Manage the Digital Workflow

The main external-drive choices are a single drive, a network-attached drive, or a RAID drive-set (which could also be network-attached). Internal drives aren't good choices because they're not sufficiently independent of the computer being backed up. They're gone if the computer is stolen, and they share the same drive controller, so a controller failure could destroy the data on all the internal drives.

FireWire, USB, and eSATA External Drives

External hard drives are great choices for backup. They're available in sizes from about 100GB to 1TB, and some of them are even entirely powered from the USB or FireWire cable, which makes connecting them and transporting them especially convenient. (Laptop and some desktop USB ports often don't put out enough power for an external drive, but some drives come with a split cable that allow you to draw power from two USB ports.)

Nearly all computers have USB 2; those that have FireWire (IEEE 1394) usually have FireWire 400, although FireWire 800 is now available on newer models. For external drives, FireWire 400 is noticeably faster than USB 2, and FireWire 800 is much faster than FireWire 400.

You'll know if you have a computer or drive with FireWire 800, as opposed to FireWire 400, because the connectors are different. (USB 1 and USB2 use the same connector). You can connect Firewire 800 disks to FW 400 ports on the computer (of course you then only get FW 400 speed).

eSATA ("external Serial Advanced Technology Attachment"), just starting to be available, is several times faster than FireWire 800, because it takes the bus commonly used inside the computer (SATA) and extends it with an external cable. Since most external drives are SATA, this bypasses the conversion to USB or FireWire and then back to SATA. (Currently, eSATA cables don't supply power.) With a $40 eSATA PC-Card, adding eSATA to a laptop is even easier than it is to a desktop or server.

Note by Uwe: We use now only drives that connect via Firewire 400 or 800 to have a standard way to connect all our many external disks. USB 2.0 is too slow for our taste for large volume drives. USB 2.0 is ok with portable bus-powered drives though. In general Firewire works often better on Macs and USB 2.0 on PCs.

Remember, though, that for backup the most important property of an external drive is that you can easily separate it from the computer, not how fast its connection is.

If the drive is going to be running all the time, put it out of sight, such as on a shelf under your desk, or on a bookshelf with some books or a family photo in front of it. Figure that a thief won't know your drive is even there and, even if he or she does, nobody wants to heist a $200 drive when there are computers, CDs, and jewelry to take instead, all much easier to fence.

We won't recommend specific brands or models of external hard drives here since they change so often. Recently, however, Uwe was impressed with the Western Digital My Passport Studio, which has both USB 2 and FireWire 400 connectors.

Marc uses large external FireWire 400 drives that are always connected and running, stored on a shelf under his desk. Two of them are used for alternating nightly images of his main drive. (They alternate in case he has a failure during backup.) He also uses a couple of 100GB - 120GB drives to save his current work that he alternates between his office and a friend's house, and he exchanges them about once a week. He carries them back and forth in foam-lined Pelican cases, which he recently wrote about. A third set of external drives back up his image archive from time-to-time and are stored in a fireproof safe, to supplement his optical-disc and online archival backups. (The drives are more vulnerable, since they're on-premises, but faster to get to).

Uwe is now mostly using 500GB - 1TB external drives (single disk, Firewire 400) from Western Digital or Seagate. Western Digital right now offers the most options for external disks (we use use single disk and RAID-1 units). The benefit using external disks from the disk manufacturers is that you know which disks are used. Also the manufacturers know very well how to handle noise and heat of their own disks.

Marc also likes the idea of getting small drives for offsite storage in bright colors, so that he'll notice right away if he left one in office by mistake.

Digital wallets for in-the-field backup are a special case of external drives and were briefly covered in Part 2 of this series, Protecting Against Threats.

Network Attached Storage (NAS)

This is a fancy name for a drive connected to your local network instead of directly to a computer, although in reality the drive has its own computer to handle the connection and storage management within the drive. NAS has been around for years, but only recently have devices been created for home and small-office users that are cheap and easy to set up.

The main advantage of an NAS device is that it can be shared among computers. The main disadvantages are that networks are much slower than even USB 2 and less reliable—brief networks hiccups and longer outages are common, even one of which can disrupt the backup. You want backups to just work, without your having to monitor them and perhaps get them unstuck and restarted.

