Sign in with
Sign up | Sign in
Your question

New to RAID, please help

Last response: in Storage
Share
August 6, 2006 11:39:01 AM

Right. I know next to nothing about RAID. I know the basics, but that's about it. (Yes, I have read the sticky!)

That being said, in september I start an important course at university - and I do not want to lose any data.

So, using that logic I should use RAID 1 for mirroring, so that I don't lose all my stuff, right?

What sort of performance gains does RAID 0 bring? If it's a decent amount, I was considering RAID 0+1, which as I understand uses 4 HDDs, using 2 of them as regular RAID 0 setup, and the other two as backups of those HDDs.

That way I have the performance of 0 and the security of 1.
Also, I read that RAID 5 is similar to 0+1 - is it?

Oh, and do I need a storage controller? Whatever that is..

Thanks.

More about : raid

August 6, 2006 12:04:40 PM

I was new to raid a few weeks ago too but I've learn't a lot since then. Now to answer your questions.

So, using that logic I should use RAID 1 for mirroring, so that I don't lose all my stuff, right?

Yes, but in Raid 1, one drive is lost to mirroring so you technically only have the capacity of one drive.

What sort of performance gains does RAID 0 bring?

Raid 0 brings increased performance as the data is able to be read and written onto two hard drives at the same time. There is a much increased read and write time using Raid 0 but no data security.

Also, I read that RAID 5 is similar to 0+1 - is it?

Raid 5 is basically Raid 0 and Raid 1 brought together in only 3 drives. It is usually only used by enthusiasts and servers which need data security and capacity. The capacity of a Raid 5 is that of all the drives added together minus the capacity of the largest drive (I think, correct me if I'm wrong). All the drives should be the same capacity for best results. In Raid 5, any drive can fail with no information lost. Just pop in a new drive and the system should rebuild itself. However, people tend to stay away from Raid 5 for home systems as I chose to do. Currently, I'm planning to go Raid 1 and later go probably go 0+1.

Oh, and do I need a storage controller?

A storage controller is the piece of hardware that controls the Raid and keeps it functional. Software Raid is possible but this sucks up sometimes much needed CPU usage and can actually slow down your computer. However, in 2006, most motherboards can with 1, 2, and sometimes 3 Raid controllers' My Asus P5W DH Deluxe I am getting comes with I believe 3, if not 2 Raid storage controllers that can do Raid 0, 1, 0+1, and 5 (Possibly more, have to check the details). Consider this when buying a motherboard.

Hope this helps.
Best of luck.
August 6, 2006 12:12:29 PM

What about real world differences to the average gamer/mp3 playing computer user? Is striping even worth it to me?

I'm unclear about how RAID 5 works.. RAID 0+1 I can understand, but since there's only 3.. if one fails, how could the other two carry on?
Say each HDD has 250gb of memory, so the total is 250*3-250=500gb spread across 3 drives. 166~gb each, leaving 84gb per drive unused. (33%)

Now if drive #3 fails, #1 and #2 have enough space to be holding mirrored data of #3, so all is well - but if #2 fails, #1 is already holding #3s data on its mirrored section.. I don't get it.

Or does the data mirroring take place as the HDD is failing?
Related resources
August 6, 2006 12:27:11 PM

Well, striping would be worth it in certain regards. Like a raptor, striping incrased read and write times but that only affects computers when it's actually using the hard drive. That means that when gaming, the only advantage will be a faster load time, but it won't affect the actual playing. Also, Windows XP will load much faster. As for MP3 playing, loading will be faster (but who notices the 2 ms load time anyway), so basically to the average person, it will not be as much of an increase other than when loading a game or XP.

Raid 5 is confusing, but I figured I won't try to understand how it works, I will just know that it works. If you have 3X250GB hard drives, your total capacity is 500GB, and if one drive fails, put another one, and the Raid will rebuild itself. Once its done, you're good to go again. My best guess is that the data mirroring happens once the drive fails but remember that data parity is spread across the three drives (whatever that means). Don't try to understand because even to me it doesn't make sense, just know that it works. :D  (With a reliable storage controller).

Unless you're a serious enthusiast:

Raid 0 for fast loading performance.
Raid 1 for security.

End of story
August 6, 2006 12:44:53 PM

Don't wast your time trying to raid.........as fast as the new drives are.....don't give up your reliability.....
August 6, 2006 12:49:41 PM

hm. I think as it doesn't affect real performance i'll go with a mirroring RAID for data security. This data is going to be very, very important to me.

