At 04:54 PM 6/09/2007, you wrote:

>I see.
>In our case, as i had earlier mentioned, each query would be within
>its own transaction. When we execute the same query for say 40
>times, the number of i/o reads is comparabable to the number of i/o
>writes. However, when the loop goes to around 10,000, the number of
>i/o writes is very much larger than the number of i/o reads.
>Any clue?

The number of disk reads will diminish as you progress through the
loop, since each data page and index page is cached in memory when it
is read for the first time, and stays there until it changes or its
memory space is needed for newer page-reads.

./hb