Re: 'redo synch time overhead ...' statistics
- From: Stefan Koehler <contact@xxxxxxxx>
- To: ORACLE-L <oracle-l@xxxxxxxxxxxxx>, Hemant-K.Chitale@xxxxxx
- Date: Thu, 3 Dec 2015 09:35:22 +0100 (CET)
Hi Hemant,
Can anyone interpret that note and help me understand that statistics ?
Oracle Patent #US20140258223-A1 got an explanation of it. Just quoting from it:
"Redo synchronization time and overhead. The latency observed by the process
committing the transaction. It is the sum of how long it takes the log
writing process to receive the log write request, the redo log write time, and
the redo synchronization overhead. The overhead is the time between the
completion of the log write and the time at which the committing process
detects that the log write is done, and consequently reflects the latency
added by post-wait or poll-wait. Thus, the ideal redo synchronization time is
the redo synchronization time minus the overhead."
More internals about post-wait or poll-wait mechanism can also be found in
Frit's slides (21 to 23):
https://fritshoogland.files.wordpress.com/2015/09/profiling-the-logwriter-and-database-writer.pdf
Best Regards
Stefan Koehler
Freelance Oracle performance consultant and researcher
Homepage:
http://www.soocs.de
Twitter: @OracleSK
"Chitale, Hemant K" <Hemant-K.Chitale@xxxxxx> hat am 3. Dezember 2015 um
04:28 geschrieben:
Apparently, v$sysstat does have statistics like
redo synch time (usec)
redo synch time overhead (usec)
redo synch time overhead count (<128 msec)
redo synch time overhead count (<2 msec)
redo synch time overhead count (<32 msec)
redo synch time overhead count (<8 msec)
redo synch time overhead count (>=128 msec)
Oracle Support note “Bug 12614085 - Diagnostic enhancement to add new
statistics for investigating "log file sync" and "log file parallel write"
relationship (Doc ID 12614085.8)” talks of an enhancement that enables this
collection but I am not sure I understand it.
Can anyone interpret that note and help me understand that statistics ? If
I see the overhead time being reported by these statistics, what really
(what overhead) does it represent ? What if we have DataGuard in the picture
? Does it also reflect DataGuard LGWR SYNC overheads ?
Hemant K Chitale
--
//www.freelists.org/webpage/oracle-l
Other related posts:
