Re: Q: V$SYS_TIME_MODEL/v_$sysstat in 12C (pluggable)
From: Hans Forbrich <fuzzy.graybeard_at_gmail.com>
Date: Sun, 29 Dec 2013 08:12:12 -0700
Message-ID: <52C03BCC.6050408_at_gmail.com>
On 29/12/2013 6:01 AM, Amihay Gonen wrote:
> Where I can see the numbers are near but not summed up .
>
> For example db time [0]= 3,991,739 <> 3,987,789+13,352+13,352+9,402 =
> 4,023,895
>
>
Are you taking into account the fact that V$ views look at real memory in real time and do not follow the Consistent Read model?
Date: Sun, 29 Dec 2013 08:12:12 -0700
Message-ID: <52C03BCC.6050408_at_gmail.com>
On 29/12/2013 6:01 AM, Amihay Gonen wrote:
> Where I can see the numbers are near but not summed up .
>
> For example db time [0]= 3,991,739 <> 3,987,789+13,352+13,352+9,402 =
> 4,023,895
>
>
Are you taking into account the fact that V$ views look at real memory in real time and do not follow the Consistent Read model?
In other words, it is quite possible for some small variations to occur simply because the V$ views (X$ tables, and the associated code) query each memory location in real time, while the system is actually performing operations.
My experience is that the 'drift' in the V$/X$ information is exacerbated in certain environments, such as virtualization. (You do not mention your environment in detail.) I tend to be surprised when the numbers actually 'add up exactly'.
/Hans
-- http://www.freelists.org/webpage/oracle-lReceived on Sun Dec 29 2013 - 16:12:12 CET