WD Raptor or Samsung SpinPoint?

[Pauli]

One other argument against a RAID setup is that it adds complexity to your system. If you start having problems that need troubleshooting, there's one more level of complexity to deal with.

[thetoad30]

Yes, RAID0 does statistically double your chance of hard drive failure.

But you have to look at the probablilities for failure and really determine if "doubling your chances" really means anything!!!

Its like saying you have 1 chance in a million to win the lottery.

Then you buy 2 tickets (lets assume that the lottery works this way).

So now you have doubled your chances to win. But isn't 2 chances in a million pretty close to one chance in a million? Sure 2 per million is now 1 in 500000... but when you get down to it, it just doesn't make that much of a difference...

Hell, at one point in time I was actually thinking of RAID0 across ALL 4 of my hard drives. The only reason I didn't do it was because it would be a bitch for me to save all 300GB I have on my computer to DVD's... that's why I decided to go with 2 RAID0 arrays... that way I can use one array to backup, while I reformat and repartition the system array.

[soujir0u]

Do your Samsung drives seek really loudly? Maybe I just got a bad sample.

[El_Duderino]

Just trying to put this horribly derailed thread back on track (while at the same time hijacking it )...

I'm debating whether to go with the WD Raptor (74 gigs) or the 200Gb Samsung 120S for my next upgrade. My rig will not be completely silent either way, I'm looking for a performance/noise sweetspot. My impression from perusing these boards is that the Raptor (rev 2) is completely acceptable for quiet (non-silent) performance systems - especially if you are less worried about seek noise (which I'm OK with) than idle noise (which I hate).

Is this (generally speaking, of course) correct? Thanks in advance!

[barleyguy]

Having two drives does NOT double your chances of failure. Statistics doesn't work that way. It's an average, not an addition. If your chance of failure is 1%, and you have 100 drives, does your chance go up to 100%? Or course not. If you make two bets on red in roulette, is the chance of winning 100%? Of course not. 50% averaged with 50% is 50%.

Also, do two Samsungs have twice the chance of failure as 1 Western Digital? In my personal opinion, two Samsungs have less chance of failure than 1 Western Digital. Just a little personal experience and bias.

One thing that is true with RAID 0: Data Recovery is much more difficult, bordering on impossible. In the unfortunate event that you lose a drive, there's a good chance of getting the data back in a non-RAID configuration. With RAID-0, the recovered data has to be matched up byte for byte with the other drive, which is darn close to impossible.

Another point about partitioning: NTFS is much less stable with partitions greater than 128 Gigabytes. With partitions over 128 Gig, NTFS relies on an OS patch and a registry key. Therefore, a corrupted registry or an overwrite of system files can cause data loss. I've actually lost all the data on a drive because my partition was over 128 gig and I didn't have the patch installed.

My 02
Harley.

[nick705]

My impression from perusing these boards is that the Raptor (rev 2) is completely acceptable for quiet (non-silent) performance systems - especially if you are less worried about seek noise (which I'm OK with) than idle noise (which I hate).

Is this (generally speaking, of course) correct? Thanks in advance!

I'd say you are correct... I have a 74GB Raptor (mounted in a standard SLK3700AMB drive bay) and an older Nidec-motored Spinpoint (SP1203N), and there's nothing to choose between them for idle noise (they're both inaudible for practical purposes). The Raptor's seeks are definitely noticeable but not intrusive or unpleasant unless you want absolute quietness, whereas the Spinpoint with AAM enabled is very quiet indeed in all respects...you can just hear the seeks if you listen hard, but you wouldn't really notice unless you were deliberately trying. And that's with the Spinpoint hard-mounted in an old Inwin case...with proper suspension I'm sure it would be effectively inaudible.

[dynamiks]

Well thanks everyone for your comments and input on the whole situation here. I guess for now I will stick with the Raptor, but I really wish I could test out both to see if the 2 Samsungs would really be that much slower than the Raptor. But hey I will probably never know. Only way I'll ever know for myself is to actually try it, everyone's opinion is really not going to make much of a difference till I try it for myself. And since I don't have money to be sampling things, and taking risks like that. I guess I'm going to be sitting here with this Raptor for main drive, and Samsung for a backup.

But thanks for all the help anyway guys. I wish there was a way to make the seeks of a raptor a little bit more inaudible, but I'll learn to live with them I guess. At least the drive is silent at idle, so I cannot complain there. Because my old WD 1200JB was loud at idle, sooooo very annoying. I'm glad I got rid of it.

[dynamiks]

Oh by the way, how do I know if my Rapor is FDM or Ball Bearing? Is there something indicated on the drive itself? I'd like to check it out, and see which version I have. I purchased mine around November/December in 2004.

[Ozy666]

Having two drives does NOT double your chances of failure. Statistics doesn't work that way. It's an average, not an addition. If your chance of failure is 1%, and you have 100 drives, does your chance go up to 100%? Or course not. If you make two bets on red in roulette, is the chance of winning 100%? Of course not. 50% averaged with 50% is 50%.

My 02
Harley.

Not quite.

