Don’t bet the store on unsafe storage

Apple and Dell don’t realise it yet, but computers aren’t very important any more. Not for printing companies anyway. Desktop computers became powerful enough to run Rips and multi-layer Photoshop files around the turn of the century. Since then Macintoshes have become cuter and PCs have shrunk a bit, but most of their extra horsepower has been absorbed by ever more ambitious operating systems and programs with extra features that most people never use. Bloatware.

No, the real issue facing printers and pre-print companies at present is storage space. Ace columnist Gee Ranasinha entertainingly outlines the storage challenges of home digital media systems on page 50, but business users have the same problems. Not everyone has an IT manager to worry about it for them, so for small businesses it’s a suck it and see approach. Or sometimes, prod it and see if it’s still alive.

All of us need more and more data storage to hold ever greater quantities of digital files, whether they’re digital cameras with ever higher resolutions, colour books, or ambitious multi-layer Photoshop files.

The value of the files to your on-going business means that you need to protect them against disaster, which might be the biblical variety (flood, hurricane, lightning strikes) or more man-made (fire, theft, sabotage or large bangs such as Buncefield in 2005).

Computers are replaceable. Stored digital data isn’t, unless you take steps to make it safe. You need to make plenty of duplicate copies, update them regularly, and spread them around widely, preferably in separate buildings, ideally in different cities. Secure off-site data storage is offered by some service companies, though it’s relatively pricey.

However, in-house storage and back-up is getting cheaper and easier. I’ve been buying new hard drives for about £100 for the past 15 years, for both Macs and PCs. When they fill up after a couple of years, in goes a £100 replacement, which will always be substantially larger for the same price. Originally my £100 would buy about 100 Megabytes. At present  it buys a 750 Gigabyte drive. I’ve just bought four of them, to replace a 300 GB and a 150 GB drive – my 10 mp digital camera files really eat up space. Next year my £100 will probably buy a Terabyte drive.

When Image Reports looked at storage technology a couple of years ago, the familiar magnetic hard disk technology seemed to be approaching its capacity limits. It looked as if the familiar 3.5 inch half-height disk drive, the standard size for desktop computers, would max out at around 250 GB. So a lot of effort was going on in labs worldwide to come up with a replacement for magnetic storage, often using exotic microscopic mechanical engines called MEMs.

But these haven’t arrived yet because traditional hard disk makers keep finding clever ways to smash the capacity barriers. Recently they’ve adapted an old idea called perpendicular recording technology. This packs data more efficiently, allowing higher capacities at lower costs than anyone thought possible three years ago. 1 Terabyte may be the final limit for 3.5 inch magnetic drives, but don’t bet on it.

Terabyte drives are impressive, but not enough for some image-intensive printing businesses that now need to keep several Terabytes on-line.

Another problem is that removable storage system capacities just aren’t keeping up with the need for digital storage and ever-increasing hard drive sizes, so it’s less practical to use them for back-up and long-term archiving than it once was. The new-generation Blu-ray and its HD DVD rival can’t manage more than 25 GB and 15 GB respectively per disc, even if anyone trusts them because of ludicrous format wars.

Iomega’s Rev cartridge drive, successor to the dear old 1 GB Jaz, has a maximum of just 70 GB per removable chunk. Tape drives, the slow but reliable staple for back-up, seem to have run out of steam too, and nobody likes them much anyway.

So today’s most convenient answer to the reliability and back-up dilemma seems to be to keep everything on duplicate sets of conventional hard drives, but to make provision for the fact that these can and do fail. The failsafe is RAID, the long-standing technique of linking a group of separate hard drives to make them behave like a single high capacity volume. With most configurations, if one of the drives fails you can replace it without losing any data, because the other drives hold enough information to restore the complete set.

RAID stands for ‘redundant array of inexpensive drives.’ The easiest but priciest way to get it is to buy a preconfigured box that’s already stuffed with disk drives. You plug it into your computer or network, which treats it as one giant disk. The best versions put standard drives onto removable caddy trays that can be ‘hot-swapped’, ie replaced without switching off the whole unit. Really mission-critical units have twin power supplies so if one blows the other keeps going, but you probably only need this if you’re running a bank or Tesco’s on-line shopping site.