I'll either go with a couple of basic HDDs in mirror, or is there a way I can get a fast raptor for the OS/games, a regular HDD for music etc, and a bigger HDD to mirror all of the previous?
August 6, 2006 12:56:41 PM

A single 10k Raptor drive will give you similar performance (better in many cases) to a 2x7200rpm RAID 0 array.

So, my personal choice would be (and thus removing the noise and heat from 4 HDD's)

1x 150GB Raptor as your OS / installation drive, defragged regularly
1x 250GB+ Seagate Barracuda Spinpoint 10 as a storage / backup drive for important data

Ever since the Raptor drives were released, RAID 0 for 2x7.2k drives lost most of its charm - more hassle, noise, and heat, and similar costs (factoring good 7.2k drives).

Unless you were considering 2x10k raptors in Raid 0 :twisted:

**********

NB for security, there's nothing stopping you ghosting your Raptor drive once a week (I'm sure you can set it up to do it while you sleep) and saving that ghost image to the storage/backup drive too. RAID 0+1 with 7.2k drives is all but pointless for most.
a b G Storage
August 6, 2006 1:02:16 PM

Quote:
Don't wast your time trying to raid.........as fast as the new drives are.....don't give up your reliability.....


pfffttt reliablility? if you have one hdd and it dies you loose everything, if you have 2 hdds in raid 0 and one dies... YOU LOOSE EVERYTHING - whats the diffrence?

my setup?

MAIN PC - 2x80gb seagates, RAID0
SERVER - 2x250gb wd's, RAID1

works for me ;) 
August 6, 2006 1:02:55 PM

Agreed. Raid seems to be a bit of a waste of money unless you are really budget tight and or want to waste a lost of money with data protection. Just making a backup of your data is a much more efficient.
August 6, 2006 1:03:20 PM

Would it be better to have RAID 0+1 with four 120GB HDDs, or RAID 1 with two 250gb HDDs?
August 6, 2006 1:05:00 PM

Quote:
Agreed. Raid seems to be a bit of a waste of money unless you are really budget tight and or want to waste a lost of money with data protection. Just making a backup of your data is a much more efficient.


Hate to be a dumbass, but how would one go about backing up all your data?
Wouldn't you need a huge external HDD for that?
August 6, 2006 1:08:11 PM

Lol, I was commenting on what DaveUK said.

1x 150GB Raptor as your OS / installation drive, defragged regularly
1x 250GB+ Seagate Barracuda Spinpoint 10 as a storage / backup drive for important data

Therfore, you can regularly backup your 150 gig raptor, which would run as fast as a Raid 0. The 250gig Seagate is more than enough to back it up with an extra 100 gigs of storage for things like Dvd's or stuff you don't want to clutter up your Raptor with.


Edit:
And why would it have to be EXTERNAL?


Also:

Raid 0+1:

Pros: Fast, Reliable

Cons: Lots of heat, power, and a higher chance of a hard drive failing especially if there's bad cooling. Harder to set up.

Raid 1:

Pros: Reliable. Less heat. Less chance of a failure.

Cons: Doesn't have the speed of a Raid 0.

Typically: 120 gig drives cost about 60-80. 250 gig area hard drives cost something like 100.
August 6, 2006 1:15:36 PM

I'll repeat.

Given that you are talking about 7200 rpm drives

I really think you would be better off running a 74/150GB Raptor and keeping larger files / ghost backup on a larger secondary drive 250-500gb.

**********

4x 120GB drives Raid 0+1
Pros
Increased performance and backup with full 240gb capacity
Cons
Hassle, noise, heat and cost from 4x 7200rpm drives
4x more likely to kill a drive (which means you need a new one to perpetuate your RAID 0+1 setup)

2x 250GB Drives Raid 1
Pros
Safe backup of all your data

Cons

More hassle to create a RAID array
No performance increase whatsoever

1x Raptor and secondary storage drive + Norton Ghost
Pros
Equal / faster loading performance than 2x 7.2k in Raid 0/0+1
Less noise/heat than four drives
More storage space overall
Ability to ghost an image for backup
Don't need to piss around with RAID

Cons

Backing up of information is manual (but can be scheduled automatically)

*****

PS. Is it me or does the guy who just posted before me seem to be COMPLETELY on the same page? lol.
August 6, 2006 1:22:00 PM

Another route to venture down is that of using a usb external backup. Most of the external drives have backup software that affords unattended automatic data backup. I have an Iomega external with Backup Pro - works well. With the relative low costs of external backup drives, this may be prefereable as you don't have additional internal drives eating up system watts and contributing to higher internal case temperatures.
August 6, 2006 1:25:55 PM

Raptor 75GB
320GB Barracuda

That seems like the best bet then, if I want security and speed.
Cheaper than buying four 7200's for 0+1, but I only get 75GB of my data at increased speeds. And I don't have to faff around with RAID controllers or setup.