P(fail_mult) = 1 - (1-P(fail_individ))^N where N is the number of drives.

So, if a single drive has a 50% failure rate, then two drives would have a failure rate of 1 - (1 - .5)^2 = 1 - .25 = .75 = 75% failure chance.

So, 100 drives with 1% failure rate would be:

1 - .99^100 = 1 - .366 = .634 = 63.4% chance for at least 1 failure.

[dontiego]

hello!

I get in this discussion, I have already 2 striped Raptors.
But I am NOT happy!
While I don't really care hearing them seeking when I open a file, I do care hearing them clicking every ten second *at idle*.

I'm a bit surprised not to see any comment about that.
Don't you guys hear that? It's like they are seeking for a tenth of a second every 10 second.

If this rings a bell, please tell me there is a solution!

I hate that noise...

[dynamiks]

hello!

I get in this discussion, I have already 2 striped Raptors.
But I am NOT happy!
While I don't really care hearing them seeking when I open a file, I do care hearing them clicking every ten second *at idle*.

I'm a bit surprised not to see any comment about that.
Don't you guys hear that? It's like they are seeking for a tenth of a second every 10 second.

If this rings a bell, please tell me there is a solution!

I hate that noise...

Mine does this also... and its more like every 5 seconds.

[ronrem]

A Spinpoint will be quieter than a Raptor BUT the idea on the table is to replace the raptor with 2 spinpoints-which then are both running full time.

I Partition because I have had OS failures,have had "Bad Sector" problems,and I find I can live with reloading the software when my data and files are seperate and thus not affected by the "C" partition's troubles.

Let us say we have 2 HDs. Each,in 5 years use,has a 15% chance of failure-possibly with no warning. There is then a 30% chance that one of the two will fail. In Raid 0, EVERYTHING is spread across the two drives.

Raid 1 is redundancy,mirroring. A full 4 drive raid compensates for the added insecurity by constantly writing a clone pair. However you then have 4 drives chirping.

I do a lot of Audio stuff. Putting music files on a Raid 1-short term,can really speed the processing of files that can be 20 mb,100 mb or more....The trade off is I then would have to silence 3 drives,the OS/software drive and the raid set-and I'd want another drive for long term storage and backups. More likely I do a drive for the OS/software-and a big quiet spinpoint SATA 2 for storage and skip raid.

[ronrem]

There are some math formulas etc,above-and they do not apply. EITHER DRIVE CROAKS and ALL the data is toast so it is not the AVERAGE failure,it IS the Failure rate X 2.

Russian roulette with 2 bullets instead of one-that's yer math right there.

I'd put the Raptor in a real good enclosure and suspend it.

[dontiego]

So if you have 4 striped drives with each probability failure of 40%, then you are sure that at least one of them will crash with a probability of:
4 * 40 = 160%

Then you're 60% more sure that they will crash than if you had 100%.
60% more sure that some of them will crash than if you were sure that one of them will crash... o_O
Doesn't make sense, you cannot exceed 100%.
Here we're speaking of playing the russian roulette with two guns, rather than one gun with two bullets.

Let's do the math again

You have a system with 2 striped drives A and B, with each the same failure probabilty F at a given time.

The probability that drive A crashes is F, *and this is independant of B crashing or not*.
This means that in your probability F, you have both the cases where A crashes alone and where A and B both crash.

Now in the probability F that drive B crashes, you ALSO have both cases where B crashes, and where A and B both crash.

So when you sum up F + F, you get the probability that A crashes alone or A and B together, plus the probability that B crashes alone or A and B together.
Twice A and B crashing together!

The solution? Sum up the probability that A crashes *alone*, the prob that B crashes *alone*, and the one that A and B *both* crash.

A crashes alone -> A crashes AND B doesn't crash = F * (1-F)
B crashes alone -> A doesn't crash AND B crashes = (1-F) * F
A and B both crashes -> F * F

Sum up all that and you get the probability of your system crashing:
-F^2 + 2*F

That matches the point that you cannot exceed a final prob of 1 (for F in [0, 1]).

Another approach: you go from your wrong result of F + F, and remove the probability that both drives crash (which you counted twice). You get:
F + F - (F*F),
which is the same result.

And it matches also the (right!) math discussed above, at F = 0.5 (50%), the final prob is 0.75 (75%).

Now if you have F very low (WD raptor have 1,2 million hours of mean time between failures!!), F*F would be even lower, almost to neglect.

It took a while for me to understand that at school, I must admit

Next, the probability that at least one of your Raptors will clicks at time t

[Shadowknight]

Too lazy to read the post above me, but the claims of increased drive failure in RAID don't make sense.

Example: You buy two hard drives, A and B.

Scenario 1: You RAID them, drive A fails, your data is lost

Scenario 2: You decide to install only one hard drive, A, in your new system build. Drive A fails, your data is lost.

Unless RAID magically causes more stress on both hard drives, I really don't see wha it's riskier using it vs. a single drive as in both scenarios one hard drive failure loses all of your data.

[dontiego]

Too lazy to read the post above me, but the claims of increased drive failure in RAID don't make sense.

Example: You buy two hard drives, A and B.

