Building a cheap 2TB RAID Server for home

I like having a bit of space to store stuff at home. Dumping the Media Centre, Music, Photos and Backups back to a central server makes sense to me. It lets me use my XBox (chipped original) as a brilliant DivX player with AC3 Surround. I also worry about losing this data, and find it impractical to back it up to CD (3000 CD’s is difficult). This is why I run RAID on my server, to help protect my data. It’s not perfect, but it’s a good start.

You’ll need an old PC. This is a great way to reuse your desktop when you upgrade. Any old piece of junk will do, but ideally it will have the following:

1. Tower Case
2. 300W Power Supply (250W may do at a pinch)
3. 4 x SATA Ports
4. 2 x IDE (PATA) Ports
5. Any CPU – doesn’t matter
6. 512MB RAM
7. On Board Video preferred

The intent is to run 5 x Hard Drives and a CDROM. You can do this with all IDE (PATA) but it gets difficult to find IDE adapter cards, and their performance can be poor. You can run two drives per PATA cable vs 1 per SATA.

You want 4 drives the same size. 500GB is about the most cost effective per GB at the moment (mid 2007), but do the numbers on whatever gives you the best deal at the time. You could use 5 – 10 old smaller drives, however finding a controller to plug them all into is next to impossible. 4 x medium size drives tend to be the most cost effective way of getting bulk space. The fifth HDD can be any old piece of junk. It’s only to run the OS, and can be replaced if it dies. It doesn’t hold any data.

The case doesn’t need to have mounts for all these drives, and it doesn’t have to be anything special. If the drives are all 7200RPM, they tend to get a little warm. They should have some space between them and ventilation. If they are 5400RPM’s you can stack them together on the base of the PC, they won’t care.

As it is a server you can shove it anywhere in the house, making super silent a lower priority. If you want to build it silent see my “building a quiet PC article“. I prefer to shove it under the stairs where noise isn’t a problem. I manage it through Remote Desktop, so it doesn’t even need a monitor, keyboard or mouse.

Install Windows Server if you have it (licensed and legitimate of course) or XP onto the single small HDD. Nothing special here. If you are running XP you’ll have to do some hacks that are covered in this article on Tomshardware. I haven’t found a hack to get Vista to support RAID 5 yet. Windows Home Server should be an interesting one to evaluate when I comes out, as it has some new takes on RAID. In the meantime however, I’ll stick with Server 2K3.

The objective is to protect the data through the use of RAID. I see plenty of people with data on a single machine at home they use as a server, but most don’t run RAID. I always run Software RAID, that way if a HDD dies, I lose nothing. I prefer software RAID in windows to any of the Hardware RAID adapters for home use. The reason is simple. If you Hardware RAID adapter fails, you have to get one the same or compatible to get to your data. Being cheap equipment, this has a fair likelihood of occurring, and then not being available. Using Software RAID on the other hand means any Windows Server can read the disks, and recover the data. We are doing this for data safety, not performance. I have always found software RAID performance to be acceptable, even when I was running an old Pentium MMX 166 box.  In a business environment it’s a very different story and cost.

Once Windows is installed I recommend you configure RAID 5 across the other 4 data disks. There are instructions here.  The disks need to be the same size to participate in the RAID pack. They can actually be different sizes, but you’ll waste space on the larger disks. Now most RAID solutions support two different types of RAID to protect your data – RAID 1 and RAID 5, which I’ll outline.

Raid 1 is a Mirror. It makes a copy of one disk onto another, and shows them as a single disk. If one dies, you have the data on the other disk, and the failure is transparent. The problem is you lose 50% of your space. Your 2 x 500GB disks (1TB total) give you 500GB useable space. Not particularly efficient. To make it practical for a home server you have to buy larger more expensive disks. This is supposed to be a cheapish exercise.

RAID 5 is referred to as a Stripe with Parity. The redundant data is spread equally across all of the disks. You need a minimum of 3 Hard Drives to run RAID 5. The way the algorithm works you lose the effective space of ONE disk to the data protection. If you think about it, the more disks you add, the more efficient it is. The data being spread over multiple disks tends to make the system faster as well, disks being the most common bottleneck. Our 4 x 500GB (2TB) gives us about 1.5TB useable. It’s actually a little less due to HDD manufacturers counting differently to the rest of the IT industry, but hey, that’s marketing for you.

With your RAID 5 setup on your old PC and some cheap Hard Drives, you’ll have one of the cheapest large data stores around. You also get the benefit of knowing if one of those drives dies, your data is safe. The gotcha is you actually have to check the RAID occasionally to be sure it is healthy. You can do this by checking the system event log. There is a risk that as it all just keeps working in the event of a failure, it will continue unnoticed, until a second disk fails, resulting in the loss of everything. I am still looking for a freeware utility to alert in the event of a software raid pack error.

It should be noted that RAID only protects against one type of disaster – loss of a single Hard Disk Drive. If you lose multiple drives, accidentally delete files, get a virus etc etc, your data will be lost. As I tend to store non critical data, and use this space to backup my other PC’s, loss of this type is not a significant problem for me. This risk should be kept in mind and managed where possible. There is no cheap large scale backup for home use that I have found on the market. The ideal would be some type of 100 disk DVD Loader, however all the library’s out there don’t have integrated burners. If anyone finds a large cheap loader with a burner, I would love to know about it (another market opportunity for some Taiwanese company).

Now this machine is going to be on 24/7 and you should think a little green. Most CPU’s aren’t too bad for power when idling, even if they aren’t the latest models. The biggest energy chewer will be the HDD’s. If you go into Power under Control Panel and set the machine to shut down the Hard Drives after 15minutes of inactivity, you’ll save a heap of power, and extend the life of your drives significantly. Consumer drivers aren’t really designed to spin 24/7. Stopping them spinning also reduces noise. Overall it will probably chew about 30-50W at idle and up to 250W under load, depending on the CPU you are running. It’s not practical to enable Standby, as there is no remote wake up command.