Author Topic: RAID Boxes (Read 4112 times)

Tatsujin · « **on:** August 02, 2009, 03:40:25 PM »

I don't understand them as much, the benefits and the disadvantages (if there is any). Also setting them up, is it friendly to set up (plug and play like putting an external HDD?) or is it complicated?

Here's a picture I found:

http://www.flickr.com/photos/dannychoo/3464325315/sizes/o/

BuriaL · « **Reply #1 on:** August 02, 2009, 05:42:26 PM »

The advantages with raid is that you ether can "double youre data" and if something goes wrong and a disk is fucked, you wont loose data.
Or you can make several disks into "one" and theyll work faster than normal.

External raid is prolly just for massive storage?

Edit: Oh its you, i guess you know it allready, dont you.

Tatsujin · « **Reply #2 on:** August 02, 2009, 05:45:16 PM »

Quote from: BuriaL on August 02, 2009, 05:42:26 PM

The advantages with raid is that you ether can "double youre data" and if something goes wrong and a disk is fucked, you wont loose data.
Or you can make several disks into "one" and theyll work faster than normal.

External raid is prolly just for massive storage?

Edit: Oh its you, i guess you know it allready, dont you.

You mean like 1TB HDD would become 2TB? And in other words, you can combine those hard drives into one, or something?

bobjoe · « **Reply #3 on:** August 02, 2009, 06:02:55 PM »

Quote from: Tatsujin on August 02, 2009, 05:45:16 PM

Quote from: BuriaL on August 02, 2009, 05:42:26 PM
The advantages with raid is that you ether can "double youre data" and if something goes wrong and a disk is fucked, you wont loose data.
Or you can make several disks into "one" and theyll work faster than normal.

External raid is prolly just for massive storage?

Edit: Oh its you, i guess you know it allready, dont you.
You mean like 1TB HDD would become 2TB? And in other words, you can combine those hard drives into one, or something?

No, a 1TB HDD would become ~500 GB.
It's just writing the data to to places instead of one so that you won't lose it.

And yes, the other kind of raid would combine two drives into one so that you can access it more quickly, easily, whatever.

a10112 · « **Reply #4 on:** August 02, 2009, 06:06:24 PM »

as per an explanation of the picture you linked: it's a "box" with "cartridges" that may be filled with HDDs and removed as such (given that the bays are hot swappable) http://en.wikipedia.org/wiki/Hot_swapping

BuriaL · « **Reply #5 on:** August 02, 2009, 06:11:41 PM »

If you had 2x 1TB drives and "Mirror" them they are exactly the same. So when one breaks down the other is backup.

If you set them up as "Striped", you get faster disks, but if one breaks down then the whole arrays data is screwed.

The Wikipedia can tell you a bunch of stuff. And motherboards manuals usualy comes with a introduction on raid.

Arveene · « **Reply #6 on:** August 02, 2009, 09:39:53 PM »

Tatsujin, I would suggest reading this. It's a fairly short explanation of what RAID is, the differences between RAID levels, and a bit of additional stuff. For your purposes, you'd probably only need to read about RAID 1, 5, and maybe 6.

The key advantages of using an external RAID for storage would be the following:
1. Data redundancy - RAID 1, 5, and 6 all allow for a hard drive failure without the loss of data.
2. You no longer have to store that data on your desktop computer. Most external raids can be hooked up to your network and you can stream media from it. (NAS)
3. Alternatively you can still have your copy on your computer and make a backup directly from your PC to the external unit. (DAS)

RAID right now usually isn't that hard to set up. It depends on your implementation and what you want to do with it. It should be a fairly simple task to set up a 4 disk external RAID and attach it to your network (NAS), or PC (DAS). A bit of a warning. A 4 disk external NAS would be fairly expensive to setup, and at that point you might be better off setting up a home RAID server instead. Network attached allows for streaming to any computer in your network, but might take a bit of tweaking to set up. Direct attached storage usually will require a PCI or PCI-X RAID card to be installed in your PC. So both implementations will require some work, but online sources for help and documentation is pretty good.

