Newbie RAID questions...

I'm thinking about getting myself a RAID setup in the near future, but have a few questions...

1. First of all, which type of RAID setup would be best for me? My whole reason for wanting RAID is to protect my data against HD failure (I've lost several HDs in the past and it's no fun). Performance increase would be nice, but it's not my objective. I'm just sick of losing data which is why RAID appeals to me.

A friend of mine told me you can set up a RAID array so that if one of the drives fails, you don't lose any data at all, and you get the storage capacity of all the drives minus one. This is the sort of thing I'm after (er, I think). Would all the drives have to be the same size for this kind of setup?

2. I don't use my computer for games — I do a lot of a video capturing, video editing, looking through photos, and photo apps. Would getting RAID improve performance with any of these? I'm hoping so, especially with regards to waiting for thumbnails to generate in my photo viewer.

3. Second, my ideal total capacity is about 2tb. I won't be using anywhere near this much right now, but I predict I'll need at least a couple of terabytes in about 3 years' time. Would it be better to use a small number of large drives (4 500gbs) or loads of smaller drives (10 200gbs)?

4. And what about expandability? Is it possible to add new drives to a raid array? That would mean I could get a smaller capacity now, then buy a new drive each time I run out of space, which would be cool. If this is possible, would I be able to add drives of larger capacity to the original ones?

5. How are RAID drives normally housed? I'm guessing all those drives won't fit in a regular computer case anywhere. Do people normally have them in a separate case with its own power source? This idea appeals to me, because I'd love to keep my drives safely tucked away in a cupboard where they're cooler and less likely to be damaged/knocked.

Any answers/advice at all would be appreciated. (Sorry again for the lameness of these questions, as you can probably tell, I'm no expert!)

before i can answer any of your questions, how much money do you want or can you spend? with raid, you can spend several thousand dollars really easy between controller card(s) and hard drives. also, how much of an issue is speed? do you need to transfer large amounts of data frequently?

EDIT: i can, howerver, answer your 5th question. external inclosures are usually for scsi arrays, but you would be amazed how many hard drives can fit inside a computer case. just do a search on newegg.com or some place and you will find plenty of cases meeting these requirements. but again, please refer to my first paragraph.

for terabytes of data, there is this huge hard drive that can store 1.5 terabytes, it is, however, extremely expensive.
for data protection, you want raid-1, whatever is writen on your first drive is copied to your second one. for speed, you want raid 0, but in this one, you lose ALL YOUR DATA if one fails. for both speed and safety, you want raid 0+1, this you need 3 hard drives. two in raid 0, same capacity, or identical drives, and another one being the storage of one drive x2. then to answer your question of how many drives. well, you want less but huge hard drives, saves money on your electrical bill and saves space as well. for terabytes, i guess theres no choice but to buy the huge 1.5 terabyte external hdd (it wont fit in 5 1/4 inch bay) or to get two 500 gb hdds if u can find some. and in raid 0, you lose space, since you are copying everythign on one to the other.

Hmmm... ideally, I'd like to spend no more than $900, but I'm willing to accept that I'll have to spend more like double that for anything approaching a terabyte, and double that again for a couple of tbs. Which is why I'd like to be able to build up the capacity as I go, by adding on more drives. Would this be possible?

As for the setup, I was thinking of something like this: 4x 250gb drives. The capacity would be 750gb, as one of the drives wouldn't count. Then maybe I could add more 250gb drives when I get low on space, but I would still only lose 250gb from the total capacity.

This is how I thought RAID worked, but I've been told this by my mate who isn't as knowledgeable as he claims to be!

cant really work that way... because logically, you can only have 4 sata and 4 pata drives in total in any system. plus you would have to get alot of power because each drive draws 25 watts. that means a bigger electrical bill. and also, the way raid 1 works, is that it copies everythign from one drive to the backup drive, meaning that if you want to raid 1 4 hard drives of 250 gb, you would only have 500 gb left because you have to copy the hard drives 1&2 to 3&4. and also, im not sure if you can do that, having many hard drives in raid 1 or 0 since controllers dont really work taht way. what your friend says is true to a certain extent. what he didnt tell you is that you can have maximum two hard drives in raid 1 and 0. more than that you want other raid arrays like 0+1 or 3, 4 or 5. however, because you are copying everything, you basically lose HALF the space you have in total. so if you buy a terabyte of space, you only have 500 gb because you are using the other half to copy the first half. two terabyte, you only have one.

and to add more and more space, might as well have no RAID at all...

