Backup your digital photos part 2, continued.
RAID for data backup
To enjoy the protection and potential speed benefits of RAID, you’ll need a RAID controller; these employ the components to build and support the RAID array, along with the necessary sockets to connect the actual hard disks. RAID controllers can actually be found on most modern PC motherboards, and are also available on optional plug-in expansion cards.
RAID facilities are also offered by many external storage devices, such as Network Attached Storage, or NAS appliances. These sometimes come with hard disks already fitted and ready to go, or as empty ‘barebones’ cases just waiting for you to fit your own choice of hard disks.
To make the most of RAID there are several rules to be aware of. The first and most important rule is that RAID controllers always treat each disk in the array as if they were exactly the same size as your smallest disk – so there’s no point using models with different capacities as it will be wasted. To avoid wastage, try and build your RAID array using disks with identical capacities, such as having two or three 500GB disks.
The second rule is that not all RAID controllers are created equal, and as with most technologies you get what you pay for. RAID requires a certain amount of processing power to work, and cheaper controllers simply use your main processor to do the legwork; these are known as software RAID controllers. While most modern computers have plenty of spare horsepower, RAID controllers which perform their calculations using dedicated processors normally deliver much faster performance – this particularly applies to the more sophisticated versions of RAID we’ll discuss in a moment. So if fast performance is crucial, go for a hardware RAID controller.
The third rule is that several different types of RAID can be used to build your storage array, each denoted by a number and each with its pros and cons. We’ll explain the main three here.
RAID 0 splits your data across two or more hard disks and reads and writes from them simultaneously. You’ll therefore enjoy improved speed and the full capacity of all the disks in the array – so a pair of 500GB disks would give you 1000GB of storage along with the benefit of running quicker than a single disk. However should one disk fail, the entire array breaks and you lose all your data. Clearly RAID 0 with its reduced reliability is far from what we’re looking for here. Instead it’s best used for applications where long-term data reliability is not important but extra speed is useful, such as virtual memory or Photoshop scratch disks.
Next up is RAID 1 which employs a pair of hard disks and simply keeps a duplicate set of data on each. The advantage is should one disk fail, the other is ready to take over immediately. The downside is you effectively only have one disk’s worth of capacity. So you may have bought, say, two 500GB disks, but you still only have 500GB of space to play with. This 50% loss in capacity can be hard to live with.
RAID 5 is a more palatable solution. It uses three or more disks and cleverly distributes backup data over all of them. The result is protection against disk failure, but with a loss of only one disk’s worth of capacity. So if you had three 500GB disks, you’d have 1000GB of storage, thereby losing 33%. If you had four 500GB disks, you’d still only lose one disk’s worth, giving you 1500GB to play with and an effective loss of only 25%. Clearly it’s still a loss in capacity, but with a 33% loss at worst, it’s much more acceptable than RAID 1. And when a disk fails, you just swap it for a new one and the controller rebuilds the array, while still allowing you to access all your data.
RAID 0 and RAID 1 are very simple technologies and are offered by virtually every RAID controller as standard; most also implement them using hardware acceleration. RAID 5 is becoming more widespread, but the distribution of data and in particular the rebuilding process following a failure are very demanding. Most cheaper RAID 5 controllers, especially those featured on motherboards, generally employ a software solution. They may read your data at a similar speed to your normal hard disk, but writing data could be at least five times slower.
If you want fast RAID 5 performance, you’ll need a hardware-based RAID 5 controller. The Promise SuperTrak EX8350 controller card which fits into a PCI Express slot is a great choice, but is inevitably more expensive than a software-based controller. I compared this card against the software-based Promise FastTrak TX4310 card for the March 2007 edition of Personal Computer World magazine in the UK and the difference in performance was quite remarkable – we'll link to the article once it's posted. Also don't assume that a controller which has hardware acceleration for RAID 0 and 1 also has hardware acceleration for RAID 5; cheaper solutions may sport hardware acceleration for the former but not the latter.
Alternatively for an easy life, you could get yourself a NAS appliance, an external box which contains one or more hard disks and simply connects to your network, presenting its storage to all your computers. As you’d expect, NAS appliances come in all shapes and sizes, some with disks already fitted and some waiting for you to fit your own. Most tend to offer several different RAID options and again it’s best to steer clear of RAID 0 and opt instead for RAID 1 or ideally RAID 5, especially if the latter's also hardware-accelerated. Check out products like the Buffalo TeraStation Pro 1TB or Infrant Technologies ReadyNAS NV 1.0TB.
Ultimately RAID is a powerful technology for protecting against disk failure and is recommended for any photographer who’s serious about their collection. Again RAID 5 solutions seem to be the most flexible and efficient, but for the ultimate performance look out for versions which are hardware-accelerated. Finally remember that while RAID will protect you against disk failure, it won’t cover you for fire, flood or theft. It’s no substitute for backing up regularly onto an external device or media and crucially keeping that backup in a different location; see Backup your digital photos Part 1 for more details.
Support this site by checking prices below or shopping via our affiliate stores