tl;dr of the link
RAID 1 = Mirrored. You take the data capacity of the smallest hard drive, multiply by the amount of hard drives and divide by 2 for your total capacity. (ex: 4x 1TB HDs = 1TB*4/2 = 2TB of mirrored storage.) RAID 1 allows for 2 disk failures in one mirrored set.
RAID 5 = Striping with parity data. You take the data capacity of the smallest drive, multiply by the amount of drives, and subtract 1. (ex. 4x 1TB HDs = 1TB*4-1TB = 3TB of storage with parity data.) RAID 5 allows for any one disk failure.
RAID 6 = Striping with dual parity data. You take the data capacity of the smallest drive, multiply by the amount of drives, and subtract 2. (ex. 4x 1TB HDs = 1TB*4-2TB = 2TB of storage with dual parity data.) RAID 6 allows for any 2 disk failures and also protects against another drive failure while the array rebuilds. If another disk goes down while rebuilding in RAID 5, you lose everything. The same can happen in RAID 1 if the "wrong" disk goes down while rebuilding. It also has an advantage over RAID 5 being able to protect data against an uncorrectable bit error rate.

Each of these RAID levels have a certain amount of redundancy, and different levels of performance. For just storage the performance hit won't be noticeable.

There's both benefits and disadvantages to each RAID level. I run with a 6x 750GB RAID 6 in my own desktop right now for a total of 2.72 TB of storage. Once I fill that up, I'll probably move my RAID card and drives to a home server instead and expand it.

Edit: I forgot to mention JBOD, that's covered in the FAQ briefly. I won't go over it though because quite frankly it's a waste of time and a waste of a storage controller.

geoffreak · « **Reply #7 on:** August 02, 2009, 10:00:30 PM »

Keep in mind that redundancy is NOT a backup solution and should NEVER be used as such.

kureshii · « **Reply #8 on:** August 02, 2009, 11:31:31 PM »

A good RAID discussion can take a long while; feel free to drop by #bakabt again if you prefer the discussion in real-time

fohfoh · « **Reply #9 on:** August 03, 2009, 01:52:42 PM »

You guys are mixing some things up. I don't know much about Raids...

But a 1 TB HDD will always be a 1 TB HDD. You can't really mux it into any thing else, and even if you can, it's pointless. The point of RAID is to make use of multiple HDDs at a hardware level. So basically, a mirror Raid (1,5,6 as mentioned above) is the amount of "mirroring" that occurs. That is, your information shows up, TWICE. Put it this way, you're basically just making a (for a lack of a better word, "backup") on your computer.
New
However, the above is assuming mostly the same HDDs inside. Let's look this way.

If you have a raid set up in which you use 500GB x2 (for a total of 1 TB) for your system, then files are distributed across those two HDDs and used simultaneously. Two HDDs spinning at the same time can read files a lot faster than a single drive. It's moreso to eliminate a bottle neck. That to an extent is what people mean by "striping". Instead of a solid piece of paper, you distribute the files across 2+ HDDs in "cut up strips" of the original file. Then, when you access the files, you read them simultaneously on several HDDs for a faster reaction. (Reduction in bottleneck). However, i've only been talking about raid 0. In 0, no mirror occurs. So Raid 0 is only about performance.

Now, looking at the pictures above, I assume you want to have a mirror "backup". This looks into raid 1, 5, 6. Now, I can't really get into the differences, but how it works is this. You have 2 sides with approx equal HDD space. So, either you can do 1TB:1TB or 2x 500GB:1TB etc. (as long as you somehow can set it up hardware wise.) One side is used as the system use. (anime watching, storing, etc.) the other side is a total copy for stability purposes. (sort of like a backup of company files, and not a backup in the sense of "backing up" random files). "Now how it works is this. " remember the days before daemon tools? The "back up burn of your favourite game"? How it works is this. You have 2. One gets scratched to shit and unusable. The other "backup" is used. You replace that destroyed copy ASAP. Same case here except the data is always updated and not "set in stone" like a burned disc.

Now with my blah blah that might not be fully correct crap done, let's talk some other specifics.

Hardware vs Software raid.