I'm assuming OS and my favorite games/applications for the raptor, and everything else including a backup of the raptors files on the barracuda?


Oh, but what happens if the barracuda fails? That would suck.
Thanks guys.
August 6, 2006 1:27:08 PM

Quote:
Another route to venture down is that of using a usb external backup. Most of the external drives have backup software that affords unattended automatic data backup. I have an Iomega external with Backup Pro - works well. With the relative low costs of external backup drives, this may be prefereable as you don't have additional internal drives eating up system watts and contributing to higher internal case temperatures.


External drives large enough for full HDD backup will always have a higher '$ per megabyte' than an internal drive. The backup would also be slower than a good internal SATA drive.
August 6, 2006 1:30:38 PM

Quote:
Raptor 75GB
320GB Barracuda

That seems like the best bet then, if I want security and speed.
Cheaper than buying four 7200's for 0+1, but I only get 75GB of my data at increased speeds. And I don't have to faff around with RAID controllers or setup.

I'm assuming OS and my favorite games/applications for the raptor, and everything else including a backup of the raptors files on the barracuda?


Oh, but what happens if the barracuda fails? That would suck.
Thanks guys.


You've got it in one buddy, hope you enjoy your new setup :) 

No problem...

...and if you're really that scared about the barracuda dying, then don't forget DVD-RW or external storage for anything thats REALLY critical.
August 6, 2006 1:44:32 PM

Would this work?

A 75gb Raptor for the OS/favorite games and applications
A 250gb Barracuda for music and a backup of the raptor etc
A 250gb Barracuda mirroring the other barracuda


That way I can't lose anything, and I have the speed for my favorite things.

But.. the price on that is £50x2, and £100 - which is the same as just buying four 250gb drives in 0+1..
August 6, 2006 2:24:50 PM

Quote:
I'm unclear about how RAID 5 works.. RAID 0+1 I can understand, but since there's only 3.. if one fails, how could the other two carry on?
Say each HDD has 250gb of memory, so the total is 250*3-250=500gb spread across 3 drives. 166~gb each, leaving 84gb per drive unused. (33%)


There's no mirroring in RAID5, it relays on parity: in a 3 drive setup (but you can have as many as you want) when you write a chunk of data it's distributed across two of the devices and the bitwise XOR between the two written blocks is stored on the third drive.
At the next chunk the drive used for parity storage changes, in a round robin fashion: this is the more secure way, because if you store the parity always in the same drive you can't say which of them failed. Let's take an example with 3 drives:

BLOCK HD1 HD2 HD3
0 data0 data1 parity
1 data1 parity data0
2 parity data0 data1

and so on.

If one drive fails the controller can reconstruct its data simply XORing the data from the other two drives.
August 6, 2006 5:50:51 PM

Quote:
At the next chunk the drive used for parity storage changes, in a round robin fashion: this is the more secure way, because if you store the parity always in the same drive you can't say which of them failed.


The controller can always tell which drive has failed. RAID 5 stores the parity in this distributed fashion instead of on a single drive because of performance considerations, not because of security.

RAID 3 (rarely used) uses XOR parity for data protection just like RAID 5, but stores all the parity information on a single drive in the array. The problem is that the parity information has to be updated on every write to the array, so the parity drive becomes a bottleneck. With RAID 5, you use distributed parity, which improves write performance. On every write to the array, the parity blocks have to be updated just like RAID 3, but since the parity blocks are on multiple drives, the parity writes are no longer a bottleneck. Instead, the bottleneck falls to the next-slowest device in the system, which is usually the speed at which the XOR information can be calculated. On newer controllers, the XOR engine has become extremely fast, so now the bottleneck is frequently the controller interface to the host system, i.e. the PCI bus, which tends to max out at somewhere around 90-120MB/sec depending on the manufacturer and design of the south bridge on the motherboard. Obviously, a controller card with a PCI-X or PCIe interface eliminates this problem.
August 6, 2006 6:21:41 PM

Yes it would, but personally. Stay out of Raid if you want to keep it simple. Put "most" of your stuff ont he Raptor and back it up to the Seagate often. If the seagate fails, buy another one or use the warranty (5 years I believe) and then just re-back it up.
August 6, 2006 6:24:26 PM

Quote:
The controller can always tell which drive has failed

No, it can't :-)
You can always say which drive has an hardware failure, but not which has inconsistent data if there are no errors in the CRC (and in 70% of cases the CRC doesn't show any error).
Do some very simple truth table calculations and it'll be very clear why.

