I thought RAID 0 was for 1 drive


Bob Eyster



"Philip Andrews" <philipfaeunst@tiscali.co.uk> wrote in message
news:uwWEZ$xCGHA.336@TK2MSFTNGP14.phx.gbl...
> Hi Bob,
>
> "Bob Eyster" <reyster1@comcast.net> wrote in message
> news:OmH7dNvCGHA.1312@TK2MSFTNGP09.phx.gbl...
>> Set the SATA HDD to RAID 0 and see what happens.
>
> Can't do it - I only have one SATA hard-drive, RAID calls for at least
> two.
>
> Regards,
>
> Philip
>

Bob Eyster wrote:

> I thought RAID 0 was for 1 drive


No, raid 0 requires two or more drives. It's called striping. It puts pieces
of files alternately on the several drives so you don't have to wait for the
first read or write to finish before starting the next. It's for increased
performance, not redundancy.

Read, for example, here: http://www.acnc.com/raid.html


--
Ken Blake - Microsoft MVP Windows: Shell/User
Please reply to the newsgroup




> "Philip Andrews" <philipfaeunst@tiscali.co.uk> wrote in message
> news:uwWEZ$xCGHA.336@TK2MSFTNGP14.phx.gbl...
>> Hi Bob,
>>
>> "Bob Eyster" <reyster1@comcast.net> wrote in message
>> news:OmH7dNvCGHA.1312@TK2MSFTNGP09.phx.gbl...
>>> Set the SATA HDD to RAID 0 and see what happens.
>>
>> Can't do it - I only have one SATA hard-drive, RAID calls for at
>> least two.
>>
>> Regards,
>>
>> Philip

"Ken Blake, MVP" wrote:
> No, raid 0 requires two or more drives. It's called striping.
> It puts pieces of files alternately on the several drives so
> you don't have to wait for the first read or write to finish
> before starting the next. It's for increased performance,
> not redundancy.


More accurately, RAID 0 does I/O faster because the
pieces are being done at the same time to more than one
drive, the read or write of each piece being completed
on its particular drive at about the same time as all the
other pieces. The speed comes from parallelism, not
from overlapping beginnings and ends of I/O tasks.

*TimDaniels*

Timothy Daniels wrote:

> "Ken Blake, MVP" wrote:
>> No, raid 0 requires two or more drives. It's called striping.
>> It puts pieces of files alternately on the several drives so
>> you don't have to wait for the first read or write to finish
>> before starting the next. It's for increased performance,
>> not redundancy.
>
>
> More accurately, RAID 0 does I/O faster because the
> pieces are being done at the same time to more than one
> drive, the read or write of each piece being completed
> on its particular drive at about the same time as all the
> other pieces. The speed comes from parallelism, not
> from overlapping beginnings and ends of I/O tasks.


Well, you said it differently than I did, but I'm not sure what you're
describing is any different . Overlapping beginnings and endings achieves
parallelism.

--
Ken Blake - Microsoft MVP Windows: Shell/User
Please reply to the newsgroup

"Ken Blake, MVP" wrote:
> Timothy Daniels wrote:
>
>> "Ken Blake, MVP" wrote:
>>> No, raid 0 requires two or more drives. It's called striping.
>>> It puts pieces of files alternately on the several drives so
>>> you don't have to wait for the first read or write to finish
>>> before starting the next. It's for increased performance,
>>> not redundancy.
>>
>>
>> More accurately, RAID 0 does I/O faster because the
>> pieces are being done at the same time to more than one
>> drive, the read or write of each piece being completed
>> on its particular drive at about the same time as all the
>> other pieces. The speed comes from parallelism, not
>> from overlapping beginnings and ends of I/O tasks.
>
>
> Well, you said it differently than I did, but I'm not sure what
> you're describing is any different . Overlapping beginnings
> and endings achieves parallelism.


