Re: Myth revisited ...

From: Geomancer <pharfromhome_at_hotmail.com>
Date: 19 Nov 2003 08:14:46 -0800
Message-ID: <cf90fb89.0311190814.6e87a537@posting.google.com>

Yes, like everything on Oracle "it depends". . . .

However, can we assume?

The on-the-margin cost of reading a 32k blocks vs. an 8k block is VERY SMALL. (I can read a 4k block in 15ms, a 32k blocks in 16ms) Hence we assert that thjere is less letency is reading data in 32k chunks, BUT ONLY IF WE NEED THE ADJACENT ROWS.
It is a waster of RAM to read 32k to get a 30 byte row.

3 - OPS and RAC (according to experts) like SMALL blocksizes, to increase granularity and reduce IDLM, and CF pinging. Received on Wed Nov 19 2003 - 10:14:46 CST