The more cash-strapped among us can build our own RAIDs. A big advantage of the Macintosh OS X operating system is that its standard Disk Utility software lets you build and maintain your own RAID sets. You simply plug in a bunch of off-the-shelf drives, either internally or externally, and OS X does the rest. However, Windows users can join the RAID party easily enough by third party methods.

Starting with a pair of drives (which can be internal or external) you can opt for ‘Striped’ RAID 0 or ‘Mirrored’ RAID 1, and if you fit four or more you can additionally go for ‘Concatenated’ RAID 5.

RAID 0 is not a back-up system – it adds the capacity of two or more drives together so they run as one. Striping (reading or writing small chunks of data from alternate drives) runs faster than a single drive. This is handy for digital video houses and possibly high volume variable data printing. However, if one drive fails you lose the data on both for ever.

RAID 1 Mirroring keeps your data more secure because identical copies of your data are written simultaneously to pairs of drives. This slows them down and halves the capacity, but if one drive fails you can keep using the other until you’ve got time to pop in a replacement.

RAID 5 is the best choice if you can afford 3 or more drives. It splits the data across all the drives in the set, so you have access to lots of storage space, but it uses a mechanism called distributed parity which in effect keeps track of where it all goes. If one drive fails you replace it and the RAID system uses the parity bits to rebuild the missing data so nothing is lost. However, you can’t access any data until you’ve plugged in the new drive, and rebuilding can take hours if you’ve got hundreds of Gigabytes. Parity bits occupy significant space too – if you install 4 x 500 GB drives into a RAID 5 array you’ll have a total capacity of 2 Terabytes, but only 1.5 TB is available for actual storage because parity bits take up the rest.

Desktop Mac Pros can take up to four drives internally, so RAID 5 is possible without needing external boxes or expensive interface cards. So I’ve put three 750 GB drives on mine running as RAID 5, with a fourth as a normal system disk – Mac OS X won’t boot up from RAID, unfortunately.

There are other minor variations such as RAID 3,4 or 6, but 5 does the job for most people. You can even link multiple RAID sets in nested arrays, which can combine safe mirroring with speedy striping. Nested arrays have labels such as RAID 10 o+r 5+1.

Worth mentioning in passing is that solid state flash memory is moving beyond digital cameras to laptop computers, where it boots up twice as fast but uses less battery power than a spinning disk. A handful of laptops offer 32 GB ‘flash drives’ already, and you can buy aftermarket units that fit the 2.5 inch bays of existing laptops. They’re pricey at about £500, but expect capacities to rise and prices to plummet soonish.

The home front is starting to have a demand for mass storage of multimedia, as Gee points out on page 50. In response suppliers are developing idiot-proof cheap storage servers, called data vaults, for home use. Basically they’re self-contained RAIDs with a bit of built-in intelligence that are loaded from broadband-linked computers and will then deliver their media wares to other networked digital devices, such as TVs and music players. A typical example is the Buffalo TeraStation, costing £500 for a box of 4 x 250 GB drives making 1 Terabyte, or 750 GB as RAID 5.

Many of them support the super-fast Gigabit Ethernet for wired networks and some also act as wireless network nodes. Note that not all of them allow users to replace their own drives, so you have to send them away if they break.

Canny businesses can use these data vaults as cheap back-up servers. Being a skinflint I have one already, a networked £250 Maxtor Shared Storage II with twin mirrored 500 GB drives and software that incrementally backs up all my computers every night.

Being on a network means access is slower than with SCSI, FireWire or e-SATA direct connections to computers, but they’re fine for automatic overnight back-ups, especially as they can sit at the end of very long Ethernet cables or can even be accessed remotely (but slowly) over broadband links, unlike direct connections that are limited to 2 m cables or less.

Back to top