If by "read" and "write" you meant the individual I/O
commands to each drive to transfer a segment of the file
to a particular drive, yes, the individual I/O operations do
overlap (100%), and any wait for the I/Os of other segments
of a group to complete (after any one of them completes)
before starting the I/Os for the following segment group is
very short or nonexistent. That is, the I/O of segment members
of a doublet (or triplet or qudruplet, etc., depending on the
number of drives RAIDed) can be done simultaneously.
But RAID striping does not implement the overlap of the OS's
high-level READ FILE or WRITE FILE instructions. I *believe*
that those must still be done synchronously, i.e. a high-level
WRITE FILE must complete before a READ FILE on the
same file can commence. Please correct me if I am wrong.

*TimDaniels*

Hello all,

"Timothy Daniels" <TDaniels@NoSpamDot.com> wrote in message
news:WPednboF4bU0DSjeRVn-jw@comcast.com...
> "Ken Blake, MVP" wrote:
> > Timothy Daniels wrote:
> >
> >> "Ken Blake, MVP" wrote:
> >>> No, raid 0 requires two or more drives. It's called striping.
> >>> It puts pieces of files alternately on the several drives so
> >>> you don't have to wait for the first read or write to finish
> >>> before starting the next. It's for increased performance,
> >>> not redundancy.

> >> More accurately, RAID 0 does I/O faster because the
> >> pieces are being done at the same time to more than one
> >> drive, the read or write of each piece being completed
> >> on its particular drive at about the same time as all the
> >> other pieces. The speed comes from parallelism, not
> >> from overlapping beginnings and ends of I/O tasks.

> > Well, you said it differently than I did, but I'm not sure what
> > you're describing is any different . Overlapping beginnings
> > and endings achieves parallelism.

> If by "read" and "write" you meant the individual I/O
> commands to each drive to transfer a segment of the file
> to a particular drive, yes, the individual I/O operations do
> overlap (100%), and any wait for the I/Os of other segments
> of a group to complete (after any one of them completes)
> before starting the I/Os for the following segment group is
> very short or nonexistent. That is, the I/O of segment members
> of a doublet (or triplet or qudruplet, etc., depending on the
> number of drives RAIDed) can be done simultaneously.
> But RAID striping does not implement the overlap of the OS's
> high-level READ FILE or WRITE FILE instructions. I *believe*
> that those must still be done synchronously, i.e. a high-level
> WRITE FILE must complete before a READ FILE on the
> same file can commence. Please correct me if I am wrong.

> *TimDaniels*

Thanks for all the input - but I finally resolved the problem by
accident, after Skype BETA (webcam) nearly wrote my system off four days
ago.

The problem of not being able to 'see' the SATA HDD apparently lay in an
incorrect BIOS setting, and when I adopted the FAILSAFE DEFAULTS in order to
get my machine back on its feet, I got my main drive back and 'discovered'
the SATA drive as well (as a bonus). After running Disk Management to format
and assign a leter to the new drive, I unloaded the RAID drivers and JAVA
RAID Console - they aren't required when SATA drives are used as individual
IDE drives - and all is now well.

Comments in passing:-

- Adding the 150 MB/s SATA drive has noticeably increased the rate of
throughput on audio- and video-file processing, by a factor of about half as
fast again as was previously possible with my ATA 100 IDE drives. 150 MB/s
definitely seems to be the way to go;

- I still can't see the point of using RAID 0 at all: because although
it may increase system speed by reducing the HDD-write 'bottleneck', the
loss of one drive in the array would mean a total loss of stored data on all
of the drives. Surely, RAID 0 must actually serve to make a system more
fragile, unless I've missed something?;

- The sooner the motherboard makers (Foxconn, in this case) include a
few paragraphs of specific information (see below) to their Mainboard
Manuals on how to connect SATA drives in a non-RAID manner, the faster
they'll hasten the revolution on SATA usage while reducing the unnecessary
load on their tech staff (and on Newsgroups such as this one).

My problem - which caused me more than ten working days of difficulties, an
unnecessary near-return-to-supplier of the drive, and which still wouldn't
have been resolved unless the PC had failed in the meantime for other
(faulty-software-related reasons - Skype, step forward for a
bottom-smacking, with your lousy webcam-ready BETA download) would have been
simplicity itself to avoid given the inclusion in the motherboard manual of
the following information . My Foxconn board features the Award Phoenix
BIOS, but equivalent instructions would be just as valid for setting up an
AMI BIOS that might be in use on any other maker's SATA-ready mainboard:-


" UP TO to four SATA drives may be used as individual IDE drives with
this motherboard. In this mode, they will be identified in the STANDARD CMOS
FEATURES section of the BIOS as the Channel 3 and Channel 4 Master and Slave
IDE drives respectively, depending on which of the motherboard SATA
socket(s) it/they are connected to.

