hard-drive raid question

[tospace]

Alright.  so I'm looking at harddrives for expanding my storage probably for the next couples years until ssd's are cheap and I only have enough money for a cheap sata ii hd.  This one thats in my price range happens to be the same model as the one that I've had for a couple years and its still working perfectly so I was thinking why not do raid 0?  Only reason I ask is because Iv'e heard its better to get all drives from the same batch for raid but would that matter too much?  hares the link http://www.newegg.com/Product/Product.aspx?Item=N82E16822136696

[Meni Rosenfeld]

For RAID 1 you certainly DO NOT want the drives to be from the same batch as they are more likely to all fail together.

But for RAID 0 I really don't think it matters. Having the same model will probably be better for performance but no need for the same batch.

[Bitsky]

expanding my storage

why not do raid 0?

[nancyfromafrica]

what do you need raid for are you running a server

[BadBear]

If you want storage for typical home use, use something like Unraid or Flexraid. I use Unraid because you don't have to use the same types of drives to expand your storage later on, and it only uses one drive for parity protection, as opposed to half of them.

You really don't want to use a standard raid array for storage. People typically use raid for access speed and uptime. Raid is a terrible choice for storage. You are greatly increasing your chances of losing everything by using it.

[Aahzman]

For RAID 1 you certainly DO NOT want the drives to be from the same batch as they are more likely to all fail together.

But for RAID 0 I really don't think it matters. Having the same model will probably be better for performance but no need for the same batch.

True dat. As long as they are the same capacity and the same RPM, it'll be fine for RAID1. amount of cache on each drive doesn't need to be identical, nor do they even have to be running the same firmware code, etc.

If you mirror a 7200RPM drive and a 10K RPM drive, the 10K will (assuming it's a relatively recent product) be throttled back to 7200RPM. if you mirror a smaller drive to a larger capacity drive, you will only see/be able to use an amount of capacity on the larger drive equal to the size of the smaller drive.

[DeathAndTaxes]

Friends don't let friends risk their data with RAID0.

For most single user applications there is insufficient queue depth (# of pending IOs) for RAID 0 vs single disk to make any meaningful difference outside of synthetic applications.  However RAID0 exponentially increases the chance of losing data because a loss of EITHER disk means a complete loss of all data.

To use an analogy it is like rolling a die and if a 6 comes up you lose data.  With RAID0 you are rolling multiple dice and if any come up six you lose.

Using RAID0 increases the likelihood of a primary data loss and thus needing to relying on backups (which could be incomplete/corrupt/outdated) without any real benefit. If you absolutely need to use RAID (and based on your OP you likely don't) go with RAID5.

[casascius]

You really don't want to use a standard raid array for storage. People typically use raid for access speed and uptime. Raid is a terrible choice for storage. You are greatly increasing your chances of losing everything by using it.

True of raid 0, definitely not true of any other RAID level since redundancy and the ability to tolerate one or more totally failed drives is a feature of the others.

[DeathAndTaxes]

You really don't want to use a standard raid array for storage. People typically use raid for access speed and uptime. Raid is a terrible choice for storage. You are greatly increasing your chances of losing everything by using it.

True of raid 0, definitely not true of any other RAID level since redundancy and the ability to tolerate one or more totally failed drives is a feature of the others.

<D&T soapbox mode on>
D&T (as a 10 year hardware manager before getting into software development) climbs on to his soapbox.  Well he used the word "everything" so it is true even in other forms of RAID.  

RAID5 vs JBOD for example trades the risk of losing something for the risk of losing everything if more than one disk dies (and yes it can happen if you get a bad batch of disks).  RAID should NEVER be used to "prevent" dataloss.  The method of preventing dataloss is backups.  

However restoring from backups is time consuming and can't be scheduled if it is a mission critical server.  RAID when used properly buys the IT department uptime and availability. When a disk fails the server doesn't go down.  Downtime is expensive.  Data PROTECTION should always be handled by backups because eventually a RAID array is going to break.  

</end D&T soapbox mode>

RAID has to be about the most misunderstood computer technology invented.  It likely has resulted in more dataloss when used incorrectly by hobbyist then just about anything other than maybe viruses/malware.

[runlinux]

You really don't want to use a standard raid array for storage. People typically use raid for access speed and uptime. Raid is a terrible choice for storage. You are greatly increasing your chances of losing everything by using it.

True of raid 0, definitely not true of any other RAID level since redundancy and the ability to tolerate one or more totally failed drives is a feature of the others.

<D&T soapbox mode on>
D&T (as a 10 year hardware manager before getting into software development) climbs on to his soapbox.  Well he used the word "everything" so it is true even in other forms of RAID.  

RAID5 vs JBOD for example trades the risk of losing something for the risk of losing everything if more than one disk dies (and yes it can happen if you get a bad batch of disks).  RAID should NEVER be used to "prevent" dataloss.  The method of preventing dataloss is backups.  

However restoring from backups is time consuming and can't be scheduled if it is a mission critical server.  RAID when used properly buys the IT department uptime and availability. When a disk fails the server doesn't go down.  Downtime is expensive.  Data PROTECTION should always be handled by backups because eventually a RAID array is going to break.  

