RAID

[gyrene81]

You are better off installing multiple drives and keeping your volatile data on one drive and the OS on another drive.  I go one more step and add a network fileserver and backup the volatile data once a week to it.  

Ditto...zero lost data using multiple drives and network drive storage.

[Bino]

Bino, and anyone else pointing to documentation of companies who sell RAID solutions, you need to understand they will always be pro-RAID.  They are in the job of selling the hardware.
...

Absolutely, sir.  In the same way that Big Blue data center "surveys" always recommend... Big Blue outsourcing!  Shock!  Amazement! 

I was just pointing to the simplest RAID explanation I knew. 

[katanaso]

For what it's worth, I use RAID0 and RAID1 in my home system.

RAID0 for the OS, apps, games, data, etc. for the speed, and RAID1 for a redundant backup of everything on the RAID0.

I use a piece of software, Acronis True Image, to do weekly and daily scheduled backups of the RAID0 to the RAID1, just in case a drive fails and I need to restore.  So it's basically a redundant backup -- kind of overkill, in a sense.

For piece of mind, I'd go with hardware-based RAID1, just in case you get that one drive that fails.  (I can't recall the last time I had a drive fail at home.  Work is another story.)

Probably the biggest reason I've setup RAID1 on family members' PCs is because they don't backup their important data to DVD or CD.  They keep all their digital photos and documents on the HDD, so RAID1 is a really cheap method for added data protection.


mir

[BoilerDown]

Despite the limitations previously pointed out, I do think Raid 1 is worthwhile if you want some protection against losing your data.  The feature is built in to just about every motherboard you can buy. 

The thing you didn't mention is how much do you want to spend?  Because you can spend practically an infinite amount of money, and someone will still come by and say you could be safer, or you could make it faster, if you only did X instead.  For example, Skuzzy advocates not using raid and doing frequent backups... true, but costly.  The cost isn't just in dollar signs either, but also in time, both to implement and to manage the backups.

So if you want to protect against a single hard drive mechanical failure, I'd say to go ahead and use your motherboard's (whichever one you get, it'll probably have it) Raid 1 feature.  Or raid 5 feature if your motherboard has it and you want to buy at least 3 hard drives.  If you want speed obviously raid 0 has the potential to be better, but its rather risky and probably not worth it except in specific applications.  Which asks the question... what are your intended applications?

You could do a raid 0+1 (or is it 1+0?  I forget the terminology), basically two raid 0s raided into a raid 1, a lot of motherboards support that.  But that's at least 4 hard drives (more money) and you're only getting the space of two of them.  And someone will quickly point out that you should just do a Raid 5 with 4 hard drives instead (maybe slower, get the space of three hard drives though).  And then someone will point out you should just get a dedicated Raid 5 hard drive controller card.  And them someone will say, "no that company sucks, get this controller card instead".  And before you know it you're spending two grand when Skuzzy pipes in with a backup solution that IS cheaper than all that.

So the final answer comes down to... how much do you want to spend, how much and which risks do you want to take or mitigate, and what are you going to use it for?

[Spikes]

But in all seriousness, I think you would be better off (and I do it too) using one or two of your hard drives for system backups over raid. Personal preference here, but I have a 120 gb drive I use for OS backups aside from my 250gb for the rest of my crap.

[Ruler2]

But in all seriousness, I think you would be better off (and I do it too) using one or two of your hard drives for system backups over raid. Personal preference here, but I have a 120 gb drive I use for OS backups aside from my 250gb for the rest of my crap.

Well I bought 6, so I don't think the number of hard drives should be a problem at all, I'm just wondering what the best way to configure them would be for performance.

[MrRiplEy[H]]

Well I bought 6, so I don't think the number of hard drives should be a problem at all, I'm just wondering what the best way to configure them would be for performance.

Make sure you install a fan to blow on the drives directly. One drive may not get too hot but 4-6 on top of eachothers will get toasty.

[Spikes]

Make sure you install a fan to blow on the drives directly. One drive may not get too hot but 4-6 on top of eachothers will get toasty.

