Re: Database Block Size

Steve,

> If you use a 16K database block size over an 8K file system buffer size then all
> single block I/O requests will be serviced by the operating system as two 8K I/O
> operations which will be performed in series. That is, the second 8K operation
> will not be started until the first has completed. There are big inefficiencies

I totally agree with you about the above. A 16K request will be serviced by two I/O operations and they will be performed serially. I can think of a few occasions where two I/O operations will be performed in parallel, but not for Oracle blocks.

> if the disk is busy enough for another request to be interposed, then you may
> lose the contents of the track buffer and sustain two seeks before the I/O
> request can be completed.

Here is where you make an assumption which is where I had a problem with the word "must" as quoted by a previous posting (not from you, but quoting you). This assumption is that the disk is busy enough and that the system will interrupt a 16K disk read in order to service another disk read on the same device. While this assumption can be a safe one under most circumstances, it is not always the case. It all depends on the individual system. And for that reason, I had a problem with the word "must". I felt that it was too strong of a word.

But all in all, it was a nice little discussion.

Nice talking to you!
Brian

-- 
========================================
Brian Peasland
Raytheons Systems at
  USGS EROS Data Center
These opinions are my own and do not
necessarily reflect the opinions of my 
company!
========================================