</end D&T soapbox mode>

RAID has to be about the most misunderstood computer technology invented.  It likely has resulted in more dataloss when used incorrectly by hobbyist then just about anything other than maybe viruses/malware.

+1

I have a raid 5 setup on my server for the off chance a drive dies. I backup my files weekly to a separate location. I can't afford to loose that stuff. I don't make that many changes to the data, so weekly is fine.

[casascius]

RAID has to be about the most misunderstood computer technology invented.  It likely has resulted in more dataloss when used incorrectly by hobbyist then just about anything other than maybe viruses/malware.

I don't think I've recommended RAID as an alternative to backups, just pointed out that other than RAID0, the array isn't subject to the loss of any single drive, which is still far more likely than a multiple drive failure.  Agreed, if you run RAID with a bad batch of drives and the whole batch goes bad, you might lose everything, but the same is true if you don't use RAID and you store your data on a single drive from a bad batch. 

The advice shouldn't so much be "don't use RAID", but rather, "don't assume RAID is a substitute for a backup", "use a RAID configuration that allows failure of two drives whenever possible", and "don't forget you need to monitor your array, and don't let a failed drive sit unreplaced for any longer than absolutely necessary".  Otherwise it's sort of like saying "don't wear your seatbelt, the extra confidence that you're safe might cause you to get in an accident".

[DeathAndTaxes]

RAID has to be about the most misunderstood computer technology invented.  It likely has resulted in more dataloss when used incorrectly by hobbyist then just about anything other than maybe viruses/malware.

I don't think I've recommended RAID as an alternative to backups, just pointed out that other than RAID0, the array isn't subject to the loss of any single drive, which is still far more likely than a multiple drive failure.  Agreed, if you run RAID with a bad batch of drives and the whole batch goes bad, you might lose everything, but the same is true if you don't use RAID and you store your data on a single drive from a bad batch.  

The advice shouldn't so much be "don't use RAID", but rather, "don't assume RAID is a substitute for a backup", "use a RAID configuration that allows failure of two drives whenever possible", and "don't forget you need to monitor your array, and don't let a failed drive sit unreplaced for any longer than absolutely necessary".  

I think you misunderstood.  He said with raid you increase the chance of losing "everything".  That is true of all forms of RAID even RAID5.  The risk of losing something decreases but the risk of losing everything (broken array) increases.

If you have files on 5 drives it would take 5 failures to lose everything (although only 1 failure to lose something).
If you have files on RAID 5 (using 5 drives) it would take 2 failures to lose everything.  The risk of losing everything has increased.   One is trading availability for data security.  In the Enterprise world given the cost of downtime for mission critical applications that is a beneficial trade but it is still a tradeoff.

I didn't mean to imply you were advocating not backing up or anything else.  Just correcting your correction.

 

[tospace]

Alright, thanks for the info guys.  I think I will get the drive and just start using it and if i get to a point where I have everything backed up so that I can reformat, maybe cloud storage? I'll try out raid 0.  What would be nice in the future is 0+1, or just 1 for important stuff and a couple ssd's in 0 would be perfect for programs and os.  It kinda sucks right now tho, hd's don't seem to be getting any cheaper, just ssd's.  Other wise i'd say short stroke 4 3tb drives and be future proof  So I will go by the rule, you mine as-well use raid 0 if its fine to loose everything

Also Iv'e heard raid 5 is terrible for performance using cpu and being slow or something if you don't have an external raid controller?  that would require 3 hard drives anyways tho?

[zenbarter]

When I first read the subject line I thought it would be about the best way to protect data from a government raid or something like that, LOL.

[SgtSpike]

RAID has to be about the most misunderstood computer technology invented.  It likely has resulted in more dataloss when used incorrectly by hobbyist then just about anything other than maybe viruses/malware.

I don't think I've recommended RAID as an alternative to backups, just pointed out that other than RAID0, the array isn't subject to the loss of any single drive, which is still far more likely than a multiple drive failure.  Agreed, if you run RAID with a bad batch of drives and the whole batch goes bad, you might lose everything, but the same is true if you don't use RAID and you store your data on a single drive from a bad batch.  

The advice shouldn't so much be "don't use RAID", but rather, "don't assume RAID is a substitute for a backup", "use a RAID configuration that allows failure of two drives whenever possible", and "don't forget you need to monitor your array, and don't let a failed drive sit unreplaced for any longer than absolutely necessary".  

I think you misunderstood.  He said with raid you increase the chance of losing "everything".  That is true of all forms of RAID even RAID5.  The risk of losing something decreases but the risk of losing everything (broken array) increases.

If you have files on 5 drives it would take 5 failures to lose everything (although only 1 failure to lose something).
If you have files on RAID 5 (using 5 drives) it would take 2 failures to lose everything.  The risk of losing everything has increased.   One is trading availability for data security.  In the Enterprise world given the cost of downtime for mission critical applications that is a beneficial trade but it is still a tradeoff.

I didn't mean to imply you were advocating not backing up or anything else.  Just correcting your correction.

Well, since we're arguing semantics, it is NOT true of all forms of RAID.

In RAID1, your risk of losing everything does not increase, and your risk of losing some things decreases (vs storing different files on each disk individually).