Announcement

**Jammrock** · 23 July 2001, 05:41

Apparently there was a Q&A issue with the 75GXP's. Just pray they send you a 60GXP back. Those babies are rock solid...so far *knocks on wood*

Jammrock

**Paulr** · 23 July 2001, 06:01

Talking about IBM Hard Drives.

At home I used to have 4 of the 45GB 75GXP units - not a single failure.
At work I've rolled out some 30+ workstations all with 75GXP drives of various sizes out and had one single failure - PC dropped.

Last week I had a 60GB 60GXP drive fail on me.
The RAID controller detected a drive failed, re-booted found the missing drive and rebuilt the RAID.
Switched off, switched on later, drive lost, re-booted drive found & RAID rebuilt.
Took drive back to where I bought it, they ran the IBM tests on it and it immediately came up with 'Failure - return to IBM'.
It's funny, near perfect reliability with the 75GXP's and a fail on a 60GXP after only a month.

All I can say is thank god for RAID 0+1.
So OK, my 240GBof HD's only give my 120GB of useable space, but I can loose a combination of 2 HD's and no data loss, if that had been on a RAID 0 array I'd have lost everything!

**Ali** · 23 July 2001, 15:40

Hey, Paulr, how do you like raid 0+1?

If got 2 60GXP drives at the moment, and when my new mobo arives (Gigabyte 7DXR) Im thinking about getting another 2 and going the 0+1 way.

There doesnt seem to be too much info about on it. Most people seem to stick to striping, with a few mentions on mirroring (which Ive been using via software for a while now).

From what I can guess, 0+1 should have 2 times the write speed of a single drive, and 4 times the read speed. Do you know if thats right, or is it only 2 time the read speed as well? (not counting overhead of 5-15%).

If Im not going to get a noticable speed increase then Ill just stick to what I have (10,000rpm, cheetah boot disk, 2X medalist pro 7200rpm software mirrored, and 2 times 20Gb IBM 7200rpm 60GXP software mirrored). All SCSI except for the 60GXPs.

I dont need any more space, but you can never have enough speed.

Ali

**Paulr** · 24 July 2001, 01:55

I've been attempting to work out exactly what kind of read & write speeds I can get from the array.
With RAID 0+1 (As apposed to 10) we have a RAID 0 (Striped) set which in turn is Mirrored (Raid 1).
So, in theory you can read data from either of the 2 RAID 0 arrays (which are mirrored) giving you a theoretic read back of 4x.
I haven't really ran any benchmarks on my ststem.
HDTach costs to use under Win2k and I've never been a fan of SANDRA - With SANDRA the scores your system gets are never consistant enough for me to really believe anything it says.

All I can say is that with 4x IBM 60GXP's the array is silent, I don't hear a peep from the drives at all, if I'm wondering if a lot of HD's writes/reads are going on I have to rely on the HD LED on the front of my case.
It's certainly fast, my array is displayed as a single 120GB partition under Win2k and copying whole directories around is done in seconds.
Lastly redundency.
HD's fail, no matter what the make.
OK with some manufacturers you'll get better milleage, but the fact still exists that HD's fail.
With RAID 0+1 I can loose both HD's from any one of the 2 RAID 0 arrays or "Diagonally" accross the RAID 1 and my machine will continue to function without data loss.
Sure the performance isn't as good as either 2 RAID 0 or one large RAID 0 array would give me, but when you're PC is storing a lot of data, even if it is replaceable, redundency to me is really important.

**Sisyfos** · 24 July 2001, 03:02

Beware

As you said; with RAID 0+1 you mirror a stripe.
So if you loose a disk in one mirror your stripe is no longer mirrored and hence you can't afford to loose a disk in the fuctioning stripe.
If you want that kind of redundancy you'd have to go with RAID 10 where you stripe the mirrors and get the "diagonal" redundancy.

Kim

**Paulr** · 24 July 2001, 07:04

Ah but no.
Because the stripe is mirrored you can still loose one disk in each stripe and the mirror image will continue to function.

For example:

RAID channel 1 - HD1 HD2 (These are striped)
RAID channel 2 - HD3 HD4 (These are striped)

RAID channel 1 and RAID channel 2 are mirrored.
Obviously I can loose any one single drive.
Now, 2 drives:

I loose HD1 & HD2 - HD3 & HD4 continue to do the work.
I loose HD3 & HD4 - HD1 & HD2 continue to do the work.
I loose HD1 & HD4 - HD2 & HD3 continue to do the work, although I have lost two halves of a stripe, because they were mirrored they can contine to function.
I loose HD2 & HD3 - HD1 & HD4 continue to do the work, as above.

Only the loss of HD1 & HD3 or HD2 & HD4 or of course the loss of 3/4 HD's will result in my loosing any data.
RAID 10 is only different because it's a striped mirror, so if the above setup was RAID 10:
The loss of HD1 & HD2 or HD3 & HD4 would be the only way data loss would occur.
RAID 0+1 & RAID 10 the only difference is which drives can be lost before data loss occurs.

