You: > On Aug 13, 2010, at 2:23 PM, Dietz Pröpper wrote: > > You: > >> On Aug 13, 2010, at 4:10 AM, Dietz Pröpper wrote: > >>> IMHO there are two problems with hardware compression: > >>> 1. Data mix: The compression algorithms tend to work quite well on > >>> compressable stuff, but can't cope very well with precompressed > >>> stuff, i.e. encrypted data or media files. On an old DLT drive (but > >>> modern hardware should perform in a similar fashion), I get around > >>> 7MB/s with "normal" data and around 3MB/s with precrompessed stuff. > >>> The raw tape write rate is somewhere around 4MB/s. And even worse - > >>> due to the fact that the compression blurs precompressed data, it > >>> also takes noticeable more tape space. > >> > >> Those problems affect software compression, too. > > > > Hmm, I can't reproduce them with gzip on the data in question. > > Maybe, then, your job is not I/O bound?
Sorry, I was a little unclear - with gzip, file size does not get bigger, regardless of the entropy of the data compressed. > If backup is limited by the speed at which you can write to tape then it > logically follows that you will get the observed behaviour you mention > above. More compressible source data will lead to faster backup rates > because those data will compress to less bits that need to be written > to tape. Of course. > Conversely, if the compression algorithm does not take steps > to guard against growth due to compressed input, backup speeds will > fall below nominal write speeds with such data because the source data > will result in more bits to be written, which will take longer > (relative to the source). That's what I observe with my ancient DLT drive. ------------------------------------------------------------------------------ This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev _______________________________________________ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
