X

An Oracle blog about Oracle Cloud

  • May 24, 2017

"physical read total multi block requests"

Abel Macias
Senior Principal Support Engineer

During an escalation where WORKLOAD statistics end up being corrupted and producing aberrant execution plans I was made aware of a blog post about to "Never gather WORKLOAD stats on Exadata…" .
The blog post goes ahead to identify that the statistic "physical read total multi block requests" does not appears to count correctly the multi-block requests.
The discussion continued on the Oracle Community Forum .

I wanted to know what this statistic really measures.

The Documentation of the statistic says :

Total number of Oracle instance read requests which read in two or more database blocks per request for all instance activity including application, backup and recovery, and other utilities.

But as I dig deeper I realized the description is wrong and filed Bug 26136778 - desc of "physical read total multi block requests" is wrong

My interpretation of the statistic is :

Total number of Oracle instance read requests which read 128kBytes or more per request for all instance activity including application, backup and recovery, and other utilities.

And this causes also the interpretation of
"physical read total IO requests" to change as well

Current Description :

Number of read requests which read one or more database blocks for all instance activity including application, backup and recovery, and other utilities. The difference between this value and "physical read total multi block requests" gives the total number of single block read requests.

My interpretation of the statistic is :

Number of read requests which read one or more database blocks for all instance activity including application, backup and recovery, and other utilities. The difference between this value and "physical read total multi block requests" gives the total number of "Small" IO requests which are less than 128kBytes down to single block read requests.

We'll have to wait for Oracle Dev to review the bug and give its ruling.

Be the first to comment

Comments ( 0 )
Please enter your name.Please provide a valid email address.Please enter a comment.CAPTCHA challenge response provided was incorrect. Please try again.