Brave new stats in 11.1.0.7
From: Mladen Gogala <gogala.mladen_at_gmail.com>
Date: Mon, 26 Jan 2009 10:38:02 +0000 (UTC)
Message-ID: <glk3qa$k6g$1_at_solani.org>
There are some very strange new statistics in 11.1.0.7, namely the
"fastpath" mechanism for consistent gets. There are 2 types
of fastpath consistent gets, one for the db blocks, one apart from that one, probably for the index blocks. That would be consistent with Jonathan's and Alex Fatkulin's blogs which have recently blogged about "consistent gets from cache (fastpath)" statistics. Now, consistent gets are at the basis of how Oracle RDBMS functions: when returning data from a query, Oracle rebuilds the blocks from the UNDO segments so that they are consistent with last transaction committed before the query had started. That process of rebuilding the blocks to the consistent state is called "a consistent get". Consistent gets are an essential part of Oracle RDBMS ever since Oracle 6. Before that, there was no UNDO tablespace but so called
"BI file", but let me digress no more.
This mechanism is so basic and essential that every change raises an alarm flag. I would like someone from the Oracle Corp. to explain the new mechanism and alleviate our stress. The illustration of the new statistics is below:
db block gets from cache (fastpath) 1117557
consistent gets from cache (fastpath) 17950678
index fast full scans (full) 9201
index fast full scans (rowid ranges) 0
index fast full scans (direct read) 0
Date: Mon, 26 Jan 2009 10:38:02 +0000 (UTC)
Message-ID: <glk3qa$k6g$1_at_solani.org>
There are some very strange new statistics in 11.1.0.7, namely the
"fastpath" mechanism for consistent gets. There are 2 types
of fastpath consistent gets, one for the db blocks, one apart from that one, probably for the index blocks. That would be consistent with Jonathan's and Alex Fatkulin's blogs which have recently blogged about "consistent gets from cache (fastpath)" statistics. Now, consistent gets are at the basis of how Oracle RDBMS functions: when returning data from a query, Oracle rebuilds the blocks from the UNDO segments so that they are consistent with last transaction committed before the query had started. That process of rebuilding the blocks to the consistent state is called "a consistent get". Consistent gets are an essential part of Oracle RDBMS ever since Oracle 6. Before that, there was no UNDO tablespace but so called
"BI file", but let me digress no more.
This mechanism is so basic and essential that every change raises an alarm flag. I would like someone from the Oracle Corp. to explain the new mechanism and alleviate our stress. The illustration of the new statistics is below:
[oracle_at_dibidus2 ~]$ sqlplus / as sysdba
SQL*Plus: Release 11.1.0.7.0 - Production on Mon Jan 26 11:16:54 2009
Copyright (c) 1982, 2008, Oracle. All rights reserved.
Connected to:
Oracle Database 11g Enterprise Edition Release 11.1.0.7.0 - Production
With the Partitioning, OLAP, Data Mining and Real Application Testing
options
SQL> select name,value from v$sysstat
2 where name like '%fast%';
NAME
VALUE
db block gets from cache (fastpath) 1117557
consistent gets from cache (fastpath) 17950678
index fast full scans (full) 9201
index fast full scans (rowid ranges) 0
index fast full scans (direct read) 0
SQL>
-- Mladen Gogala http://mgogala.freehostia.comReceived on Mon Jan 26 2009 - 04:38:02 CST