Oracle8 Tuning Release 8.0 A58246-01 |
|
Dynamic performance views are useful for identifying instance-level performance problems. Whereas the underlying X$ tables represent internal data structures that can be modified by SQL statements, the V$ views allow users other than SYS read-only access to this data. This chapter describes the views of the greatest use for both performance tuning and ad hoc investigation-for example, when users report a sudden deterioration in response time.
See Also: For complete information on all dynamic performance tables, see Oracle8 Reference.
These views concern the instance as a whole and record statistics either since startup of the instance or (in the case of the SGA statistics) the current values, which will remain constant until altered by some need to reallocate SGA space. Cumulative statistics are from startup.
The single most important fixed view is V$SYSSTAT, which contains the statistic name in addition to the value. The values from this table form the basic input to the instance tuning process.
These views either operate at the session level or primarily concern transient values. Session data is cumulative from connect time.
The structure of V$SESSION_WAIT makes it easy to check in real time whether any sessions are waiting, and if so, why. For example:
SELECT SID , EVENT FROM V$SESSION_EVENT WHERE WAIT_TIME = 0;
You can then investigate further to see whether such waits occur frequently and whether they can be correlated with other phenomena, such as the use of particular modules.
This section describes procedures for:
Key ratios are expressed in terms of instance statistics. For example, the consistent change ratio is consistent changes divided by consistent gets. The simplest effective SQL*Plus script for finding the current value of a statistic is of the form:
col NAME format a35 col VALUE format 999,999,990 select NAME, VALUE from V$SYSSTAT S where lower(NAME) like lower(`%&stat_name%') /
Note: Two LOWER functions in the preceding query make it case insensitive and allow it to report data from the 11 statistics whose names start with "CPU" or "DBWR". No other upper-case characters appear in statistic names.
You can use the following query, for example, to report all statistics containing the word "get" in their name:
SQL> @STAT GET
It is preferable, however, to use some mechanism that records the change in the statistic(s) over a known period of time.
You can adapt the following script to show the rate of change for any statistic, latch, or event. For a given statistic, this script tells you the number of seconds between two checks of its value, and its rate of change.
set veri off define secs=0 define value=0 col value format 99,999,999,990 new_value value col secs format a10 new_value secs noprint col delta format 9,999,990 col delta_time format 9,990 col rate format 999,990.0 col name format a30 select name,value, to_char(sysdate,'sssss') secs, (value - &value) delta, (to_char(sysdate,'sssss') - &secs) delta_time, (value - &value)/ (to_char(sysdate,'sssss') - &secs) rate from v$sysstat where name = '&&stat_name' /
Note: This script must be run at least twice, because the first time it is run, it will initialize the SQL*Plus variables.