Hardware raid has HDDs preset as used and the mirror sides. Just put in the right amount of HDD space on each side, and it will happily chug along with you without much push on the cpu.

Software raid just sits in the back and copies the required changes to the other side when necessary. Uses a little bit of CPU, but basically the same function.

More gibberish

Why the heck did I say 1TB: 1TB? Raids aren't really Raids without 3+ more HDDs. The main points of the Raid is #1, multiple HDDs seeking system, file etc info simult so you reduce HDD spin rate bottleneck. #2, goes into mirroring for a "sort of backup" so you don't lose data.

Performance Raid focuses on point 1
Storage/backup Raid focuses on point 2

Any wrong info here, please correct. I admit I'm a Raid n00b.

bcr123 · « **Reply #10 on:** August 03, 2009, 04:03:42 PM »

Raid can accomplish different goals depending on how it's setup:

For small raids (2-4 disks)

Raid0 - maximum performance -- by distributing the read/write IO across multiple physical disks, for systems that need to move a lot of data but don't need to store it long term (think of scratch disks for example) It's not a big deal any more with TB+ disks available but it also allowed creation of very large partitions by combining multiple disks back when individual drives were small. The big disadvantage of RAID0 is that failure in any one drive loses all data on the set, so you are adding risk to gain speed.

Raid1 - maximum redundancy -- by duplicating the data on multiple drives hard drive failure is a non-event (usually), same speed as a single disk though.

Raid5 - Better than stock speed, minimal hardware to achieve redundancy, IO, data and parity is spread across multiple disks, minimal amount of "wasted" capacity to achieve redundancy, can recover from a single disk failure easily but performance is reduced during the rebuild of the failed disk, generally used where high-availability and good performance is crucial (servers). RAID5 requires 3 disks minimum to operate, it's practicality is limited to the 4-8 disk range, above 8 you start thinking Raid6 for enhanced redundancy.

halfelite · « **Reply #11 on:** August 03, 2009, 06:38:20 PM »

A good thing to remember software raid cannot always be moved to a new system. IE if the software is part of the bios. If you run nix and use something like mdadm then you can easily move it to a different machine.

Arveene · « **Reply #12 on:** August 03, 2009, 07:33:13 PM »

Quote from: fohfoh on August 03, 2009, 01:52:42 PM

Now, looking at the pictures above, I assume you want to have a mirror "backup". This looks into raid 1, 5, 6. Now, I can't really get into the differences, but how it works is this. You have 2 sides with approx equal HDD space. So, either you can do 1TB:1TB or 2x 500GB:1TB etc. (as long as you somehow can set it up hardware wise.) One side is used as the system use. (anime watching, storing, etc.) the other side is a total copy for stability purposes. (sort of like a backup of company files, and not a backup in the sense of "backing up" random files). "Now how it works is this. " remember the days before daemon tools? The "back up burn of your favourite game"? How it works is this. You have 2. One gets scratched to shit and unusable. The other "backup" is used. You replace that destroyed copy ASAP. Same case here except the data is always updated and not "set in stone" like a burned disc.

I bolded the important part. If I'm reading that correctly, you're mentioning 3 drives? 2x 500GB drives, and 1x TB drive? Unless you're using JBOD (waste of time), any RAID setup using those 3 size drives will only use 500GB of the 1TB drive, not to mention you can't use RAID 1 or 10 (01) with 3 drives. I don't know of any RAID controller that can use the 2x 500GBs as a TB and mirror it onto another.

Then there's hardware vs software RAID, and even with a RAID card for an external solution it's not always a full hardware RAID. It's important to look into the details and differences between software, hardware, and fake hardware RAIDs. Like kureshii said, we could discuss RAID for days.

halfelite · « **Reply #13 on:** August 04, 2009, 12:01:36 AM »

and if you are going to start a raid project make sure you budget it out first. good hardware raid cards run in $500-1000 depending on how many ports, what the proc speed on the card is and how much cache it comes with.

zfs in opensolaris has been showing some good points but its very confusing to setup for first timers.

fohfoh · « **Reply #14 on:** August 04, 2009, 11:38:54 AM »

