On Sun, 26 Jun 2011 16:13:47 -0400 "John Stoffel" <j...@stoffel.org> wrote:
John, > Neat. What type of speedup or space savings are you seeing? I just > recently played around with lrzip (Long Range Zip by Con Kolivas) and > came away quite disappointed by it's speed. It would only give me > around a 10-15% space savings over bzip2, but at a cost of having to > use lots more memory and lots more CPU (multiple cores actually) to > get anywhere close to the native bzip2 compression speed. I have not done real speed test inside bacula. You can find a lot of compressor benchmark on the net. In bacula, you can expect the same kind of speed and compression ratio. > Of course I was compressing already compressed files in one test, but > in another test I was compressing pretty regular 5gb Magma Talus > volvanos (EDA design tool file format) and just not seeing good > results. > > I had also compiled in the LZO library and tried that, but bzip2 and > even gzip were just as good if not much better. Sorry I don't have > numbers handy. LZO is design for speed (both compression and decompression). It is not design to maximize the compression ratio. IMHO, LZO in bacula should be used if your CPU can compress data at least at the same speed as (your disks can read and your bandwidth between the fd <-> sd). LZO should be nice for users that only use disk storage. I will save disk space at almost no cost. > Now onto the patch commentary. Would it be possible to change the > code to support more compression schemes, specifically the BZIP2 > standard would be good to be able to implement. Your code is so > nicely done, that it shouldn't be hard to do, except the #ifdefs are > going to get ugly quickly as more are added. When the patch will be accepted, I will start working on liblzma. It should provide smaller and faster results than bzip2. IMHO, bacula should include all compression schemes floating around. A very fast and a very good (in term of compression ratio) seems a good starting point. -- Laurent Papier ------------------------------------------------------------------------------ All of the data generated in your IT infrastructure is seriously valuable. Why? It contains a definitive record of application performance, security threats, fraudulent activity, and more. Splunk takes this data and makes sense of it. IT sense. And common sense. http://p.sf.net/sfu/splunk-d2d-c2 _______________________________________________ Bacula-devel mailing list Bacula-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-devel