+10 couldn't agree more. I make sure that all my cases have a front intake regardless of amt of HDD's. Luckily all the NZXT cases have them, and that's all I will ever buy anymore!

[Skuzzy]

BoilerDown, RAID will not protect your data.  This is the fallacy of RAID and the Windows filesystem.  When a drive starts to go bad, it is highly likely to start corrupting data before it finally dies.  That corrupted data will migrate over the RAID system.  All it takes is one read and one write of corrupted data to completely wreck a RAID system.

This all traces back to the Windows filesystem which will merrily propagate bad data.

In all cases that stack of RAID drives has a higher chance of failure than a single spindle disk.  Statistically, a hard mechanical failure of a spindle without data corruption is rare.  Extremely rare.

I worked on RAID design when I was with Adaptec.  The onboard chip solutions is a non-solution.  They are wrought with pitfalls and bad error handling.  As long as the drives are perfect, you may not have a problem.  Then again, if you are depending on the drives being perfect all the time, why bother with RAID to begin with?

The only filesystems designed to work with RAID are all on UNIX based servers.

[katanaso]

Like Skuzzy says, corrupted data will be corrupted on the entire array.  So bad data is just that -- bad data.  That's a chance you'll take.

The RAID 1 or 5 will protect you against a sudden physical blowup of a drive, but it won't do anything if your data is corrupt.

In our server environment, we rely more on SMART technology than anything to know when the drives are prone to failure so that we can replace them, but that's not a realistic option for home users.

We've had a few servers need to be rebuilt and restored because of a corrupted RAID system, but they were mainly older servers (think 10k 36GB SCSI HDD) that have been running 24x7x365 for 6 or 7 years straight.

On the home side, for me, it's RAID0 for the increase I/O, and just a backup to the other array at night.

mir

 

[BoilerDown]

BoilerDown, RAID will not protect your data.  This is the fallacy of RAID and the Windows filesystem.  When a drive starts to go bad, it is highly likely to start corrupting data before it finally dies.  That corrupted data will migrate over the RAID system.  All it takes is one read and one write of corrupted data to completely wreck a RAID system.

This all traces back to the Windows filesystem which will merrily propagate bad data.

In all cases that stack of RAID drives has a higher chance of failure than a single spindle disk.  Statistically, a hard mechanical failure of a spindle without data corruption is rare.  Extremely rare.

I worked on RAID design when I was with Adaptec.  The onboard chip solutions is a non-solution.  They are wrought with pitfalls and bad error handling.  As long as the drives are perfect, you may not have a problem.  Then again, if you are depending on the drives being perfect all the time, why bother with RAID to begin with?

The only filesystems designed to work with RAID are all on UNIX based servers.

You don't need to be in the business of convincing people to not buy on-board solutions, you're not working at Adaptec any more.   (To paraphrase your own previous statement.)

The truth is, I've worked in IT for a small to mid sized pharmaceutical contracting company for about as long as you've been at HiTech, and I remember a lot of hard drives failing in our servers, but I never remember one corrupting the whole array.  We had three IT people there, and we each got our hands dirty in every aspect of IT you can think of.  The backups in particular was my responsibility.  Our servers were NT or 2003, including our fileservers, and we never lost everything in the way you described.  IMO its just a bunch of FUD.

Now back on topic, we know you've got 6 hard drives.  Its definitely true that if you corrupt your data, either in the way Skuzzy describes, or by a virus attack, or accidental deletion, or whatever, that your data is gone without good backups.  What a raid (other than raid 0) does is protect against an unexpected hard drive failure.  But you asked what's the best way to configure them for performance.  You still haven't said what you're using them for though.  If you really need some performance, I'd probably take 2 of them and make a raid 0, and raid 5 the other 4.  Since you can afford to buy 6 hard drives, I'll just assume you're going to use one of the backup solutions suggested in this thread.  If you believe that no raid setup is safe enough, then get a tape backup solution and do backups to that, don't trust your raided hard drive to hold your backups.  Personally, I think tape backups are overkill for someone who doesn't have a business interest in their computer, such that losing the data could cost you money and customers.  Raid 1 or 5 for hard drive redundancy, and backing up your Raid 0 to that, should be sufficient.