The best-known new NAS device is Apple's new Time Capsule, which is a 500GB or 1TB drive combined with a wireless base station. It allows you to back up wirelessly, with or without Time Machine (OS X Leopard's built-in backup software). This means it can be farther from your computers than a directly-connected drive, such as in a closet, on a high bookshelf, or in another room. (Friendly neighbors might even be able to put their Time Capsules in each others' houses, although we haven't tested that arrangement.)

NAS drives contain small computers for administration and storage management. You can also add a desktop or server computer to the network to act as an NAS. But such a computer is much more expensive than a self-contained NAS device, requires much more power and space, and introduces another machine to be set up, booted, maintained, and possibly even backed up. A little box you can just attach and forget about is much more convenient.

To find NAS boxes, go to Amazon and search for "nas drive", "nas disk", "ethernet drive", or "ethernet disk". You'll find drives with network attachments, drive cases with network attachments (you have to add the drive), cases with WiFi and with and without drives, and even cheap ($85) devices to which you can attach a USB drive. Some examples, none of which we've tried or are recommending, are: Linksys Storage Link, Iomega Home Network Storage, Simple Tech SimpleShare, and Neatgear ReadyNAS NV+.

Some things to watch out for when you buy an NAS device:

• Other than Time Capsule (which also works with Windows), even the easy-to-set-up devices can be complicated to set up. If you're used to setting up networks, downloading drivers, and dealing with strange preference dialogs, you'll be ok.
• Some backup software organizes the entire backup as a series of giant files (>4GB), and some NAS drives don't allow files greater than 4GB, so you may have a problem.
• According to reviews on Amazon, some of these devices are very slow.
• Some devices won't work with a Mac and those that do might not work with OS X 10.5 (Leopard).
• Even if they will work with a Mac, they won't work with Time Machine because it doesn't support networked drives (other than Time Capsule).
• Apple's AirPort wireless base station has a USB port for an external drive, so you might be able to use that instead of getting a Time Capsule.

Marc tried a Time Capsule when it first because available, but had so many operational problems with it that he sent it back. Since then the firmware in the Time Capsule and the version of OS X that supports it (10.5) have been updated several times, so his problems may have been solved.

RAID

RAID stands for Redundant Array of Inexpensive Disks. The advantage of RAID is that it's more reliable than a single disk of the same size. The disadvantages are that it costs extra for the same amount of storage (at least one extra disk and some fancy electronics) and that the disks aren't nearly independent enough because they share the same power source, controller, driver, and connecting cable.

Sometimes the "I" in RAID is said to stand for "independent," but Marc strongly dislikes that substitution.

In the 1980s there were two categories of hard drives: expensive and reliable, and inexpensive and less reliable. RAID was a clever scheme for using an array of inexpensive disks to match, or even exceed, the reliability of expensive disks, with much lower costs. Today, even inexpensive disks are orders-of-magnitude more reliable than the expensive disks of the 1980s, so the advantages of RAID are much less compelling. The chief benefit of RAID today is not to make disks reliable enough for critical work, since even a single disk is more than good enough, but to reduce downtime. RAID allows the system to keep running even if a disk fails. You can replace the failed disk and the RAID system will automatically bring up-to-date in the background while you can go on with your work. This is advantageous because disks are cheap and missed deadlines are expensive. So RAID is about increased productivity, not about preventing failure.

There are various RAID arrangements, the two most popular of which are RAID-1 (mirroring) and RAID-5 (striping with parity). The idea is that when a disk fails it can be removed from the running system and replaced without the system going down or any data being lost. When you replace the defective disk, the RAID system automatically restores the data that was on it. Unfortunately, the restore often takes hours, during which time you are vulnerable to a total data loss if there's a second failure.

So-called RAID-0, also called striping, isn't really RAID at all because there's no redundancy. The loss of either disk destroys all the data on both disks. RAID-0 is for performance, not for reliability, which is actually reduced. For Uwe RAID-0 is a clear no-no for any backup solution. RAID-0 can be perfect for scratch disks using eSATA if the highest speed is needed and data security is no real issue.

Drobo is an especially clever, easy-to-manage, and nicely-packaged implementation of RAID. It can take from two to four drives of different sizes. It uses variations on RAID-1 and RAID-5 and sophisticated volume management to partition the drives automatically. You can increase the size automatically by just adding a drive or replacing a drive with a bigger one, and Drobo wil adapt to the increased space. But, as smart as it is, Drobo is still RAID. Uwe likes Drobo because it can grow in storage size over time and is a good value for the money. There was a new Drobe announced that supports now Firewire 800 and USB 2.0.

