which is an excellent all-round lzma based compression app, gzip which is the benchmark fast standard that has good compression, and bzip2 which is the most common linux used compression. In the following tables, lrzip means lrzip default options, lrzip(lzo) means lrzip using the lzo backend, lrzip(gzip) means using the gzip backend, lrzip(bzip2) means using the bzip2 backend and lrzip(zpaq) means using the zpaq backend. linux-2.6.31.tar Compression Size Percentage Compress Decompress None 365711360 100 7z 53315279 14.6 2m4.770s 0m5.360s lrzip 52372722 14.3 2m48.477s 0m8.336s lrzip(zpaq) 43455498 11.9 10m11.335 10m14.296 lrzip(lzo) 112151676 30.7 0m14.913s 0m5.063s lrzip(gzip) 73476127 20.1 0m29.628s 0m5.591s lrzip(bzip2) 60851152 16.6 0m43.539s 0m12.244s bzip2 62416571 17.1 0m44.493s 0m9.819s gzip 80563601 22.0 0m14.343s 0m2.781s These results are interesting to note the compression of lrzip by default is only slightly better than lzma, but at some cost in time at the compress and decompress end of the spectrum. Clearly zpaq compression is much better than any other compression algorithm by far, but the speed cost on both compression and decompression is extreme. At this size compression, lzo is interesting because it's faster than simply copying the file but only offers modest compression. What lrzip offers at this end of the spectrum is extreme compression if desired. Let's take two kernel trees one version apart as a tarball, linux-2.6.31 and linux-2.6.32-rc8. These will show lots of redundant information, but hundreds of megabytes apart, which lrzip will be very good at compressing. For simplicity, only 7z will be compared since that's by far the best general purpose compressor at the moment: Tarball of two kernel trees, one version apart. Compression Size Percentage Compress Decompress None 749066240 100 7z 108710624 14.5 4m4.260s 0m11.133s lrzip 57943094 7.7 3m08.788s 0m10.747s lrzip(lzo) 124029899 16.6 0m18.997s 0m7.107s Things start getting very interesting now when lrzip is really starting to shine. Note how it's not that much larger for 2 kernel trees than it was for one. That's because all the similar data in both kernel trees is being compressed as one copy and only the differences really make up the extra size. All compression software does this, but not over such large distances. If you copy the same data over multiple times, the resulting lrzip archive doesn't get much larger at all. Using the first example (linux-2.6.31.tar) and simply copying the data multiple times over gives these results with lrzip(lzo): Copies Size Compressed Compress Decompress 1 365711360 112151676 0m14.913s 0m5.063s 2 731422720 112151829 0m16.174s 0m6.543s 3 1097134080 112151832 0m17.466s 0m8.115s I had the amusing thought that this compression software could be used as a bullshit detector if you were to compress peoples' speeches because if their talks were full of catchphrases and not much actual content, it would all be compressed down. So the larger the final archive, the less bullshit =) Now let's move on to the other special feature of lrzip, the ability to compress massive amounts of data on huge ram machines by using massive compression windows. This is a 10GB virtual image of an installed operating system and some basic working software on it. The default options on the 8GB machine meant that it was using a 5 GB window. 10GB Virtual image: Compression Size Percentage Compress Time Decompress Time None 10737418240 100.0 gzip 2772899756 25.8 7m52.667s 4m8.661s bzip2 2704781700 25.2 20m34.269s 7m51.362s xz 2272322208 21.2 58m26.829s 4m46.154s 7z 2242897134 20.9 29m28.152s 6m35.952s lrzip 1361276826 12.7 27m45.874s 9m20.046 lrzip(lzo) 1837206675 17.1 4m48.167s 8m28.842s lrzip(zpaq) 1341008779 12.5 4h11m14s lrzip(zpaq)M 1270134391 11.8 4h30m14 lrzip(zpaq)MW 1066902006 9.9 At this end of the spectrum things really start to heat up. The compression advantage is massive, with the lzo backend even giving much better results than 7z, and over a ridiculously short time. Note that it's not much longer than it takes to just *read* a 10GB file. Unfortunately at these large compression windows, the decompression time is significantly longer, but it's a fair tradeoff I believe :) What appears to be a big disappointment is actually zpaq here which takes more than 8 times longer than lzma for a measly .2% improvement. The reason is that most of the advantage here is achieved by the rzip first stage. The -M option was included here for completeness to see what the maximum possible compression was for this file on this machine, while the MW run was with the options -W 200 (to make the window larger than the file and the ram the machine has), and it still completed but induced a lot of swap in the interim. This should help govern what compression you choose. Small files are nicely compressed with zpaq. Intermediate files are nicely compressed with lzma. Large files get excellent results even with lzo provided you have enough ram. (Small being < 100MB, intermediate <1GB, large >1GB). Or, to make things easier, just use the default settings all the time and be happy as lzma gives good results. :D Con Kolivas Sat, 19 Dec 2009