[katanaso]

You want really fast performance?  All 6 in RAID 0.   

But depending on what you want, with the assumption that you want the performance from a RAID 0 array:

4 drives RAID 0, 2 drives RAID 1
3 drives RAID 0, 3 drives RAID 5
2 drives RAID 0, 4 drives RAID 5
2 drives RAID 0, 4 drives RAID 10

This isn't getting into RAID 1/RAID 5, RAID 1/RAID 10, RAID10/RAID1, etc...other options that will increase or decrease the amount of space you have while trading I/O to achieve it.

Honestly, with 6 drives, you have quite a lot of options, so it all depends on what you're doing and what you need.  It is a lot of drives for a gaming PC.  (But I'd do it too.. )

Btw, what type of drives are they?  That can help people help you too.


mir

[Ruler2]

2 are WD1600JSes, the other 4 are Seagate DB35s

[Ghastly]

...I remember a lot of hard drives failing in our servers, but I never remember one corrupting the whole array ..<SNIP>.. What a raid (other than raid 0) does is protect against an unexpected hard drive failure

-Boilerdown

The problem is the difference between what it should do     ^^^^^     and what often actually is the case when the hardware involved is SATA drives and the onboard Intel/Promise/Silicon Image controllers (or the inexpensive fake RAID controller cards - including Adaptec's).   

We're also a small firm with 3 of us in my department, and we utilize the onboard Intel or Nvidia chipset based (depending upon motherboard) fake raid controllers if they are available in the motherboard to mirror the drives in the workstations that we roll out at our company, too. And in my experience, the FUD factor is genuine and based upon cleaning up messes - while they often work exactly as you'd expect in the event of a drive failure, far more often than seems "reasonable" they do not.  Roughly 15% of the time they fail in such a way that the mirror is corrupt in some fashion after the failure of it's paired drive, and about 1 time in 5 they work as expected when one drive fails until you attempt to add a new replacement drive to the array, at which point the entire array becomes unusable, including the previously usable drive that you've only a new mirror to. 

I've even utilized the onboard controllers at times in server configurations, but again, you do so fully cognizant of the fact that it's a cost/risk/reward trade-off (over getting a real RAID solution, I mean).  Hopefully and most likely your servers used something other than the onboard fake raid controllers (and I know they did if they were NT-based!) ... and bear in mind that SCSI RAID solutions intended for use in server solutions are a completely different class of hardware than the onboard/fake RAID SATA hardware under discussion.

Anyway,  I have experienced the loss of an entire array due to a drive failure - once on a server with an Adaptec controller, and many times using the onboard controllers.  Redundancy provided by RAID should be considered a hedge that you hope will buy you some breathing room - time in the event of a failure to deal with it on your own schedule - rather than the "GOOD!  I'm covered so I won't need to backup my stuff" that many folks treat it as.

The onboard chip solutions is a non-solution.  They are wrought with pitfalls and bad error handling.

-Skuzzy

I agree absolutely they are wrought with pitfalls and bad error handling - but I do disagree that they are a non-solution.  For just the cost of a second drive, roughly 80% of the time they save us a significant amount of time when a hard drive goes belly up - and more importantly, most of the time, we get to schedule when we're going to address it for when the user doesn't need the system, rather than the "Good Grief!! Drop everything!!  - go fix Charlie's system!" that used to be the case.

<S>

[katanaso]

2 are WD1600JSes, the other 4 are Seagate DB35s

So two 160GB and four 250GB drives? 

From what I've briefly read, the DB35's would be faster, so I'd probably go with two DB35's in RAID 0, and two DB35's in RAID 1 for my backup drive/partition.  Perhaps use the WD's as another RAID 1 for backup, but not what I constantly access.

Or if you're just gaming, and want speed, and aren't paranoid (probably descibes my liking of RAID 1 at home) -- four DB35's in RAID 0 and the two WD's in a separate RAID 0 (to maximize space) and backup there.