Backups ... Just Do It!

Backups ... Just Do It!

Backups ... Just Do It!

If you are unfamiliar with the terms RAID 1 and RAID 0, don't get too concerned just yet. They will be explained below. For now, just know that RAID 1 is drive mirroring where all data is written to a pair of drives such there are always two copies of every file.

I have used a number of backup schemes over the years. In the not-too-distant past, most of them relied on some form of redundant, automatic backup such as RAID 1 (mirroring) or RAID 0 + 1 (striping plus mirroring). However, a while back, I decided that was not good enough and moved to a scheme of using software - in my case, CrashPlan (Plus) - to back up my desktop, my laptop, and my wife's laptop to a separate PC in my house just for that purpose. Additionally, I have CrashPlan back up the same machines (and my backup machine) off-site to their servers (called CrashPlan Central). The bottom line is that it's almost a requirement to have some form of active hard drive backup, and the good news is that it's not that hard.


Do Not Trust Your Data to Any Hard Drive Manufacturer
I don't receive any compensation for the hardware or software products mentioned. I may not have even used some of the products mentioned. Those are included only as illustrative examples where needed.

The following is my personal experiences with computer storage over the last 30 years or so. Sadly, the necessity of having some backup scheme for my hard disks to guard against hard drive failure increased dramatically for me in the early 2000s. Originally, I would set up a RAID 1 array fully expecting to never replace either disk. The rare occasion when I did have to replace one justified the cost in my mind. I could be quite smug when I slapped in the new replacement disk, chose to rebuild the array, and lost nothing (save the time to rebuild and the cost of the new disk). That fact was that in the early days, the majority of RAID 1 arrays that I set up would go through their entire life with no failures. I know of a couple of systems in which I included a RAID 1 array that were 4-5 years old and still running on the original drives.

Unfortunately, that changed. By 2005 or so, I set up a RAID 1 array fully expecting it to be necessary. I'm not sure what happened with the hard disk manufacturer's, but reliability just took a dive. I suppose it could be me; I'm using twice as many hard drives in the systems I build, so it follows I would see twice as many failures. I believe, however, that's not entirely the cause. As I mentioned above, my early RAID 1 arrays, for the most part, ran their entire life without any replacements. In the last few years, more than half of the RAID 1 arrays I've set up have required at least one disk replacement.

This trend doesn't seem to be limited to a particular manufacturer either. I typically use Seagate or Western Digital (WD), but I have occasionally used Samsung and IBM (now Hitachi) drives as well. I don't use the latter two often enough to include those two in the following statement: In my opinion, no hard drives can be trusted. It seems to be Seagate's opinion, too, as they first reduced their warranty from five years to three, and now that warranty is down to two years on standard models. WD has three years on most drives with their premium models at five. What that should mean to anyone with a drive older than three to five years is that it should be expected to fail at any time.

For a long, long time I was an utter Seagate fanboi; it's the only brand of hard drive I would use in builds. When I first started using RAID 1 arrays, WD had a couple of embarrassing manufacturing issues, which they initially denied and eventually acknowledged. I myself wouldn't touch the things. In 2006 or 2007, I built myself a new desktop PC with a RAID 1 array using a pair of Seagate 120GB drives. One of the drives failed fairly quickly. Because I had used so many of their drives over the years with no failures, I figured it was just my time to get a bad one and replaced the drive. Then the other drive failed within a year. Both were covered by the (then five-year) warranty, but that's not the point. Had I not used a RAID 1 (mirroring) setup, data would have been lost. I didn't feel I could trust that particular model and tried my luck with a larger 250GB Seagate. One of those failed just under two years later. Like the others, it was still in warranty, but this was not what I expected from Seagate. RAID 1 was saving the day again and again, but I really wished it wouldn't have had to.

I had a similar experience in a desktop PC I built for my daughter to take to college. One of the 250 GB Seagate drives (in RAID 1) failed after less than two years. During this same time, the company I worked for bought three WD250 GB external USB drives for a few of us working offsite to use for backing up our laptops. All three died; mine lasted the longest - almost two years. In my opinion, all hard drives should be regarded with the belief that they will fail before they are out of warranty, so having a backup is absolutely required. Sometime during all of this, I decided to use RAID 0 + 1 in my gaming desktop. That required I add another pair of Seagate 250GB drives in order to effectively get 500 GB of space. Of the four drives in use, I had three failures including one where the replacement for a failed drive also failed. I switched to using Western Digital Black drives, which had (and still have) a five-year warranty.

I now use SSDs (solid-state drives) as my main boot and file drives. Their cost is low enough to be viable replacements for HDDs (standard hard disk drives), but still high enough that doubling the number of drives needed (as used in RAID 1 mirroring) is prohibitive. I also have a pair of WD Scorpio Black drives, which have a five-year warranty. Still, I don't have enough faith in any of them not to have a complete backup elsewhere. That is where CrashPlan comes in. In the rest of this article, I'll go through the ways I have (and you can) backup your hard disks.

Approaches I Have Used for Hard Drive Backup

Hopefully, I made a convincing case above that hard drive backup is absolutely essential. It was always a good idea, but with what I'm seeing as the current state of hard drive reliability, it's no longer an option. There are lots of effective ways of maintaining a backup. In the links below, I will go into a number of them that I have used.


