Re: OPS on NT (was: Native Oracle-port on Linux)

Extracted from what Billy Verreynne wrote:  

> The local Oracle Support tech team leader told me that they did an OPS
> installation with NT. They used Wolfpack and some OEM version of NT and it
> seems to be working fine. This raises interesting possibilities with NT. My
> biggest beef with NT has always been the lack of scalability - you need to
> purchase special OEM versions of NT to scale it to more than 4 processors.
> And I was doubtful if Wolfpack was an answer. However, if Wolfpack does work
> and can run Oracle Parallel Server (OPS) then Microsoft may have a very
> competitive product for large systems. A cluster of 200 Intel Pentium II
> machines is (almost?) equal to a supercomputer in raw power, and should be
> much cheaper than a 200 node MPP system from a Unix vendor. Imagine running
> OPS on that!
>
> But then again, there are questions about how stable and bugfree the initial
> versions Wolfpack will be and how easy will it be to administer a NT cluster
> and 200 Oracle OPS instances. The NT commandline sucks and there are no
> standard telnet with NT.

How about a 64 node parallel Linux machine like the Beowulf boxes that are running at Los Alamos National Laboratory and Caltech? After all 64 Pentiums running Linux should be able to perform about as well as 200 running NT :-). There is no way you are going to beat either the hardware or OS price. See the January 1998 issue of Linux Journal for more information on these systems.

I believe that the major hurdle is that a department in even the largest organization that puts together multi-processor Linux boxes is not very likely to buy Oracle (or any other commercial RDBMS). As a result I doubt if we will ever see SUPPORTED Oracle on a Linux box -- and, in the business world, if you can't get support from the vendor -- don't buy the product :-(

Jerry Received on Fri Dec 19 1997 - 00:00:00 CST