Quote from: Arveene on August 03, 2009, 07:33:13 PM

Quote from: fohfoh on August 03, 2009, 01:52:42 PM
Now, looking at the pictures above, I assume you want to have a mirror "backup". This looks into raid 1, 5, 6. Now, I can't really get into the differences, but how it works is this. You have 2 sides with approx equal HDD space. So, either you can do 1TB:1TB or 2x 500GB:1TB etc. (as long as you somehow can set it up hardware wise.) One side is used as the system use. (anime watching, storing, etc.) the other side is a total copy for stability purposes. (sort of like a backup of company files, and not a backup in the sense of "backing up" random files). "Now how it works is this. " remember the days before daemon tools? The "back up burn of your favourite game"? How it works is this. You have 2. One gets scratched to shit and unusable. The other "backup" is used. You replace that destroyed copy ASAP. Same case here except the data is always updated and not "set in stone" like a burned disc.

I bolded the important part. If I'm reading that correctly, you're mentioning 3 drives? 2x 500GB drives, and 1x TB drive? Unless you're using JBOD (waste of time), any RAID setup using those 3 size drives will only use 500GB of the 1TB drive, not to mention you can't use RAID 1 or 10 (01) with 3 drives. I don't know of any RAID controller that can use the 2x 500GBs as a TB and mirror it onto another.

Then there's hardware vs software RAID, and even with a RAID card for an external solution it's not always a full hardware RAID. It's important to look into the details and differences between software, hardware, and fake hardware RAIDs. Like kureshii said, we could discuss RAID for days.

Oh, it can't be done? I guess I was given wrong information then. Someone said they set up a raid with 2x500GB+1TB with 2x 1TB on the other side and was recommending those WD Scorpio green or w/e for low power use on the other side of the raid.

Sorry, wrong info. Thanks for the correction.

Talapus · « **Reply #15 on:** August 04, 2009, 05:02:53 PM »

I understand how parity works on RAID 2-5. But can anyone explain how the double parity in RAID 6 works? I've done some reading on R-S code, and while I can understand most of the math, I can't seem to put the pieces together.

halfelite · « **Reply #16 on:** August 04, 2009, 05:15:39 PM »

It works the same as raid 5 except the parity is on two drives. its like the parity chunks is raid 1 and mirrors to another drive. So you can technically lose 3 drives before you lose data.

Talapus · « **Reply #17 on:** August 04, 2009, 05:33:45 PM »

Quote from: halfelite on August 04, 2009, 05:15:39 PM

It works the same as raid 5 except the parity is on two drives. its like the parity chunks is raid 1 and mirrors to another drive. So you can technically lose 3 drives before you lose data.

No. Normal parity (including mirrored parity) cannot compensate for the loss of two data blocks for a given stripe. And since RAID 6 rotates parity, that would mean the the loss of ANY two drives would compromise the data integrity.

If you had mirrored parity and dedicated parity drives, you could loose both parity drives, or one parity drive and one data drive (but not two data drives).

Arveene · « **Reply #18 on:** August 04, 2009, 07:09:10 PM »

RAID 6 will remain intact with the loss of any two drives. One more drive loss before it's rebuilt will wipe out the array.

The wikipedia page has a little bit on the math and links this as it's source. It might be worth a read.

halfelite · « **Reply #19 on:** August 04, 2009, 08:34:14 PM »

Quote from: Talapus on August 04, 2009, 05:33:45 PM

Quote from: halfelite on August 04, 2009, 05:15:39 PM
It works the same as raid 5 except the parity is on two drives. its like the parity chunks is raid 1 and mirrors to another drive. So you can technically lose 3 drives before you lose data.

No. Normal parity (including mirrored parity) cannot compensate for the loss of two data blocks for a given stripe. And since RAID 6 rotates parity, that would mean the the loss of ANY two drives would compromise the data integrity.

If you had mirrored parity and dedicated parity drives, you could loose both parity drives, or one parity drive and one data drive (but not two data drives).

That is what I ment. 2 drives down and you still have a functioning system to rebuild with. One more loss and you are out of luck.