> Hi there, > I've been having some problems attempting to increase the write speed to my > tape drive through Bacula. > > If I use the operating system to communicate directly with the tape drive, > I get the appropriate read and write speeds but using Bacula, I get a third > of the speed. I have tried spooling the data to a separate physical drive > before writing, no luck. I have played around with the block sizes using > 64K, 128K, 196K and 256K but still the same performance. I have tried > various backup sized ranging from under 1gb to 80gb but the speed stays > constant on all tests. > Changing the block size should only make a small percentage change. Also spooling of a single job will not help because bacula fills the spool file completely then stops the backup to send the spool file to tape. When it does that you should see transfer rates of (30 to 45MB/s) for the despooling but the whole job will have a lower backup rate because of the lack of both spooling and despooling running at the same time.
> > My average write speed seems to stay around 20mb/s give or take a few > megabytes. I should be getting at least double that speed for the drive and > SCSI card that I am using. > 20 to 30 MB/s is totally reasonable for a full backup on LTO2 drive depending on compression. > My software setup is as follows: > FreeBSD 5.5 x86 > Bacula 2.2.5 (Installed from source) > MySQL 5.0.45 > > My hardware setup is as follows: > CPU - AMD AM2 5600+ > Motherboard - Asus M2N-LR > SCSI Card - Adaptec 29160N > Tape Drive - HP StorageWorks Ultrium 448 > Data Cartridge - HP LTO2 Ultrium 400GB > > Here are some examples: > ####DD Read/Write Test#### > server1# dd if=/dev/zero of=/dev/nsa0 bs=65536 count=300000 > 300000+0 records in > 300000+0 records out > 19660800000 bytes transferred in 299.338943 secs (65680729 bytes/sec) > server1# mt -f /dev/nsa0 rewind > server1# dd of=/dev/null if=/dev/nsa0 bs=65536 count=300000 > 300000+0 records in > 300000+0 records out > 19660800000 bytes transferred in 291.253620 secs (67504054 bytes/sec) > > These numbers are artificially high because zeros will compress very highly so very few tape blocks are actually written to the tape so all you are testing here is the speed of the SCSI bus. A more realistic test would be to substitute /dev/sda for the input assuming you have data on your sda drive... John ------------------------------------------------------------------------- SF.Net email is sponsored by: The Future of Linux Business White Paper from Novell. From the desktop to the data center, Linux is going mainstream. Let it simplify your IT future. http://altfarm.mediaplex.com/ad/ck/8857-50307-18918-4 _______________________________________________ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