To install it/them, proceed as follows:

1) Install and secure the SATA drive(s) using appropriately-sized
mounting-screws;

2) Connect the SATA drive(s) to its/their data and power cables;

3) If a known-defective SATA drive has just been replaced, and if you're
quite sure that the BIOS settings are correct, go straight to para (8) below
from here. If you wish to check and confirm the BIOS settings, or if a new
SATA drive (or drives), have been added to the machine for the first time,
boot the PC and then use the <DEL> key to obtain the BIOS settings screen;

4) Select the 'INTEGRATED PERIPHERALS' section, and make sure that IDE HDD
Block Mode is enabled;

5) In the 'OnChip PCI DEVICE' section of 'INTEGRATED PERIPHERALS', make sure
that Serial ATA Controller is enabled, and that Serial ATA Mode is set to
IDE (both settings are defaults, but should be checked and verified);

6) In the 'STANDARD CMOS FEATURES' section, confirm that the SATA drive(s)
is/are now visible;

7) Save the changes to the BIOS settings and exit the BIOS setup. The BIOS
settings will declare the new drive(s) to Windows, and make it/them
available for use;

8) Boot the machine into Windows XP, wait until it settles, and then click
on 'Start ... Control Panel ... Administrative Tools'.

9) Double-click on the 'Computer Management' icon and then select the Disk
Management section of 'Storage';

10) Right-click the the new drive and select 'Format' to format it to match
the file-system that's already in use on your existing IDE drive(s) (this
should preferably be NTFS under Windows XP, but Windows XP will also accept
the FAT32 file-system format);

11) Assign a suitable letter to the new drive that allows Windows a degree
of flexibility:- for example, 'Z' may be appropriate in the case of a single
SATA drive that's dedicated for use as a video-project scratchpad, in a
machine that already features an IDE drive-caddy as well as two fixed,
multiple-partition IDE drives;

12) Close 'Component Manager', open 'My Computer' and see that the drive(s)
is/are now present and visible, and can accept data (by copying a few small
files across to it/them, and then logging-on to the drive(s) to make sure
that the files have arrived);

13) Run SCANDISK THOROUGH on the new drive(s) to make sure that no bad
sectors are found: you should untick the 'Fix Errors' box beforehand, to
make sure that Windows reports any faults that it finds rather than simply
fixing them and moving on. A hard-drive 'surface' must be perfect from the
outset, or the drive in question will fail early and will probably take your
data with it if its ailment isn't recognised in time;

14) If no surface errors have been found, the SATA drive(s) is/are now ready
for use - otherwise, replace the defective drive(s) and start again at para
(1) above;

15) If drive-formatting or drive-lettering problems are encountered, reboot
the PC, go into the BIOS screen, select 'LOAD OPTIMISED (or FAILSAFE)
DEFAULTS' and then complete this setup procedure from paras (4) to (14), to
determine whether the drive itself or the BIOS settings are the cause of the
trouble. If a BIOS setting unexpectedly turns out to be the cause, go back
into the BIOS when the problem has been resolved and make whatever minor
adjustments might be necessary to allow the PC's hardware to work at its
best - and note them down somewhere. "


I may be biased of course - but given that style of instruction or
something very similar, it's my conviction that NO-ONE could possibly have
setup problems with SATA hard-drives being used singly or as independent,
non-RAID units. Simple, aren't they? - once you get past the stilted style
of prose, of course.


Is that the sort of thing you meant by 'setup instructions', Anna?


