DBWR ave. scan depth?

[jgmagnus]

I have never heard of this but some crazy DBA Tool is telling me that maybe my DB_FILES are too high. Here is some basic info. Any help is appreciated.

DBWR Ave. Scan Depth = 1028.9114

db_files integer 100

thanks in advance,
Magnus

[pwoneill]

After a bit of research,

DBWR Avg Scan Depth = (DBWR summed scan depth) / (DBWR lru scans)

Those values come from v$sysstat.

A quote I found from basically the same question was

<i>The DBWn scan depth is both a function of the scan depth boundaries and the write batch size. Further, if I recall correctly, the scan depth boundaries are primarily based on the write batch size too, and the db_block_buffers limit only applies if the write batch is very large. Of course, the default write batch size is a function of db_files and thus the comment (I assume).</i>

From [url]http://www.ixora.com.au/q+a/dbwn.htm[/url]

So basically because db_files determines write batch size, the scan depth boundries are large, causing a large DBWR summed scan depth. Divide that by a smaller DBWR lru scans and it gives the 1028 which I suppose is too high a ratio.

However I would wait for someone more informed to reply before changing anything :)

[sambavan]

That is pretty cool pwoneill. Btw, I was reading the oracle 24x7 chapter 7 and I noticed it to talk a lot on the

_DB_WRITER_SCAN_DEPTH : the number of buffers that DBWR will scan in the LRU list when looking for buffers. This is influenced by _DB_BLOCK_WRITE_BATCH


do this query and will show you all the hidden parameters and their values.

select ksppinm, ksppdesc, ksppstvl
from x$ksppi i,
x$ksppcv v
where v.indx = i.indx;


There are around 486 parameters and be v.v. carefull when playing with the uderscore parameters (not advisible!).

Sam