Understanding RAID: Internal and External RAID-ing

[Earendil]

So I have a Powerbook with an 80gb 4200RPM drive. I recently bought an external FW 800, 7200RPM drive weighing in at 160gb.
Right now I have the external partitioned in half, using one of the 80gb portions as a backup for my internal drive.

Perhaps my question is ridiculous, because I don't understand the intricacies of RAIDs. But I'm going to ask anyway
Would it be possible to use one of the 80gb partitions with my internal 80gb in RAID form?
If it is possible, would it be any faster for me?
Would it provide the same backup functionality if my internal failed?
How well would it handle being *occasionally* disconnected?

Thanks guys!

~Earendil

 

[mellow2bits]

So I have a Powerbook with an 80gb 4200RPM drive. I recently bought an external FW 800, 7200RPM drive weighing in at 160gb.
Right now I have the external partitioned in half, using one of the 80gb portions as a backup for my internal drive.

Perhaps my question is ridiculous, because I don't understand the intricacies of RAIDs. But I'm going to ask anyway
Would it be possible to use one of the 80gb partitions with my internal 80gb in RAID form?
If it is possible, would it be any faster for me?
Would it provide the same backup functionality if my internal failed?
How well would it handle being *occasionally* disconnected?

Thanks guys!

~Earendil

Hi Earendil,

to set up RAID you use Apple's Dsik Utility, it's got fairly good help files explaining RAID.

My experience with RAID is with a 1.44 mhz Mini with an internal 80gb hard disk and an external 160 Gb firewire disk in two partitions of 80 Gb. I use one of these two partitions for a RAID set together with the internal 80Gb disk and the other 80Gb as extra space for less critical files.

To answer your questions:
yes it is possible.
no, it would probably be slower (there are three options to use raid: mirroring (creates two exact copies of all your data); concentenated (spreads out large files over two ore more disks in your RAID set - improving speed when using large files); striped (combining two or more disks into one larger (virtual) disk).
If you want a back-up functionality, choose mirrored RAID. the RAID disk you create will most likely be slower than your single 80Gb internal disk (4200rpm).
I think that only mirrored RAID will allow you to disconnect one disk. When you reconnect, the RAIDset will have to be rebuild (checking wich files have changed in the mean time and synchronising both disk). This takes at least an hour and a half on my Mini, not something you want to do very often. Of course, you can always disconnect and reconnect when your Mac is shut down (and not turned on inbetween).

to summarize: no speed gain, yes back-up, and rather do not disconnect.

Good luck and let us know what you decide!

-m2b

 

[Earendil]

Thanks for the run through.
What I don't understand is how you can say that it would provide NO improvement. Test at barefeats have shown that a FW800 External drive is damn near as fast as an internal IDE drive. So how can adding a 7200RPM drive in RAID form to my slow clunky 4200RPM not make it any faster, or even slower?! surely not because of the FW800 connection...

Is there some principle where the RAID will still use the internal for a large majority of the tasks instead of relying on the external to pick up information faster?

~Earendil

 

[strydr]

RAID ?'s

Thanks for the run through.
What I don't understand is how you can say that it would provide NO improvement. Test at barefeats have shown that a FW800 External drive is damn near as fast as an internal IDE drive. So how can adding a 7200RPM drive in RAID form to my slow clunky 4200RPM not make it any faster, or even slower?! surely not because of the FW800 connection...

Is there some principle where the RAID will still use the internal for a large majority of the tasks instead of relying on the external to pick up information faster?

~Earendil

-The reason for NO improvement, is simple. The system now has to write each bit to two different drives, as opposed to one. This operation will require more CPU, and therefor, will slow down a little. When the RAID is done through software, it takes even more CPU- most hardware RAID solutions offload the task to the RAID card. The improved speed of the FW drive will be degraded because it is only writing the same data as the internal drive (4200rpm). Now, when you access data on the other partition (non-RAID), you should have good speed, but on write to the mirror set, you will have a slight slowdown.

-The idea behind a Mirror set, is it reads the data off both drives (the OS only sees one drive, the RAID handles all the interchange of data), so when you access data, both drives are active.

