%wcache is to low

[GLOT]

hello,

We have a server HP-UX 11.0 64 - bit with two processors and 2GO of RAM.

On this server there is a base Oracle 8.1.7

The parameters
Dbc_max_pct = 20
Dbc_min_pct = 5

Size current of the buffer:

echo "bufpages /D" |adb /stand/vmunix /dev/kmem
bufpages 104857

I have a problem when I make a sar -b the result shows that the value of %wcache is very lower than that advised by HP.

My value of %Wcache is on average :7 and HP advice >=70

However my %rcache value is good.

I joint you my output of sar -b.

which are the parameters for resolve this problem ?

Thanks for your help

[Steven E. Protter]

Based on you post I have two thoughts:

1) dbc_max_pct should be dropped to between 7-10

2) You need to collect more data.

Attaching scripts to accomplish goal #2

SEP

[A. Clay Stephenson · Accepted solution]

First of all, 2GB is considered pretty low for Oracle. In your case, dbc_max_pct doesn't matter because bufpages is set to a non-zero value. 104857 X 4096 ~ 400MB -- a decent value for 11.0 and 2GB. If your database is doiung raw/io then you really don't care about the %wcache hits because you are bypassing the buffer cache. On boxes that do lots of random i/o (e.g database servers), you can't really expect high write cache hit rates so the 70% stuff is typically unrealistic. The real question to ask is: How is my system performing?" Are queries being handled reasonably well? Are inserts and updates okay?

On database boxes you really want to look at the cache hit rates within the database itself.

[GLOT]

hi steven,

I joint you my report file on my system

thanks

[Tim Sanko]

With An Oracle 8.1.7 I have seen as little as a 3 I have not seen more than 25 on any oracle server with a load on. If you use raw partitions the value goes way down.

The real concern is the performance. Dump an hour of sar data during peak and look for magic problem areas such as swap/in/sec, swap/out/sec, and process switching.

Hope that helps.

Tim