Oracle Overload

Oracle Tips by Burleson Consulting
Don Burleson

Oracle is optimized to work on any database server from an Mainframe to a Macintosh, but you must be careful to avoid server overload conditions.

Because Oracle10g AWR tracks server performance over time, we produce management reports that show you exactly when your server was overloaded.

You can also get AWR RAM, I/O and CPU trend reports to help you predict when you need new hardware.

Tip! Take two AWR/STATSPACK snapshots, 30 seconds apart and generate a one minute elapsed time report. This will often reveal the source of a server loverload.

If you have not yet completely optimized your SQL, the following conditions are generally true:

I/O overload - This is evidenced by high "db file sequential read" and "db file scattered read" waits and can be detected in the Oracle10g dba_hist_filestatxs view. You may want to look at SQL that issues unnecessary table block access, possibly due to missing indexes or poor CBO statistics. Assuming that your SQL is optimized, your only solutions are adding additional RAM for your data buffers or switching to solid-state disk.

CPU overhead - With the advent of 64-bit Oracle and super-large data block buffers (db_cache_size, db_keep_cache_size), the main bottleneck for many databases has shifted from I/O to CPU. If you see CPU in your top wait events, you should look at SQL that may be causing unnecessary Logical I/O against the data buffers. You also want to look at the library cache to see if excessive parsing might be causing the CPU consumption. Assuming your have optimized Oracle, your option is to add more CPU's or faster CPU processors.

http://www.dba-oracle.com/oracle_tips_intel.htm

RAM overload - The Oracle 10g Automatic Memory Management (AMM) utility has facilities in Oracle10g Enterprise Manager for detecting too-small SGA regions (db_cache_size, shared_pool_size, pga_aggregate_target, etc.). You can re-allocate RAM within these regions, reducing pga_aggregate_target if you have no disk sorts or hash joins, reducing shared_pool_size if you have no library cache contention and reducing db_cache_size if you have low disk I/O.

Detecting I/O Overload

Here is a sample Oracle 10g script to detect all files with physical reads over 10,000 during the snapshot period:

break on begin_interval_time skip 2

column phyrds              format 999,999,999
column begin_interval_time format a25

select
   begin_interval_time,
   filename,
   phyrds
from
   dba_hist_filestatxs
natural join
   dba_hist_snapshot

where

   phyrds > 10000
;

Below we see a running total of Oracle physical reads from phys_reads.sql. Note that the snapshots are collected every hour in this example, and many DBAs will increase the default collection frequency of AWR snapshots. Starting from this script, we could easily add a where clause criteria and create a unique time-series exception report.

SQL> @phys_reads

BEGIN_INTERVAL_TIME FILENAME PHYRDS
------------------------- ---------------------------------------- ------------
24-FEB-04 11.00.32.000 PM E:\ORACLE\ORA92\FSDEV10G\SYSTEM01.DBF    164,700
                          E:\ORACLE\ORA92\FSDEV10G\UNDOTBS01.DBF    26,082
                          E:\ORACLE\ORA92\FSDEV10G\SYSAUX01.DBF    472,008
                          E:\ORACLE\ORA92\FSDEV10G\USERS01.DBF      21,794
                          E:\ORACLE\ORA92\FSDEV10G\T_FS_LSQ.ORA     12,123

24-FEB-04 12.00.32.000 PM E:\ORACLE\ORA92\FSDEV10G\SYSTEM01.DBF    164,700
                          E:\ORACLE\ORA92\FSDEV10G\UNDOTBS01.DBF    26,082

��