|
From: Radosław K. <rad...@ko...> - 2010-12-07 12:18:43
|
2010/12/7 Christoph Litauer <li...@un...> > > I can see no bottleneck at all ... but "status storage" still reports > 40MB/s write performance. > > AFAIK, when bacula-sd run a copy job it perform this by single thread, synchronous. So, bacula-sd read a block from source volume, then write it into destination volume, repeats until end of source volume. This leads to about a half of expected throughput. There was a thread on bacula-devel list about it: "slow copying from tape to tape." At this thread Kern confirmed a possible improvement: Yes, I am working on this now (a very low priority project for the moment) > but > it will probably only be in the Enterprise version and not in the community > version. > bye -- Radosław Korzeniewski rad...@ko... |