FreeLists / oracle-l / Re: *Measuring sql performance (elapsed time and scalability) by number of logical reads

["Morten Egan" <meg@xxxxxxxxxxxx>]

A logical IO is not a logical IO :)

There are many diff. types of logical IO in oracle, and many of them involve 
completely different code paths, taking diff. amounts of time to complete.

Regards,
Morten Egan
-----Original message-----
From: "Radoulov, Dimitre" cichomitiko@xxxxxxxxx
Date: Tue,  2 May 2006 14:08:53 +0200
To: "Bernard Polarski" bpolarsk@xxxxxxxxx
Subject: Re: *Measuring sql performance (elapsed time and scalability) by 
number of logical reads

> >I think you have been a bit short in the problem description.
> >
> > You just meant that all the requested data is already in buffer and no 
> > physical read is needed.
> > Thanks but we have no information on the nature of the sql, the amount of 
> > data, the expected goal.
> > Bad or good SQL is a ratio of  these. What if I read one million blocks 
> > from my multi gig db block buffer
> > to return a tiny rowset for the worse ever seen SQL,  it will satisfy your 
> > prerequisite and still be very bad.
> 
> Excuse me for not being clear, I meant, theoretically speaking:
> 
> SQL 1 reads n1 blocks from buffer (no physical read) to complete, elapsed 
> time t1
> SQL 2 reads n2 (where n2 > n1) blocks from buffer (no physical read) to 
> complete, elapsed time t2
> 
> t1 is greater than t2
> 
> Always theoretically/hypothetical speaking:
> could anyone comment the possibile reasons behind such behaviour.
> 
> 
> 
> Regards,
> Dimitre
> 
> 
> 
> --
> http://www.freelists.org/webpage/oracle-l
> 
> 
> 
--
http://www.freelists.org/webpage/oracle-l