And finally if you have a slow HD it would be *always* a bottleneck, no matter if you use it for storing only parity data or you use round robin: every write to this drive will be slow and it'll always slow down the transfer rate beacuse the controller can't absolutely start a new write operation before the preceeding has finally completed (except in recovery mode).

In ANY RAID array of ANY type the performance will be which of the worst device in the entire array.
August 6, 2006 6:41:05 PM

Quote:
You can always say which drive has an hardware failure, but not which has inconsistent data if there are no errors in the CRC (and in 70% of cases the CRC doesn't show any error).

And finally if you have a slow HD it would be *always* a bottleneck, no matter if you use it for storing only parity data or you use round robin: every write to this drive will be slow and it'll always slow down the transfer rate beacuse the controller can't absolutely start a new write operation before the preceeding has finally completed (except in recovery mode).

In ANY RAID array of ANY type the performance will be which of the worst device in the entire array.


You're correct, there are instances where a controller can tell that the array has a problem, but not which drive is causing it. I have see RAID 5 failures where a drive developed lots of bad sectors, but still stayed on-line. The controller would log CRC events, but not be able to tell which drive was causing it. In those cases, removing the drives from the array one at a time and testing them with the manufacturer's low-level diagnostic tool or with a third-party low-level diagnostic tool such as SpinRite will isolate the failed drive. Replacing it and rebuilding the array then restores the array to full redundancy.

I disagree that a slow drive will ALWAYS be a bottleneck. Yes, it will be the limit on individual writes (because the controller will always be waiting on that particular drive). But if the wait time for the slow drive is still shorter than the wait time for the interface data transfer, then the data transfer path is still the bottleneck in the overall system. You can see this effect in systems such as a 4-drive RAID 0. Even though one drive might be slow, together the drives could still read/write at 130MB/sec, but are limited to 100MB/sec by the PCI bus or south bridge. Old RAID 5 controllers can limit the transfer rate even more severely via the under-performing XOR unit (old Adaptec controllers in particular).

Under those conditions, replacing the "slow" drive with one that matches the others still doesn't improve the transfer rate, thus by definition it can't be the bottleneck in the system.
August 6, 2006 6:49:40 PM

I agree with you with the bottleneck problem with the bandwidth in regular desktop systems: with desktop Intel chipsets and PCI controller there is really the 130MB/s interface limit (but not with AMD+PCI-e).

Sorry if I missed the PCI/SouthBridge limitations, but I usually work on Opty servers with U320 and PCI-X and in such systems the only bottleneck is effectively the slowest drive.
a b G Storage
August 7, 2006 8:12:25 AM

Quote:
I agree with you with the bottleneck problem with the bandwidth in regular desktop systems: with desktop Intel chipsets and PCI controller there is really the 130MB/s interface limit (but not with AMD+PCI-e).

Sorry if I missed the PCI/SouthBridge limitations, but I usually work on Opty servers with U320 and PCI-X and in such systems the only bottleneck is effectively the slowest drive.


ahem?
August 7, 2006 9:25:06 AM

It's well known that AMD hasn't the Intel bandwidth limit simply because don't use that silly FSB that shares a single bus for RAM and every other peripheral.
AMD hasn't this problem because it uses a dedicated HT link for the southbridge and has IMC.
August 7, 2006 3:37:25 PM

Most of the time, the southbridge is only the limit in older chipsets (8xx series from Intel, for example). Newer chipsets have solved their bandwidth limitations in that regard, from both Intel and AMD.

However, the PCI bus is still a large limit. Obviously, as you have found, PCI-X and PCI-e cards eliminate that limitation.

