RAID. What is it? Whats the point?

Raid can make backup easier if you set them up for redundancy. Or you can have files stripped on various hard drives so access times are cut down.

http://en.wikipedia.org/wiki/Redundant_array_of_independent_disks

 

Here's a crash course on RAID:

It's a great technology for those that need it, using hardware and/or software to improve reliability and/or performance of hard disks by making multiple hard disks act as one. If you don't know what it is, chances are you don't need it and shouldn't use it. I know of too many performance-seekers who set up a cheap RAID 0 array and lost data, and RAID 0+1 starts to get pretty pricey.

 

For typical desktop use, RAID is not better than individual disks; in terms of speed, reliabiltyh and/or expense.

There are two main types of RAID - striping and mirroring.

Mirroring writes whatever you save immediately to 2 disks at once. This can help you recover if one disk fails catastrophically. It does NOT help you if you accidentally delete a file, or save changes when you didn't want to, or write corrupted data as a result of a crash -- whatever data damage that is written to Disk 1 is also written to disk 2. Also, running a mirror "wastes" at least one drive's capacity, so a 2 drive mirrored RAID 1 set holds only one drive worth of data.

The second type of RAID is Striping -- where data is divided up into pieces, and part written to one drive and part written to another. This can speed up disk reading and writing, because while one disk is busy getting one block of data, the other disk can already be setting up to read/write the next block.

Two problems: It is risky: if you have a failure of either drive, ALL of your data on both drives is gone.

Second problem: Turns out that running a RAID on typical desktop machinery doing typical desktop applications, doesn't gain you much, if any, real world speed. Only intensive disk operations typical of servers, high end video capture, and very high end graphics, stresses the disk systems enouogh for the benefit of striped RAID to outweight the computation loss in calculating where the blocks of data go. Also, if you are running RAID through a single Firewire controller (which you are unless you install a PCI Firewire card), then the botttleneck of the controller nullifies part of the speed improvement.

More complex RAID systems, of 3 to 10 drives, combine the concepts of striping and mirroring in various ways. They are expensive, and again, only pay back on high-end and multi-user scenarios.

The simplest way to improve performance of a Mac';s hard drive system is just to have two drives -- one exclusively for Data, and one for your System and applications. If you have three drives, you can divide them up by System vs Applications vs Data. IF you have 4 drives, you can dedicate the fastest one to the OS scratch files and scratch files for applications like Photoshop. Obviously, there are variations depending on the expandability of the Mac, the speed and size of the drives, and the busses available. Best case on a PowerMac G5 is to install a SATA card, and have 2 SATA internal drives on the motherboard SATA controller, and 2 external SATA drives on the card.

The reason to divide Data and System is so that one set of heads can be reading/writing data, while the other drive's heads are simultaneously accessing the scratch (virtual memory) files for the OS.

 

Did you just cut n' paste all that, or have you had too much coffee this evening?

I'm considering a RAID1, for ensuring I have a backup should a drive fail (I've never had this happen to me before though) - it's an expensive option.

 

I, ummm... typed it....


A second, bootable hard drive is a very good thing to have. You can get most of the benefits of RAID1 (plus a rollback capability) by having a large external drive and sceduling periodic backups wtih SuperDuper, or Sync, or Retrospect. Your loss in case of drive failure would be what you worked on since the last backup, so you scedule the backups often enough to be within the acceptable period of loss, and not too often to interrupt your work unduly.

One benefit of this is that it can be a differerent size drive than your internal, it could be a network vs. a directly conncected drive, it could be a drive in a machine you use as a server, etc. It could be an external drive or a removeable drive in a dock that you could carry away with you and store in a differrent place.

 

Here's a simple explanation of the various types of RAID.

I agree with CanadaRAM. RAID - particularly the higher levels - are not necessary for desktop use unless you are dealing with critical real time data.

 

You would have to ask your network administrator what variety of RAID it was, but large server arrays are commonly RAID5, which combines striping and mirroring.

 

couple of quick questions about RAID and backup, ect...

1) Is a raid scenario the only way to in essence "combine" smaller drives into one large single partition drive? (That is, I have a bunch of old 20GB and 80GB drives I'd like to combine into one large partition instead of having like 4 mounted drives).

2) I also have an external drive. I backup, but before I did any type of "combining" or RAID setup I would want to create a bootable backup of my HD that I could reload onto the new drive. Is there any way to do this with standard OSX disk utility? Or do you have to purchase some sort of third party software??

(Or is it just as easy as copying the entire contents of a HD to a backup drive, drag and drop style??)

--Thanks for all the help

 

Yes, it's the only way. But in a RAID, the volumes have to be all the same size. It is possible to partition a drive and use that partition in a RAID, but the net result is, if you have 4 drives of varying sizes, with the smallest being 20 Gb, and you make a RAID of them, the maximum RAID size is 80 Gb (4 x 20), and less if you want any redundancy. Hardly worth it. (there is a spanning protocol which isn't RAID, really, but allows spanning from one drive to another, however I don't know what, if any, OSX support there is for it)

CarbonCopyCloner or SuperDuper. Check www.versiontracker.com

 

Not the only way. There are plenty of multi-drive firewire enclosures out there (as in 1 enclosure that contains multiple drives) that do what is called "drive spanning". Drive spanning means that the multiple individual drives are viewed by the OS as a single drive.

The drawback of drive spanning (particularly with HFS+ and other inode based filesystems) is that when one drive goes bad the entire volume can be corrupted since the directory entries are scattered across each of the drives in the enclosure. So, assuming that the chances of failure of a single drive is X, then the chances of the spanned volume failing is X/n where n is the number of physical drives in the spanned volume.

