David
Mon May 03 15:32:48 CDT 2004
First, look at the definition of Raid 0 at
http://www.acnc.com/raid.html.
Second, remember that INT 13h does not have any knowledge of Raid ?
unless it is a special BIOS for a specific Raid controller. While that
could be embedded within a system BIOS it would require a work to
validate the hardware on each boot. What happens if the READ_CAPACITY
of the two or more drives are different? How do you handle when one
drive has exhausted all its spares and the drive has to reduce the total
number of sectors reported via READ_CAPACITY to indicate how many
sectors it has that work?
Third, Raid 0 does increase your failure probability by the number of
drives in the Raid 0 configuration. Two drive doubles, but four drive
quadruples the possibility of failure.
Software Raid 0 increases the total size of the 'single' hard drive as
it appears to the OS. If the drives are on channels that can run at the
same time, it can increase the speed, but you usually see this with SCSI
controllers where you have not only multiple drives, but multiple
controllers. I think SATA could provide some of the same speed
capabilities, but I don't know if the SATA controllers are normally
duplicated or just the same controller on multiple drives.
P.S. I think Raid 0 should be avoided. Bad idea! Raid 1 for data
protection makes a lot of sense. Raid 5 for capacity, speed and data
protection are indicated for servers, especially if multiple IO requests
can be sent to the various drives for processing before the previous
requests are completed.
Peter is an expert at the problems with storage. His logic makes sense.
Some of your ideas are probably implemented by various SAN vendors to
reduce cost, but the PC is not a mainframe. You can get close, but the
costs become too high, while mainframe buyers are used to being thought
of by their suppliers as a bank.
"Alex" <AlX@a> wrote in message
news:ea623wTMEHA.2252@tk2msftngp13.phx.gbl...
> No it's not easy, but, what the hell are you talking about?
> The boot time MBR loader dosn't have anything to do with any WDM
thingy?!
> You simply hook the int13, make sure you report one 1 drive, and
handle all
> the reads/writes using the original int13.
>
> It's not trivial but it's not rocket science ether.
> It's duabale and if done properly it should work just fine.
>
> But what I don't know is, is there any merit to a software raid0? I
mean,
> will you see any signifiant increase in speed?
>
>
> "Peter Wieland [MSFT]" <peterwie@online.microsoft.com> wrote in
message
> news:%2307MLmTMEHA.2976@TK2MSFTNGP10.phx.gbl...
> > Sure it's possible - just about anything is possible. But you make
it
> sound
> > like a trivial piece of work to take a big chunk of the WDM storage
stack
> > and compile it into a real-mode component that can replace the bios
> > functionality.
> >
> > It would be a lot of work, and it would require a lot of testing to
make
> > sure it worked right in all scenarios and with a significant set of
system
> > BIOSs (which have all sorts of interesting behaviors). And changing
the
> > boot loader risks introducing new bugs in a component that must work
right
> > on every Windows box in order for anything else to run.
> >
> > I'm not commenting on the rightness or wrongness of Windows not
supporting
> > RAID-0 system disks ... that's a question the storage team needs to
> evaluate
> > based on customer demand, development cost and overall risk. I'm
just
> > trying to point out that features aren't always as simple to add as
one
> > might think.
> >
> > -p
> >
> > --
> > This posting is provided "AS IS" with no warranties, and confers no
> rights.
> >
> > "Alex" <AlX@a> wrote in message
> > news:OGxzkbTMEHA.1468@TK2MSFTNGP12.phx.gbl...
> > > What if you combine the stripping driver for windows, with a boot
time
> > > BIOS
> > > hook. The BIOS hook can be implemented with a replacement for the
MBR
> > > loader.
> > >
> > > "andrea catto'" <acatto@dataflight.com> wrote in message
> > > news:%23dfeyzGMEHA.2792@TK2MSFTNGP10.phx.gbl...
> > >> My topic here is to discuss why microsoft does not implement the
> ability
> > > to
> > >> utilize at OS-Setup two disks as raid0/striping.
> > >>
> > >> Yes I know, raid0 is not safe, or at least is just as safe a a
1-disk
> > > which
> > >> is the most popular solution for 99% of the users in the desktop
> > > computing.
> > >>
> > >> So why after all not allowing this ???
> > >>
> > >> I think users should be allowed doing this, assuming the risks on
their
> > > own.
> > >>
> > >> the fact of the matter is that nowadays everybody may afford a
couple
> of
> > >> hard drives and if put on different IDE channels/perhaps
controllers
> they
> > >> could 'fly' with raid0.
> > >>
> > >> This limitation indeed exixts within the Redhat/Fedora
implementations
> > >> too
> > >> as well as others, but just because of 'lack of care' for those
> > > requirements
> > >> that many users would appreciate instead if given the chance.
> > >>
> > >> Slackware Linux on the other hand SHOWS that it is possible to in
fact
> > > have
> > >> two separate devices paired as raid0 mounted on / (equivalent to
c:)
> > >>
> > >> This raid0/striping on boot/root is way possible, I am a linux
guru
> too,
> > >> although I love and prefer Microsoft, and since it's possible
because I
> > > made
> > >> it and used it, I am wondering if this limitation is made perhaps
> because
> > >> the microsoft's disk structure/software could not handle this at
all
> > > unlike
> > >> linux.
> > >>
> > >>
> > >
> > >
> >
> >
>
>