Intel's architecture for PCI-e is very similar to AMD's now. PCI-e comes off the north bridge chip (MCH) vice the south bridge. Thus the interlink between the two is no longer an issue if you're using a PCI-e storage controller. Take a look at Intel's block diagram of the 975X chipset. The 32 available PCI-e lanes are off of the MCH, and can be divided up amongst PCI-e slots as the motherboard manufacturer sees fit. Generally, you take 16 of them to make your PCI-e x16 slot for graphics, and use the other 16 in various other slot combinations from 1x to 8x.
August 7, 2006 3:52:11 PM

I have no idea what you guys are talking about now.
All I know is I play games, mp3s, and do a lot of very, very important college work on my computer. I need to keep this data secure.
I can't really afford raptors tbh.
August 7, 2006 4:11:51 PM

Lol, ya, little bit lost too now :? . Basically, if you want speed, it's either a raptor (which you stated you can't get), or a raid 0. However, raid 0 is not secure. So raid 1 is your best choice, but even then, the easiest option of all, is to get one 7200rpm Sata 2 hard drive and then an even larger one to manually back it up onto and for additional storage. I've decided to do the same and skip out the whole headache of RAID.
August 7, 2006 4:27:37 PM

I say only this,,, back up your data to dvd or cd or whatever. Disks can fail no matter what your back up should not if you are smart carefull about doing it.
August 7, 2006 6:15:57 PM

Quote:
I have no idea what you guys are talking about now.
All I know is I play games, mp3s, and do a lot of very, very important college work on my computer. I need to keep this data secure.
I can't really afford raptors tbh.


My apologies. We did digress just a bit, didn't we? :?

As I understand it, you need data security as the primary goal, low cost as a secondary goal, and tertiary goals would be ease of setup, capacity, and speed.

I see 3 different options for you here:

Option 1: 2x250GB in a RAID 1 on a Promise FastTrack TX2300

- Very secure
- Reasonably low cost
- Setup is not trivial, but we'll walk you through it if you like
- Not especially fast. Will be slightly slower than a single drive on writes, somewhat faster on reads.
- 250GB of total storage

Option 2: 3x250GB in a RAID 5 on a Promise FastTrack TX4310

- Very secure
- Cost is higher than option 1, both due to 3 drives vice 2, and more expensive storage controller
- Setup not trivial, but we'll walk you through it if you like
- Reads faster than single drive, but not as fast as RAID 0. Reads faster than writes. Writes could be slow on this controller, XOR is not done in hardware.
- 500GB of total storage

Option 3: 3x250GB, 1 for C:, 2 for D: in a Windows software RAID 1

- D: drive is secure. C: drive is not.
- Reasonably low cost
- Easier setup, since there is no storage controller (assumes you have 3 available SATA connectors available on your motherboard)
- C: drive is single drive. D: drive somewhat faster on reads than single drive, slightly slower on writes. Not as fast as RAID 5 or RAID 0.
- 250GB unprotected storage, 250GB protected storage.
- Slightly higher CPU utilization during drive access from D: than other options because Windows is doing RAID work.


For you, I would dismiss options involving RAID 0+1 (requires 4 drives, likely too expensive) and expensive storage controllers that have hardware XOR. You said you didn't want to pay for a Raptor, so I left that out of these options as well.

Also, allhell has a good point. RAID can only protect you to a certain degree. You're protected against a drive failure, but that's it. RAID won't protect you from viruses, spyware, OS corruption, another user deleting all your files, fire/flood, etc. The only solution against those threats is a well-thought-out backup strategy.
August 7, 2006 7:12:32 PM

Option 1 seems decent. I'm not too bothered about how long it takes me to write things to the HDD, I can wait an extra hour to install the game for all I care - but if the loading times once it is installed are quicker [read times] then that's good.

I didn't know RAID 1 read quicker than a vanilla HDD.
August 7, 2006 8:47:02 PM

Quote:
I have no idea what you guys are talking about now.
All I know is I play games, mp3s, and do a lot of very, very important college work on my computer. I need to keep this data secure.
I can't really afford raptors tbh.


No need for expensive RAID stuff.

For School work (critical data):
Your best bet is to store your important information on the school's network. It's free and backed-up with fault-tolerant hardware (assuming it's a reasonably equipped university). Keep a copy locally and/or burn a CD for extra insurance when you finish a piece of work. This is the safest since anything short of a meteor is unlikely to take out both your's and the school's computers at the same time. If school has no storage for you, sign up for free gmail account (or other provider) and use their online storage, many providers give gigabytes of free storage.

For mp3's, pr0n, etc (non-critical data):
Burn to CD or DVD. Buy a 50-pack and use back-up software (lots available for cheap or free). The software will keep track of back-ups so only the new files need to be burned.

Don't waste time backing up OS and programs, they can be easily replaced from install disks. Just protect your data.
!