I have to disagree with these statements. I think SCSI will handle up to 32 devices on a single controller. What about PCI or PCI-Express slots? Sure the motherboard controller can only use up to 4 SATA and 4 PATA. However, you still have room using even old PCI SATA Controllers on the PCI bus. You'll constain the drives to the PCI's 133MB/s transfer rate. I only get a max of around 100MB/s (sustained) on my motherboard's SATA RAID controller anyhow. I'm sitting here looking at my PCI SATA Controller right now. It's a Promise that'll take 4 drives on its own, using RAID 0, 1, 0+1, and RAID 5. Another advantage of using a RAID controller is that they're available with onboard RAM and a co-processor, making the RAID less strenuous to your systems CPU and RAM.

RAID levels can more easily be explained by looking here. I would suggest RAID 1, 0+1, or the bulletproof RAID 5 for data retention. And sorry, no, you do not lose %50 of your hard drive space by using RAID. RAID 5 has the highest overhead I've experienced and in most cases, about %30 of the total volume's space was used for parity. My RAID 0 uses 2 80GB disks. I still retained about 150GB of usable space after RAIDing them together.

I just don't understand that last sentence...
So, you mean that somehow terabytes of space for enterprise applications are constrained to only using the onboard ports? That doesn't make any sense. Especially when I've took apart a 42U server rack completely full of fiber SCSI enclosures, each holding 14 drives. Not to mention I've built single machines that can handle 10 onboard drives.

I've had a look at a couple of RAID tutorials and I'm pretty sure the type of RAID I want is RAID 5. This one uses one parity bit for every 4 regular bits, so you have a reduction of 20% of the total disc capacity. But for that you still get complete redundancy. The only way you can lose your data is if two drives fail at the same time. Which, I'm guessing, is unlikely if you take good care of them.

what i was thinking is in raid 1 you lose 50%. i was in a rush there, so i forgot to mention the add-on cards. but yeah, once you put loads of space (and hard drives to acheive that), you are looking more at a server than a desktop... and yes, your raid-0 statement is correct, or else i dont see why people would want to try it anyways, including myself
and for more space, well, if you add more drives, to me, its a bit more work to add it in and configure it in the raid array and all, im more of the "plug 'n play" guy.

and speaking of "daisy chain scsi" you can always buy firewire or usb external drives, but i dont know how you are going to raid usb drives... and also, they arent the fastest ones either, atleast i dont think unless you spend alota cash.

SATA intends to replace SCSI, particularly in the RAID department, so throwing in a controller for additional SATA ports would work well, but he's looking for a helluva lot of redundant storage and would be better served looking at a corporate-level server storage rack.

Sorry, this thread has gone over my head a little and I'm confused about who's addressing whom!

Let me ask some new questions...

How would someone like me (non-techy) go about setting up a RAID system? What hardware will I need apart from the drives? Do I need extra power units/plugs? I'd really like the array to be housed in its own case, separate from my main computer case, but if it's a lot easier to put them in my main computer, I will.

I think RAID 5 would suit my needs best. I just want my data to be safe really, although a performance increase would be nice. Is RAID 5 the right choice?

Do all the drives have to be the same size for RAID 5 array? And is it possible to expand the array by adding more drives at a later date (or is the number of drives fixed from the moment you turn it on)?

Oh, and could anyone recommend a cheap HD supplier in the UK?

the drives don't have to be the same size, but the entire array is limited by the smallest drive. If you think you've got enough power, use this good chunk of information to test and see if your amount of usage will work with your power unit:

http://www.devhardware.com/forums/t55173/s.html

If your motherboard supports RAID 5, then use your motherboard booklet and follow the instructions. What type of motherboard do you have? If not, you could use windows software for software RAID (I believe), or I do believe they sell RAID adapter cards for the PCI bus.

this card will give you the ability to have raid 0+1
http://www.newegg.com/Product/Product.asp?Item=N82E16816118007

but as you said that you wanted RAID-5...

here's a raid-5 card for IDE devices:
http://www.newegg.com/Product/Product.asp?Item=N82E16816115016

here's one for SCSI:
http://www.newegg.com/Product/Product.asp?Item=N82E16816118012

and here's one for SATA:
http://www.newegg.com/Product/Product.asp?Item=N82E16816115023

I hope that helped

oh crap I just realized you're in the UK. Damnit....could always take those product numbers and put them into a place over in the UK that ships over there.