Scenario 1: You RAID them, drive A fails, your data is lost

Scenario 2: You decide to install only one hard drive, A, in your new system build. Drive A fails, your data is lost.

Unless RAID magically causes more stress on both hard drives, I really don't see wha it's riskier using it vs. a single drive as in both scenarios one hard drive failure loses all of your data.

Scenario 1: data lost if A crashes OR if B crashes.

Scenario 2: data lost only if A crashes (less probability but not exactly half)

[xilencer]

If I understand RAID 0 correctly it's pretty simple. The problem in this thread seems to be that drives are not separated from data.

* RAID 0 does not increase or decrease the chance of hard drive failure.

* RAID 0 multiplies the chance of data loss with the number of drives used.

In the case of RAID 0 with two drives the chance of data loss is doubled compared to a single drive. No math needed for drive failure as it is not affected.

The drives don't care if they are in RAID 0 or not, however as drives in RAID 0 share all data, it only takes one drive failure to destroy all data on all drives. After drive failure the rest of the drives in the array can still be re-partitioned/re-formatted and used, in another RAID 0 or whatever.

It's called RAID 0 as it's not a proper RAID level, because in respect of preventing data loss it goes the opposite way. The chance of data loss is easy to calculate, 2 drives means 2 times the chance compared to 1 drive, 3 drives means 3 times the chance compared to 1 drive, 100 drives means 100 times the chance compared to 1 drive, etc.

If 1 drive out of 100 in RAID 0 fails, that's 1 drive failure.
However, all data has been lost across all 100 drives.

Or am I the one who doesn't understand?

[dontiego]

Or am I the one who doesn't understand?

Yes, you are. Or at least you don't have everything right.

* RAID 0 does not increase or decrease the chance of hard drive failure.

-> Yes and no. If one disk in your RAID0 fails and is destroyed, the other ones are safe, but the data on them is lost.

* In the case of RAID 0 with two drives the chance of data loss is doubled compared to a single drive. No math needed for drive failure as it is not affected.

-> False. If you read my earlier explanation, you'll understand that the probability for data loss is then greater, but not quite doubled. However it is quite close to it since today's hard drives have low failure rate.

* The drives don't care if they are in RAID 0 or not, however as drives in RAID 0 share all data, it only takes one drive failure to destroy all data on all drives. After drive failure the rest of the drives in the array can still be re-partitioned/re-formatted and used, in another RAID 0 or whatever.

-> Exact!

* It's called RAID 0 as it's not a proper RAID level, because in respect of preventing data loss it goes the opposite way.

-> I don't understand. RAID 0 is striping your data to two disks (or more), RAID 1 is mirroring the data of one disk to another. The RAID word without any number groups these 2 concepts.

* The chance of data loss is easy to calculate, 2 drives means 2 times the chance compared to 1 drive, 3 drives means 3 times the chance compared to 1 drive, 100 drives means 100 times the chance compared to 1 drive, etc.

-> False. As explained earlier, having 100 drives in RAID0 with a failure probability of 2% will never give you a system with failure probability of 200%. A probability higher than 100% does not exist. As Ozy666 stated (and I later), the formula is the following:
with probability p of one disk failing, and N the number of disks in a RAID0 array, the probability of your system failing is:
1 - (1-p)^N.

* If 1 drive out of 100 in RAID 0 fails, that's 1 drive failure.
However, all data has been lost across all 100 drives.

-> yes. If one doesn't see a data loss as a disk failure

[Shadowknight]

Suppose you buy drive A.

It fails, all data is lost permanently.

Suppose you buy A AND B.

A crash, all data is lost permanently.

Whenever you buy a hard drive there is a chance it will fail and lose all your data.

There is no way you can prove RAID is necessarily worse than non-RAID. Any hard drive you buy, ANY, can go bad at any time.

[nick705]

There are some math formulas etc,above-and they do not apply. EITHER DRIVE CROAKS and ALL the data is toast so it is not the AVERAGE failure,it IS the Failure rate X 2.

Russian roulette with 2 bullets instead of one-that's yer math right there.

I'd put the Raptor in a real good enclosure and suspend it.

That's not quite the same thing, as if you played Russian roulette with all six chambers containing a bullet, you'd be assured of blowing your brains out (ie a probability of 1). If however you had 100 drives, each with a 1% chance of failing withing a given time period, the probability of one of them failing during that period is not 1...there's still a chance they might all end up working OK. A better analogy is playing Russian roulette twice with two different guns, each gun having a single bullet in one of its six chambers... you're not actually doubling your chance of dying, just as if you did it six times you might still come through it alive.

Statistics makes my head hurt though, so I'm going to leave it there and download some porn.

edit: in a minute, anyway...

There is no way you can prove RAID is necessarily worse than non-RAID. Any hard drive you buy, ANY, can go bad at any time.

I think you're misunderstanding statistics...it's the likelihood of something happening we're talking about. Every single lottery ticket has an equal likelihood of winning, but someone who buys 1000 tickets is more likely to win than someone who buys just one. That's unrelated to the fact that there are just two possible outcomes (you win or you don't win).