-If your goal is to have a regular backup, you should try using a backup solution like Dantz, or just figure out how to use Rsync- just create a CRON job to run your backup at a specified time (or create a script to run it whenever you want)

 

[Earendil]

-The reason for NO improvement, is simple. The system now has to write each bit to two different drives, as opposed to one. This operation will require more CPU, and therefor, will slow down a little. When the RAID is done through software, it takes even more CPU- most hardware RAID solutions offload the task to the RAID card. The improved speed of the FW drive will be degraded because it is only writing the same data as the internal drive (4200rpm). Now, when you access data on the other partition (non-RAID), you should have good speed, but on write to the mirror set, you will have a slight slowdown.

-The idea behind a Mirror set, is it reads the data off both drives (the OS only sees one drive, the RAID handles all the interchange of data), so when you access data, both drives are active.

-If your goal is to have a regular backup, you should try using a backup solution like Dantz, or just figure out how to use Rsync- just create a CRON job to run your backup at a specified time (or create a script to run it whenever you want)

That makes some sense I guess. Thanks for the info!
And Strider, err Strydr, nice computer names
My Powerbook is named Elrond, my iPod Elros, and of course my user name is Earendil

 

[CanadaRAM]

RAID Basics

RAID - Redundant Array of Inexpensive Disks

Redundancy (or mirroring) means that the data you write to the hard drives is stored with a backup so that if you suffer the loss of one drive in the array, you can reconstruct all of your data from the remaining drives.

Array means more than one drive working together and appearing to your machine as one storage device

Inexpensive related originally the difference in price between a group of 3 to 10 small hard drives and the expense of a single gigantic drive used in servers. This distinction is largely academic now.

The RAID function is controlled either by software or by a dedicated RAID controller card. There are two components of a RAID array; striping and redundancy (or mirroring).

Striping takes your data and divides the bytes across two or more hard drives. The theory is that the writing and reading speed will be faster because as drive A is dealing with one byte, drive B is already getting a head start finding the next byte to read or write. This works because transferring the data takes only 20% of the time needed, the other 80% of the time the drive needs to move the heads and wait for the data on the disk to rotate around to the head.

The simplest RAIDs are called RAID 0 (striping) and RAID 1 (mirroring)

RAID 0 simply uses two identical hard drives, and stripes the data between them to create a single volume. There is no redundancy, therefore if one drive has a failure, ALL the data is lost on both drives. Cheapest but risky. 2 x 100 Gb drives in RAID 0 = 200 Gb space

RAID 1 (mirror) just means two drives, where data is written simultaneously to both. Drive B is a mirror of drive A. This provides good redunancy, at the expense of half or the drive space. 2 x 100 Gb in RAID 1 = 100 Mb space and at the cost of some extra writing time as the data must be written twice.

RAID 2 through 5 are various arrangements of 3 to 10 drives, where data is striped across the drives, but "parity" data is also stored, to be able to reconstruct any single drive. Typically one drive is consumed by the parity data, so
3 x 100 Gb drives in RAID = 200 Gb space
5 x 100 Gb drives in RAID = 400 Gb space

You can also combine RAIDs, such as a striped pair (RAID 0) mirrored by another identical striped pair (RAID 1), which is sometimes called RAID10 or RAID 1+0 or RAID 0+1
4 x 100 Gb drives in RAID 0+1 = 200 Gb space.

Bottom line:
For desktop computer use, RAIDs are almost never worth it for speed improvement, unless you are doing something that needs massive data streaming, like video production.

There is an overhead to RAID in calculating where each byte of data is going to go. This means that RAIDs are slower than a single drive for everyday use.

They only really shine when put under the loads of a server, or if you need a single volume that exceeds the size of available single drives.

RAID 0 is terribly risky unless you maintain a fulltime backup onto another drive.

RAID 1 - 5 provide additional security but have significant costs for drives, controller cards, and drive enclosures. But remember that RAID 1 - 5 only protect against drive failure. If you delete a file, have a crash or get a virus, the damage is done instantaneously to the mirror as well as the primary data, so RAID gives you no protection against that.

 

[neocell]

Thanks for the informative posts mellow2bits, strydr and CanadaRAM. I feel a little bit less ignorant now

 

[strydr]

That makes some sense I guess. Thanks for the info!
And Strider, err Strydr, nice computer names
My Powerbook is named Elrond, my iPod Elros, and of course my user name is Earendil

Thanks, my Domain (internal) is Fanghorn.. I need more computers so I can have the entire middle earth represented.. Never thought of naming my iPod.. good call, I think it should be Samwise..