Well, I would suggest using the same (size, speed, brand) drives for a RAID. Ideally, all of them should be new.

The case interests me too. You would like to make them external. Then you'll need a storage enclosure for standard size hard drives. These are hard to find IMO. I've looked several times for other people asking the same questions you are. Instead, I would suggest the CoolerMaster Stacker. I found this one to be out of stock at the particular site I mentioned. However, you should be easily able to find it somewhere in the UK. There's little caddies that are internal mountable, shown near the bottom. So, you could still use SATA drives and multiple RAID controllers in the same box. There's alot of SATA cables involved, so a large SATA RAID (8 drives) would best be housed in this case IMO. The cost involved is something to consider though.

RAID 5 is really hard to defeat through drive failure. Promise Technology makes very nice SATA RAID controllers with onboard RAM and co-processors. The other thing about the Promise brand is that their controllers make SATA drives hot-swapable. Meaning you could change out a failed drive while your PC is still on.

correct me if I'm wrong, but can you use RAID-5 with only 4 discs?

Well, RAID 5 is the best method for a user who wants to protect their data and doesn't want to spend tons of money on hard drives and then lose half their capacity. Here are the types of RAID:

RAID 0 takes your data and puts half of the file on one HDD and the other half on the other HDD. When it reads the data it comes off of 2 HDDs simultaneously, therefore increasing the sequential read performance substantially. If one drive of this setup fails, all data is gone. This is definitely fast, but not what you want.

RAID 1 takes any data written to one drive and doubles it onto another one. Both operations are done simultaneously and so transfer rates you would expect from one single drive are what you should get. You lose half of your maximum theoretical storage capacity because you need 2 drives to write what one drive can take. Very secure, very expensive.

RAID 5 (and please somebody correct me if I'm wrong) stripes the data across all drives of the array, minus one. There is no one drive set exclusively for parity information, this information is kept on each drive and is the equivalent of 1/X of the maximum possible storage capacity of all the drives combined, with X being that maximum possible capacity. The tolerance of RAID 5 is quite good. You can lose a drive and still get your data back without too much of a problem. No more than one drive may fail however, though this should not be a problem. This type of RAID is fast as well, since it is striping information across the drives. This is why it is widely used for data protection and speed.

Now as for whether you should use many small drives or a small handful of larger drives, I would say that less is almost always better. You both reduce the chance of failure of any single drive, since there are less, and you keep your power requirements at a minimum. For your ideal "2TB" of storage, you would need 5 500GB drives. You would of course fall slightly short of 2TB because 500GB drives are more like 450GB I believe, but that's another story altogether. This presents another problem: most controllers handle UP TO 4 drives, not five. You would need to get a higher end controller to accommodate the fifth HDD. You could also stick with 1.5TB of storage and run it all on a cheaper controller. This of course, is up to you.

As for expandability, I don't think that that is going to happen. I'm pretty sure that you can't just drop another drive onto your array and make it work. You would need to move all your data over to another medium, rebuild the array and then copy it all back over again. This is of course no easy task with 2TB of data to move. I don't think that you should expand an array later on anyways, since you would be putting year old drives with new drives and you would then have different life expectancies for all the drives. Kind of like changing the tires on a car at different times.

For housing your drives, it would be easiest to do in a standard ATX case. If you go with 4 drives then all cases that you buy should be ready to handle that. If you are worried about heat then you could also get an HDD cooler for each, though that would take up a 5 1/4" drive bay every time. I still think that you are better off having all your drives in your case. I don't see any other way, though I know that it can be done.

One last thing: make sure that you get SATA drives and not IDE drives. Sounds stupid, but just make sure that you do that. lol Good luck with your RAID array!

You can add to a RAID 5 array (depending on your controller and supporting software), but it takes forever for the array to expand. I expanded one for a client at my old job once. I added 3 72GB SCSI (Ultra160) drives and watched it take something like 4 hours for the array to rebuild. Still other controllers will let you do it in a hot-swap manner.

RAID 5 is possible with only 3 drives, but I wouldn't suggest it. Using a case like the stacker, you could easily add more space by adding another controller and more disks. Plus, the stacker comes dual power supply ready. More disks need more power, a reduntant PSU may be needed. And, you could build your ideal 2TB of space using more, (hopefully) less expensive disks.

I should mention that if one drive in a RAID 5 fails, you'll lose performance while your array rebuilds after you swap the bad drive out. The same goes for adding in more disks, if it's allowed.