Hi again,

13.10.2009 20:35, Arno Lehmann wrote:
> Hello,
> 
> 13.10.2009 14:27, justAck wrote:
>> Arno,
>>
>>
>> Arno Lehmann wrote:
...
>> Maybe some cache is involved, need to test deeper, just wanted feedback
>> about possible scenarios.
> 
> See above - I'd try vmstat 1 during a restore and subsequent read 
> performance test, then wait a while, and do a nother read performance 
> test.

Found something interesting... you can get information about the 
blocks of a file linux keeps in memory.

http://insights.oetiker.ch/linux/fadvise/ has a small sample program, 
which I could easily compile and run, and it gives useful output.

Run that on a file while it is restored to see if Bacula uses the 
fadvise function (which Bacula in principle uses) to make sure files 
it writes are not buffered (which makes sense, as the probability that 
the same file is used immediately again is almost zero during a restore).

Then run it while you do your read performance test.

I don't know if the results actually tell us something, but you might 
be able to find a pattern...

Arno

> Arno
> 
>> Thanks again!
> 

-- 
Arno Lehmann
IT-Service Lehmann
Sandstr. 6, 49080 Osnabrück
www.its-lehmann.de

------------------------------------------------------------------------------
Come build with us! The BlackBerry(R) Developer Conference in SF, CA
is the only developer event you need to attend this year. Jumpstart your
developing skills, take BlackBerry mobile applications to market and stay 
ahead of the curve. Join us from November 9 - 12, 2009. Register now!
http://p.sf.net/sfu/devconference
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users