Under linux you can use LVM (Logical Volume Management) to handle drive spanning at the OS level rather than in hardware.

 

couple more for ya:

1) So if I just hooked up 3 or 4 drives to a controler card, does anyone have any good tips on how to move files around. That said, doen't certain programs get to acting funny if their libraries are moved (i'm thinking iphoto photos and itunes music in particular) or if you move documents out of the user account folder how will OS X know where to look for it next time.

Pretty much, any good organization / how to move things around ideas. AND

2) say you move IDE drives off the Motherboard and onto a controler card. Will the computer have any hard time at all finding that new location, or will it simply see the same volume mounted, just from a different location? (I ask, because I wouldn't want to go through some massive data shift / reinstall procedure simply to hook a drive up to a faster bus on a PCI card.

--thanks again

 

In the iTunes preferences you can specify the location of the iTunes library. I don't use iPhoto, but I assume it would have a similar preference to set.

Depends on the size of the drives available, the use of the machine and the size of data involved.

Personally, I always move my home folders off the boot drive and move my iTunes library out of my home folder (often onto a different drive if available). I also install all my applications into one of several categorized folders that are not on the boot volume (unless the application insists on being installed on the boot drive). These application folders are named "Applications", "Games", "Languages", "Network", "Utilities" and so on.

A good reason to move your iTunes Library off onto a separate drive is so that you reduce head movement and therefore speed up disk operations when playing music at the same time as do other tasks. When you are playing an mp3 the hard disk must keep reading from the mp3 file. If at the same time you start an application, or surf the web, then the hard disk must jump back and forth between reading the mp3 file and loading the application/saving the web pages to the local cache. By moving the iTunes library onto a separate drive, that drive can keep reading the file and the other drive is busy loading the application/saving the web page.

You could also move your swap space and any scratch files onto a separate drive.

If you are just letting the OS X automounter mount the drives then they will just mount correctly (possibly in a different order).

 

True -- although one enclosure will cost more than a single 300 Gb drive, so it is of doubtful value for a collection of 20 - 80 Gb drives. LaCie and others offer large Firewire drives that come with two (or more) mechanisms and handle RAID in the enclosure hardware. To the Mac it looks like one drive and the Mac doesn't know the difference.

 

RAID or daisy chain?

I use a G5 for recording. I have reached the stage where it is crucial to save all recording files on two hard drives in case one drive crashes and I lose valuble files.
My question is: will I use two hard drives daisy chained using Firewire 800 or use a RAID system. I need to save simultaneously to both drives. On re-opening files I need any changes made/new information to save to both drives. I'm sure you get the drift of my question. As you can gather from my question I am fairly new to Apple computors. I would really appreciate your help on this.

 

mmm... I love the smell of resurrected thread in the morning.

 

There are many others who know so much more, but ...
You want to have two copies that both register any changes made to any file - which means you want one drive which is a mirror of the other - which means RAID 1 see here. The problem with RAID 1 is that if you mess up the file (in oh so many ways), then that corruption / mistake will be mirrored onto the other drive, so it's not a perfect way to protect your data, as you may end up with two copies of a bad file.

What i would probably do is have one drive that holds the current data, and have that copied over to the other drive every couple of hours, depending on how much time you want to loose if things go pear shaped (done with superduper on a schedule). Also have good drives which fail less often, meaning you'll be running less risk of disc failure.

 

RAID /= Backup

Given what you want to do, you'd want:
1. Mirror (RAID 1) two drives. What this actually does for you, is if one of the drives dies, you still have a complete copy that allows you to continue to work. It's called availability, which is how often the system is operational (uptime needs of 24/7). NOT a backup. The primary reason it's not suitable as a backup, is if you make a mistake, it will be duplicated on the second drive in the mirror (though a software glitch could do the same thing without your knowledge).

Please understand, you can use a RAID to backup a RAID, but the method is via scheduling software, not RAID functions. Just as if it were a single drive for backup. This is a bit different, as you're not trying to depend on the RAID to cover itself. They're not capable of this under every circumstance, and it's foolish to do so.

In your case:
2. You also want a 3rd disk as a proper Backup. This isn't part of the RAID set, and the files are sync'ed via software, such as Time Machine or a 3rd party product. This covers your butt if you accidentally erase a file, and later discover you needed it. Or if there's a catastrophic failure that blows the data on both (i.e. software glitch, both drive controller boards burn out,...). Ideally, you'd even want a copy of the data in an online source (totally separte physical location), to cover you from things like flood, fire,... where the entire system is destroyed.

I understand there's budgetary limitations, so the bare minimum is a separate drive (listed above) in the same system or an externally attached unit. Either method will cover you for most things.

You've got it right.

 

My plan is by later this year to have four drives setup

Mac Pro
1. SSD - Operating System, Applications, Documents, etc.
2. SATA HDD (RAID1) - Aperture Libraries
3. SATA HDD (RAID1)
4. SATA HDD - Windows Partition, Video Games

I am also interested in either a Lacie or Drobo solution for backing up my Aperture files.

Question: For my setup (two drives utilizing RAID1) is it preferable to use the basic Disc Utility RAID setup, or should I invest in the Apple RAID Card?

 

I don't know if this will help, but here are some benchmarks for a WD 1TB drive. I assume increases are similar when raiding any particular maker. (About half way down the page)

http://www.pcstats.com/articleview.cfm?articleid=2404&page=2

 

The Apple RAID card is junk, and you don't need it.

Just use Disk Utility.