**Sisyfos** · 24 July 2001, 07:16

RAID 0+1 is NOT to be confused with RAID 10. A single drive failure will cause the whole array to become, in essence, a RAID Level 0 array

Here's the site I quoted

Kim

**Paulr** · 24 July 2001, 07:38

Well, as we are in the mood for quoting, I will do so from the manual for my Promise Fasttrack 100 RAID controller.
As this information appears on Page 73 of a PDF document, I'm going to have to go for cut and paste I'm afraid:

About Dual Data Redundancy
One unique (though rarely occurring) feature of RAID 0+1 is dual fault tolerance. In some cases, two drives can fail simultaneously and still maintain the integrity of data. There are six combinations in which two drives can fail. FastTrak100 protects the data array in four of those cases depending on drive type (some drives do not permit the Slave drive to continue to function if the Master drive fails).
Assume the drives are configured as follows (M = Master, A/B indicates which striped pair the drive belongs to, # indicates which part of stripe data):
IDE 1 IDE2
M Drive A1 Drive B1
S Drive B2 Drive A2

Under RAID 0+1, the array maintains data integrity if any 1, 2 combination survives.
Event Failed Drives Array Status Why?

1 A1/A2 Working B1/B2 retain array integrity
2 B2/B1 Working A1/A2 retain array integrity
3 A1/B2 Working B1/A2 retain array integrity
4 B1/A2 Working A1/B2 retain array integrity
5 A1/B1 Offline B2/A2 contain only half of array data
6 B2/A2 Offline A1/B1 contain only half of array data

As you'll see this is exactly the same way as I described above.

**Sisyfos** · 24 July 2001, 08:56

We agree that RAID 0+1 is a mirror of a stripe, right?
So if A1 fails while you are writing to the disks, then the mirror cannot be maintained.
Even though you hurry out and buy a new disk as fast as you can, if B2 fails while you are away then A2 and B1 can't "cross stripe" because A2 is out of date.

If the Promise controller maintains the A1>B1, A2>B2 mirror on a per disk basis, then it's RAID 10 and not RAID 0+1.

This is a good discussion btw

I just put together a 4+4 RAID 10 at work, so this is *cough* enhancing my understanding of the differences between the two types of RAID.

Kim

**Ali** · 24 July 2001, 18:04

Wow, good info.

I know what you mean about the 60GXPs being quiet. I cant hear the read/writes, but the spin noise is still there.

I think Ill get the other two drives, and put the SCSI HDDs in a different computer. Then I can get rid of the noise of that stuff.

I have all my CD type things in an external SCSI case, and only turn it on when I need to burn a CD, watch or burn a DVD or install something.

It will be good to get rid of the noise from my Alpha P3125 as well!

The mobo should be here tomorrow, so Ill have a busy weekend building my new computer by the looks of things.

Also, do you know if its possible to set up a stipped array, then add the mirror later? I dont know if I can wait for my other 2 drives to turn up (probably 3-4 days if I order them now)

Ali

**Paulr** · 25 July 2001, 02:08

No, it is RAID 0+1.

Taking my last post as example.

You have the primary RAID 0 or simply striped aray:

A1 & B2

This is in turn mirrored to the second array:

B1 & A2

This is the Mirrored Stripe.
Because A1 & B1 are simply mirrors of one half of the striped array, loosing both of those drives will result in the loss of the whole array.
Same goes for B2 & A2, once again they are simply a mirror of one half of the array.

Now, if it was RAID 10 the following would be true:

RAID 10 is a striped mirror.

A1 & B2 are a mirror of one another, as are B1 & A2.
The 2 arrays are in turn striped.
So, if it was a RAID 10 array the following fails would result in the loss of the array:

A1 & B2
B1 & A2

Because, if it was a RAID 10 array the two drives on each of the controllers are mirrors of each other and in turn are both one half of a stripe set.

In fact, RAID 0+1 is a better option than RAID 10, for the following reason:

If one of the IDE controllers on the RAID card was to fail, lets take the example above and we lost IDE1, we would loose A1 & B2.
This would result in the RAID 10 array failing.
However, in a RAID 0+1 array, the array could continue, as A2 & B1 can continue to function.

Promise are being correct, it is a RAID 0+1 array they are using, RAID 10 gives the same amount of protection, but just allows for different combinations of failure.

Do what I did, get some paper & pencil out and draw it all, you'll see the Promise is indeed using RAID 0+1.

**Technoid** · 27 July 2001, 16:05

Bad write speeds

Hi everybody!

I'm back again!

Reinstalled W2k! (It died when it couldent access the raid!)

But now the Write speeds are limited to ~10mb/S!!!

And that is for both the DTLA and the Samsung RAID array!

In W98 I have normal 20~25MB/s write speeds but In W2K its throthled!

I am realy stumped hear!

Can somebody please give me some tips?

If it was only the drive on the stabdard ide port I would blame it on VIA, But it's the same for the raid!!

**Technoid** · 1 August 2001, 14:50

They have arived!

Both HDD's have arived at IBM!

Now, what was that word again.......

AH....

Patience!

**Technoid** · 27 August 2001, 12:57

They'r Back

The HDD's are back!!!

But they are repaired ones

what to do?

Announcement

75GXP does Netherlands!

75GXP does Netherlands!

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment