Re: Looking for Oracle server

In article <dtb.732498052_at_otto>, dtb_at_otto (David Bath) writes:
|> szhu_at_prism.poly.edu (Shu Bin Zhu) writes:
|>
|> >We need to decide on a server to run oracle. The choice seems to narrow
|> >down to R/S6000 570 and the NCR 3450. I don't have the SPEC mark for
|> >the NCR 3450. Does anyone know which is a better machine. The NCR 3450
|> >has two 486-50 cpus.
|>
|> The UNIX on the RS6000 series is stuffed primarily because it has a non-
|> standard malloc(). Instead of getting the memory and giving you a pointer
|> it gives you a pointer and then when you start using the memory, it
|> then allocates it. Of course, the memory may not be there (you may have
|> asked for 200 Zigabytes), and then it send a non-standard signal to your
|> process.
|>
|> Thus, many standard UNIX programs need to be reworked in the signal handling
|> parts and this is among the trickiest areas of applications programming to
|> work in. Try other UNIX related newsgroups to get other's opinions, but
|> the general consensus is that while the RS series perform well, the core
|> of the OS is brain damaged, primarily because of the malloc.

I also seem to recollect being told by one of Oracle's techies not so long ago that if you do go RS6000 you _must_ use raw partitions as the OSYNC flag is not supported (ie you can't guarantee a synchronous write).

|>
|> [ I guess you do not want all your programmers wondering why their
|> programs "blue" up :-) ]
|>
|>
|> --
|> David T. Bath | Email:dtb_at_otto.bf.rmit.oz.au (131.170.40.10)
|> Senior Tech Consultant | Phone: +61 3 347-7511 TZ=AEST-10AEDST-11
|> Global Technology Group | 179 Grattan St, Carlton, Vic, 3153, AUSTRALIA
|> "The robber of your free will does not exist" - Epictetus
 

-- 
                               John Peach
                           Chevron (UK) Ltd.
   Ninian House, Crawpeel Road, Altens, Aberdeen, AB1 4LG, Scotland.
   Internet: epeas_at_aberdeen.chevron.com       Phone: +44 224 242637