ekke85 wrote:

> The spooling attribute was not enabled, I have enabled it now. This is what I 
> get writing to disk and then also writing that file to tape with tar:

Tape speed testing while writing a repetitive file is useless as 
hardware compression makes it go a lot faster than natually.

For tape tests use /dev/random

NFS is a dreadful protocol and usually very slow (highly inefficient). 
If you were streaming off the netapp at 22MB/s then you are doing well.

There are mount parameter you can apply to NFS to speed things up, we 
well as moving to 9000 byte packets on a Gb network (don't do this in a 
mixed speed environment, there are always problems)


Concentrate on getting the spooling working well - if you are using 
anything higher than LTO2 your spool area will have to be a dedicated 
striped disk array (or a SLC SSD) as LTO3 and onwards can easily outrun 
a single hard drive's ability to stream data (if the drive has to do any 
seeking, forget it....)

Bottlenecks appear all over the place with high speed tape. The only way 
to find them is to test every component for throughput and latency as 
well as testing the complete chain.

Don't forget to test the source speed of /dev/zero and /dev/random - 
check throughputs when piping to /dev/zero. Ditto for despooling area. 
If the tape speed is anywhere near approaching these speeds then your 
results will be distorted.





------------------------------------------------------------------------------
The Palm PDK Hot Apps Program offers developers who use the
Plug-In Development Kit to bring their C/C++ apps to Palm for a share
of $1 Million in cash or HP Products. Visit us here for more details:
http://p.sf.net/sfu/dev2dev-palm
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users