 

[belvdr]

RAID Basics

RAID - Redundant Array of Inexpensive Disks

Redundancy (or mirroring) means that the data you write to the hard drives is stored with a backup so that if you suffer the loss of one drive in the array, you can reconstruct all of your data from the remaining drives.

Array means more than one drive working together and appearing to your machine as one storage device

Inexpensive related originally the difference in price between a group of 3 to 10 small hard drives and the expense of a single gigantic drive used in servers. This distinction is largely academic now.

The RAID function is controlled either by software or by a dedicated RAID controller card. There are two components of a RAID array; striping and redundancy (or mirroring).

Striping takes your data and divides the bytes across two or more hard drives. The theory is that the writing and reading speed will be faster because as drive A is dealing with one byte, drive B is already getting a head start finding the next byte to read or write. This works because transferring the data takes only 20% of the time needed, the other 80% of the time the drive needs to move the heads and wait for the data on the disk to rotate around to the head.

The simplest RAIDs are called RAID 0 (striping) and RAID 1 (mirroring)

RAID 0 simply uses two identical hard drives, and stripes the data between them to create a single volume. There is no redundancy, therefore if one drive has a failure, ALL the data is lost on both drives. Cheapest but risky. 2 x 100 Gb drives in RAID 0 = 200 Gb space

Don't forget you can use more than 2 drives in a RAID 0 as well.

RAID 1 (mirror) just means two drives, where data is written simultaneously to both. Drive B is a mirror of drive A. This provides good redunancy, at the expense of half or the drive space. 2 x 100 Gb in RAID 1 = 100 Mb space and at the cost of some extra writing time as the data must be written twice.

RAID 2 through 5 are various arrangements of 3 to 10 drives, where data is striped across the drives, but "parity" data is also stored, to be able to reconstruct any single drive. Typically one drive is consumed by the parity data, so
3 x 100 Gb drives in RAID = 200 Gb space
5 x 100 Gb drives in RAID = 400 Gb space

You can also combine RAIDs, such as a striped pair (RAID 0) mirrored by another identical striped pair (RAID 1), which is sometimes called RAID10 or RAID 1+0 or RAID 0+1
4 x 100 Gb drives in RAID 0+1 = 200 Gb space.

In a combined RAID scenario, you use the first RAID number to create volumes, then use the second RAID number across the volumes. So, RAID 1+0 (aka RAID 10) is not the equivalent of RAID 0+1.

In RAID 1+0, you have at least two mirror sets that are striped. So you can have 2 x 72GB in a mirror and 2 x 72GB in another mirror, then you have both 72GB volumes in RAID 0 for 144GB of space.

RAID 0+1 is the opposite. You have 2 x 72GB drives in a RAID 0, and 2 x 72GB in another RAID 0 volume. The two striped 144GB volumes are then mirrored.

Bottom line:
For desktop computer use, RAIDs are almost never worth it for speed improvement, unless you are doing something that needs massive data streaming, like video production.

There is an overhead to RAID in calculating where each byte of data is going to go. This means that RAIDs are slower than a single drive for everyday use.

They only really shine when put under the loads of a server, or if you need a single volume that exceeds the size of available single drives.

RAID 0 is terribly risky unless you maintain a fulltime backup onto another drive.

RAID 1 - 5 provide additional security but have significant costs for drives, controller cards, and drive enclosures. But remember that RAID 1 - 5 only protect against drive failure. If you delete a file, have a crash or get a virus, the damage is done instantaneously to the mirror as well as the primary data, so RAID gives you no protection against that.

I agree totally. If you're looking for speed and are not worried about redundancy go with RAID 0. If you need redundancy, RAID 5 would be your next best option. However keep in mind that read speeds will not necessarily equal write speeds. In RAID 5, reading is much faster than writing.

 

[strydr]

RAID 10 is going to be a little faster than RAID 5 because the system does not have to calculate parity or break up your data. Only issue is you need at least 4 disks. I run a RAID 5 (with no HotSpare) on my file server- It's not a speed demon, just a repository. If you need redundant data, and speed, I would go with a RAID 10.

 

[rickvanr]

I've been thinking about setting up a RAID 0 on my PM. Do both drives have to be identical sizes, ie. 250GB + 250GB?

With Canada RAM's warning of it being the most dangerous method, does RAIDing put an extra strain on drives? Or are the chances of a drive failing just as likely as any other time?