There was some discussion last week about tape write speeds, including
one message that said something like "I guess if you're using an LTO-3
drive, you must have spooling turned on."  Since I'm using an LTO-3
drive and had not been using a spool file, I thought I'd enable this
and see what effect it had on performance.

To my surprise, it took over twice as long to perform a full backup
when spooling was enabled.  The numbers for two full backups made two
weeks apart are:

    Without spooling: 3.8 TB in 24:57 at 44.8 MB/s
    With spool file:  3.9 TB in 55:43 at 20.4 MB/s

Both jobs were on the same hardware, with everything -- the client,
director, storage server, console, and database -- on the same
machine, and with no other backup jobs running.  The only significant
difference was the addition of a 60 GB spool file on the fastest RAID
array on the system just prior to the second run.

What seems to happen is that without a spool file, Bacula does a very
good job of overlapping the read of data with the write to tape and
the update of the database.  When spooling is in use, these three
operations are performed sequentially, significantly slowing down the
speed of the job.  On the other hand, I did notice a speed improvement
in the concurrent backup of several remote clients with the spool file
enabled.

So while there are many good reasons for using a spool file listed in
the "Data Spooling" section of the manual, speed isn't always one of
them.

-- John Kodis.


-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems?  Stop!  Download the new AJAX search engine that makes
searching your log files as easy as surfing the  web.  DOWNLOAD SPLUNK!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=103432&bid=230486&dat=121642
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users

Reply via email to