RAID is ok if you are using it to replace what otherwise would be a single drive (primary or backup), but it doesn't substitute for backup. Moreover, if your budget is limited, the extra money you spend on RAID might mean that you have less to spend on backup—a poor tradeoff if your backup is less than adequate.

To see why RAID doesn't diminish the need for backup, we can do a quick threat analysis, as we do for all other backup methods:

1. User Error: No help; with RAID there's still just one logical copy of the data.
2. Computer Failure: RAID protects only against disk failure. That's probably the most common hardware failure, but as RAID doesn't protect against other hardware failures, such as the RAID controller itself and the rest of the computer, nor against software failures, the need for backup hasn't changed.
3. Surge: No help, as all RAID hardware has to be plugged in.
4. Disappearance: No different than a single disk. An external RAID cabinet can be hidden below the desk, but so can a single-disk cabinet.
5. Office Destruction: No help.
6. Regional Disaster: No help.

Anyone who thinks RAID provides sufficient protection is focused too narrowly on a single kind of failure, the disk itself, and is ignoring the other threats. Still, if you have the money, a RAID device is more reliable than a single disk. (Not RAID-0, however, which is less reliable).

It's true that your system is somewhat at risk during the rebuild that follows the replacement of a failed disk, but, if your backup design is right, you're not depending only on the RAID system. If you find yourself nervous during a rebuild, perhaps your other backups are inadequate. Still, it's a good idea to immediately back up your current work - what you've done since the last backup - as soon as a RAID failure occurs, just in case.

Uwe is using a Western Digital RAID-1 (2TB means 2x 1TB) disk as an always-on extension to his main Mac workstation. This disk will be also have offline copies (single disks).

Western Digital My Book Studio 2TB (2x 1TB) in RAID-1 setting
Ports: 2x FW 800, USB 2.0 and eSATA

Optical Discs

Optical discs are in theory more stable than hard drives, even unconnected ones, because they don't rely on magnetic recording and they have no moving parts. Discs you write on a computer use dyes, however, so they're not indestructible (see below).

Optical discs are much cheaper than hard drives, which means that you can afford to make several backup copies and store them in very safe places. The disadvantages of optical discs relative to hard drives are that they're much slower to write to (burn), and their capacity is much lower, ranging from only about 700MB for CDs to 4.7GB for single-sided DVDs to 25GB for single-sided Blu-ray discs.

Single-layer DVDs are at the sweet spot of the capacity-cost curve, at about 6 cents per gigabyte, compared to about 30 cents per gigabyte for a hard drive (or a little less if you shop carefully).

Double-layer DVDs hold almost twice as much as single-layer, but cost at least 5 times as much and can't be burned on all burners. Also, the best archival discs (see below) aren't usually available as double-layered. As of this writing, blank Blu-ray discs cost about 40 cents per gigabyte, but that's bound to come way down over the next several years. CDs are cheaper than DVDs but hold much less, so aren't usually practical for image backup.

There are two DVD formats in common use for backing up data: DVD-R and DVD+R, and nearly all DVD burners can burn both. You can search around the web and find various arguments in favor of one or the other, but we haven't seen anything definitive, although last time Marc did his research he decided DVD+R was more reliable. (That was web research, not scientific research.)

The manufacturing quality and the materials used make a huge difference in reliability. Most major manufacturers are now offering archival quality discs, sometimes with a silver or gold coating. Last time Marc bought discs, he got Taiyo Yuden Shiny Silver Thermal 8X DVD+R Media, but he's sure that there are many other good choices for discs and sellers. What's never a good choice are bargain discs labeled with strange brand names and, at 30 cents a disc for the good ones, there's no point in using cheapies for image backup.

How you protect the discs once you burn them is also important. Write only on the hubs, store them with the written surface suspended in air, keep them in the dark, and keep them cool. More information is in the US National Institute of Standards and Technology (NIST) publication Care and Handling of CDs and DVDs.

However well you care for optical discs, they won't last forever. You should plan to re-copy them every few years, and then verify that the images on the copy are still good with a tool like Marc's ImageVerifier.

Backup Overview

Manage the Digital Workflow