本文主要是介绍db block gets (current gets) VS. consistent gets，希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！
当前读（db block gets / current gets）与一致读（consistent gets）统称为逻辑读，逻辑读可能需要物理读把块读到cache中。
当前读指读现在已提交了的数据，一般在全表扫描读数据字典、更新、删除时发生。
一致读指读发出SELECT的那个时间点SCN的数据，一般在查询（增删改都可能有隐式查询）时发生。
db block gets are blocks read in 'current' mode. meaning, get me the block as it exists right now,
no read consistency. do NOT undo changes. Again, it is a mode, not an indication of how many
blocks didn't need changes -- rather how many blocks we asked for in that MODE.In general, a statement will either get a block from the cache in current mode or consistent read mode.
Those are logical IO's. The logical IO might need to do a physical IO in order to get the block into the cache -
but the statement is doing either a consistent read or current mode read. You Askedhi tomi your discussion regarding consistant gets and db dblock gets
you mention about block read in CURRENT MODE AND consistent modeCould you please explain exact meaning of this words by giving
some examples?Thanks and we said...Ok, when you turn on autotrace in sqlplus you can see these stats. Lets run a query:ops$tkyte@ORA817.US.ORACLE.COM> set autotrace traceonly statistics
ops$tkyte@ORA817.US.ORACLE.COM>
ops$tkyte@ORA817.US.ORACLE.COM> select * from emp;14 rows selected.Statistics
----------------------------------------------------------0 recursive calls4 db block gets2 consistent gets0 physical reads0 redo size1979 bytes sent via SQL*Net to client430 bytes received via SQL*Net from client2 SQL*Net roundtrips to/from client0 sorts (memory)0 sorts (disk)14 rows processedHere we had 4 db block gets. Those were blocks read in CURRENT MODE. The blocks that
were read are actually the blocks that tell us how to FULL SCAN the dept table (data
dictionary type of information). We need to get that in CURRENT MODE (as of RIGHT NOW)
to get an accurate picture of what the table looks like. We also had 2 consistent gets -- these are blocks we read in "consistent read" mode --
also known as query mode. This means we were reading them as of the POINT IN TIME the
query began. Seehttp://download-east.oracle.com/docs/cd/A81042_01/DOC/server.816/a76965/c23cnsis.htm#17882for a great discussion of this.Now, if we do a delete:ops$tkyte@ORA817.US.ORACLE.COM> delete from emp;14 rows deleted.Statistics
----------------------------------------------------------0 recursive calls20 db block gets1 consistent gets0 physical reads4220 redo size1009 bytes sent via SQL*Net to client796 bytes received via SQL*Net from client4 SQL*Net roundtrips to/from client1 sorts (memory)0 sorts (disk)14 rows processedwe do lots more db block gets -- why? because in order to delete the data we need to get
the block in CURRENT MODE, as it exists right then.why did we do a consistent get? because the "read" part of the delete uses the
consistent read mechanism -- we only delete data that existed in the table as of the
point in time the delete began. Consider if DEPT was a 1,000,000 row table instead.
It'll take a while to delete all of those rows. As you are deleting however, other
sessions are inserting and committing data. This consistent read mechanism makes it so
that we only delete the rows that existed WHEN WE BEGAN the delete. We will not delete
this new data being inserted.How Oracle Manages Data Concurrency and Consistency
http://docs.oracle.com/cd/B10501_01/server.920/a96524/c21cnsis.htm#2570
这篇关于db block gets (current gets) VS. consistent gets的文章就介绍到这儿，希望我们推荐的文章对编程师们有所帮助！
