RE: Oracle issue with Ultrasparc CPU's

NUMA as in Sequent NUMA (non-uniform memory architecture, or maybe it is access). There is more to server performance that just the CPU speed. Our old machine had a very fast buss however the vendor put everything across that one buss. Everyting being both memory access and IO. Having a separate IO buss and memory buss means two slower speed dedicated busses may well outperm a higher speed but over-burdened buss.

A reduction in disk platters may not have the negative effect on IO that you might expect when you consider that it was not possible to write to all the disks at once. The disks were more or less daisy chained together so that you could only write to one disk in a chain at a time. Hence the reduction in the absolute difference in number of platters has much less of a negative effect than you might expect. Then throw cache in front of the new disk, plus the new disk probably ran at a higher RPM and you get more IO's per second from less disk platters.

The devil is in the details that most of us do not have time or the interest to really get to understand. Besides just like Oracle they change.

-----Original Message-----
From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org]
Sent: Tuesday, July 18, 2006 4:13 PM
To: oracle-l_at_freelists.org
Subject: RE: Oracle issue with Ultrasparc CPU's

 >>>

oops...I meant wouldn't...Could it be

--



--
http://www.freelists.org/webpage/oracle-l