Regards,

Philip Andrews

"Philip Andrews" wrote:
> - I still can't see the point of using RAID 0 at all: because
> although it may increase system speed by reducing the
> HDD-write 'bottleneck', the loss of one drive in the array
> would mean a total loss of stored data on all of the drives.
> Surely, RAID 0 must actually serve to make a system more
> fragile, unless I've missed something?;


Yes, RAID 0 does roughly double the probability of data
loss by hard drive failure. But RAID 0 is for speed. RAID 1
(mirroring) is for reliability. See the RAID Primer at:
http://www.finitesystems.com/PRODUCT/raid/raidlevel.htm .


*TimDaniels*

Philip Andrews wrote:

> - I still can't see the point of using RAID 0 at all: because
> although it may increase system speed by reducing the HDD-write
> 'bottleneck', the loss of one drive in the array would mean a total
> loss of stored data on all of the drives. Surely, RAID 0 must
> actually serve to make a system more fragile, unless I've missed
> something?;


You haven't missed anything at all. Yes, RAID 0 increases speed, but hurts
reliability.

However, that doesn't mean RAID 0 has no point. Most people think that
increasing speed is a valuable thing to do. However, if you use it, the
always-present need for backup becomes even greater.

--
Ken Blake - Microsoft MVP Windows: Shell/User
Please reply to the newsgroup

On Sat, 31 Dec 2005 23:36:21 -0000, "Philip Andrews"

> - I still can't see the point of using RAID 0 at all: because although
>it may increase system speed by reducing the HDD-write 'bottleneck', the
>loss of one drive in the array would mean a total loss of stored data on all
>of the drives. Surely, RAID 0 must actually serve to make a system more
>fragile, unless I've missed something?

No, you missed nothing; RAID 0 is purely a performance thing, it has
no positive value (and indeed, much negative value) in survivability.

I generally adopt RAID 0 only when it becomes the cheapest way to
attain a high capacity data workspace, as is often the case with video
editing etc. A RAID 0 of two small HDs doesn't make much sense, if
the same capacaity is cheaper as a single HD.

There may be additional speed benefits, aside from strapping two HDs
together so as to double the data capacity per cylinder (as buying a
single twice-as-large HD would likely do, too). Not sure on that, or
how applicable they are (i.e. whether something that looks brilliant
on benchmarks will necessarily translate well to real-world use)

> - The sooner the motherboard makers (Foxconn, in this case) include a
>few paragraphs of specific information (see below) to their Mainboard
>Manuals on how to connect SATA drives in a non-RAID manner, the faster
>they'll hasten the revolution on SATA usage while reducing the unnecessary
>load on their tech staff (and on Newsgroups such as this one).

Um... I've been building with native (non-RAID) S-ATA for a while now,
and I don't remember any issues. In general, you are right, in that
practical documentation of CMOS settings is rare. Invariably you get
"help" like "Spread Spectrum: Disable - this disables Spread
Spectrum"; there's no task-orientated (e.g. "how to set up with S-ATA
as first bootable HD") or background info (e.g. "what the hell is
Spread Spectrum, anyway?") help.

In particular, S-ATA vs."legacy" IDE is a bitch. We really need
proper documentation on how these interact under the various modes,
e.g. "legacy", "enhanced" and so on. If you have S-ATA HDs, an "IDE"
optical drive, and an IDE HD you don't want on the same channel as the
optical drive, things get very messy, very quickly.

It seems as if two S-ATA will overlay each IDE controller, so that you
can't (say) use a primary IDE plus S-ATA 0 or 1 at the same time.
There can be flakiness on boot order vs. enumeration order that can
cause XP to "see" the non-boot HD "first", etc.

And all that is before you get to RAID issues...



>---------- ----- ---- --- -- - - - -
Don't pay malware vendors - boycott Sony
>---------- ----- ---- --- -- - - - -

On Sat, 31 Dec 2005 23:36:21 -0000, "Philip Andrews"
<philipfaeunst@tiscali.co.uk> wrote:

> - I still can't see the point of using RAID 0 at all: because although
>it may increase system speed by reducing the HDD-write 'bottleneck', the
>loss of one drive in the array would mean a total loss of stored data on all
>of the drives. Surely, RAID 0 must actually serve to make a system more
>fragile, unless I've missed something?;

While you are correct, how often does a drive fail?
--
Top 10 Conservative Idiots:
http://www.democraticunderground.com/top10/