Hi
You always got two IDE controllers which are separated from each other
(even not very old boards); the IDE
controllers are "occupied" by the PCI bus system, which also controls
PCI slots and ISA bridge, also the
data flow from the controllers to the memory (DMA use).
As I said before (and I still think I am right in a way) the bandwidth
of the data flow between memory and
IDE controllers is limited by the speed of the PCI bridge and the IDE
controllers themselves.
So you always got a maximum of throughput, where can get above - SCSI
controllers show the same behaviour: if you got a 10MB/s controller you
will never get 15MB/s out of, if you got an UW (40MB/s) you will never
get 50MB/s out of it.
Greetings, Dietmar
Joachim Zobel wrote:
>
> At 23:17 23.05.99 +0200, you wrote:
> >
> >raid-1 also increases read performance. It can do reads just like raid0,
> >because both disks contain the same data. It doesn't read the same block
> >from both disks and compare, instead it reads like from a stripe set (raid0).
> >
> >At least that's the theory. Your points about several drives on the same
> >bus etc. are ofcourse valid. If two IDE disks are put on _separate_ busses
> >and the IDE chipset doesn't stink too much, you should see good read
> performance
> >both from raid0 and raid1.
>
> Seperate buses? How do i find out? Seperate ide controllers is not enough?
>
> Probably i do not have seperate buses. My raid0 device shows exactly the
> same behaviour if i test it with hdparam.
>
> Thanx,
> Joachim
>
> --
> "... ein Geschlecht erfinderischer Zwerge, die fuer alles gemietet werden
> koennen." - Bertolt Brecht - Leben des Galilei
--
"Wissen ist ein Transportproblem" W. Roesler
http://home.t-online.de/home/dstein2203
[EMAIL PROTECTED] [EMAIL PROTECTED]
Systemadministrator UNIX / Linux
