FETCH -- what exactly is it?

I am in the process of designing some performance benchmarks for Oracle V7.0.15 over SQLNet1. Part of my data capture is the result of an 'ALTER SESSION SET SQL_TRACE = TRUE' and the resulting '*.trc' file.

I will be running the test scripts on the database server
*and* on a remote client.

In my preliminary testing to see if what I have wrought actually works, (and to see how well tkprof works) I have noticed something very interesting.

When I compare the server test results (TWO_TASK is not set), the number of "Fetch"es done is about 7 for every 100 rows returned.

However, when I review the client test (TWO_TASK is set to the host:database) results, I find that the count of the "Fetch"es is actually one greater than the number of rows returned.

Upon further investigation, I found that the number of packets that contain the response to the initial select statment is
*exactly* the same number of "Fetch" count. (This will have
a very negative impact on the performance of distributed Client Server application based upon Oracle/SQLNet v1. if the number is not a coincidence.)

Does anyone have an explanation of what exactly is the "Fetch" count? And what is the relationship of "Fetch" to SQLNet?

Thanks,

Tom Hackwood
thack_at_antares.Tymnet.COM Received on Wed Jul 06 1994 - 03:51:04 CEST