Craig Prall

RAID 1 - Disk Mirroring

RAID 1 - Disk Mirroring

RAID 1, for those who don't know, uses a pair of disks whereby anything written to one disk is also written (i.e., mirrored) in parallel to the other. I used RAID 1 in desktop builds - both for gaming and business use - for over a decade with very good results. On RAID 1's minus side, half the disk space is lost since, for example, two 500GB drives appear as only one 500GB drive. Also, the writes don't occur completely in parallel and the two drives require synchronization to ensure that everything was faithfully copied to both drives. This slows the write speed of a RAID 1 array down to being a bit slower than writing to a single disk.

If you decided to go with a RAID array, consider purchasing the hard disks from two different sources. (Make sure the model number is the same, however.) Doing so decreases the likelihood of getting two faulty drives from the same "bad" batch.

On the plus side, is that in the event either one of the two disks fails, no data is lost. That's the big plus that saved me from data loss a number of times. Also, when reading files from disk, the RAID 1 array can read alternate blocks of a file from both drives simultaneously to nearly double the read speed. (This is similar to how RAID 0 striping works all the time. See below.) In a gaming system, significantly more time is spent reading from the disks than writing, so RAID 0 gave me a noticeable increase in speed when booting, loading games initially, switching maps/level, etc. When and if a disk fails, that boost in read performance (and the security on having a constant backup) is lost until a replacement is installed and the files are copied from the remaining drive to the new drive. (This is called rebuilding the RAID array, and while this is in progress, it's a very good idea not to use the system. It's going to be really slow anyway, so just give it the necessary time to finish.)

A RAID 1 array can be built as a "hardware" or "software" array. Hardware arrays are those that use a motherboard with a chipset that supports RAID. Most current motherboards, even lower-end ones, support some form of hardware RAID. Sometimes, it's only RAID 0 and 1. (More on RAID 0 in a bit.) Optionally, we can buy a RAID controller add-in card to create a hardware RAID 1 array. The higher-priced add-in cards are often faster than the RAID solutions found on motherboards, so it's a good alternative. Windows 7, Windows 8 and Windows 8.1 also support creating RAID arrays though the operating system software. I personally have not used a software RAID, and it wasn't clear from what I read if a software RAID can be used as the boot drive. I know a hardware RAID array can be the boot drive as that's pretty much the only way I have done it.

The second thing we need is a matched pair of disks. Technically, the disks don't have to be the same, but it's more efficient if they are. If a 1.5 TB drive is paired with a 2 TB drive, the resulting RAID 1 array will be 1.5 TB with the extra 500GB of space on the 2 TB drive wasted. If the two drives have significantly different write times, the effective write time is that of the slower drive (plus the overhead for synchronization). This would waste the potential speed of the (presumably more expensive) faster drive. Typically, I just get a matched pair of drives from the outset. That said, it's a good idea to get each drive from a different source. That way, if the vendor happened to be the unlucky recipient of a bad batch of drives, the system we build with them won't inherit all the same bad luck.

When installing the disks, it is best to leave airspace between the drives if possible. If the drive cage has a fan that moves air across the drives, that's even better. The power and data cables should be kept out of the way as much as possible. The picture below shows a pair of drives with a fan in front of them that draws cool air in from the front of the case. (The case is an Antec Nine Hundred.)

RAID 1 drives installed - image 1 0f 1 thumb

Setting up a hardware RAID array varies by motherboard, but the basics tend to be the same. Rather than write my own instructions here, I'll just point to one of the many references I found on the subject at PCWorld. Really, the main trick for setting up a hardware RAID in Windows is having the correct drivers onhand (on a USB stick, CD or DVD) during the Windows installation process. For an example of setting up a software RAID array, this article at Overclockers seems to be pretty complete. It's not something I have done, myself.

Although RAID 1 has saved my hide on a number of occasions, it's only part of a good backup scheme. For example, RAID 1 won't protect against accidental deletion of (irreplaceable) files. Nor does it protect against viruses. Therefore, it's still a good idea to perform a periodic backup of critical files to CD-Rs, DVD-Rs, an external disk, a spare internal disk, another computer, a network file server or a backup server on the Internet. This need not be a backup of everything on the disk (like RAID 1 does), but it should be those documents, audio files, pictures, and movies that you yourself have created and would not be able to replace. Several of the following sections talk about these alternatives.

Craig Prall

RAID 0 (Striping) Plus a Backup Drive

RAID 0 (Striping) Plus a Backup Drive

RAID 0 is the preferred hard disk solution for the fastest disk access possible and is very popular in gaming desktops (and some laptops). With RAID 0, files over a certain size - called the stripe size - are broken into "stripes" and the stripes are written to alternating disks. Generally, the number of hard disks in a RAID 0 array is two, but it can be more. If it is two, the first stripe-sized block of a file is written to one disk and the second is written to the second disk. The third stripe is written to the first disk, and so on. The stripes can be fed to the two disks in parallel nearly doubling both their read and write speed.

The first and major plus as mentioned is the greatly improved read and write speed. Also on the plus side, unlike with RAID 1, no disk space is lost in RAID 0. Two 2 TB drives would appear as a single 4 TB drive. On RAID 0's minus side, if either disk fails, all files over the stripe size are effectively lost since half of a file (assuming a two disk raid) is little better than no file. One way to resolve this problem is to have a third non-RAID disk that matches the size of the RAID 0 array (or larger). This drive will be used to contain one or more backups of the RAID 0 array. Some form of backup software such as Acronis True Image Home 2015 would then be installed and configured to perform backups on a regular - typically nightly - basis.

Acronis True Image has a feature named Acronis Nonstop Backup that automatically creates incremental backups of files and folders (or even partitions) every five minutes allowing users to roll back their system, files, and folders to a point in time in the past. While this sounds good, it may well defeat the purpose of a RAID 0 drive or at least reduce it. There would need to be some sort of synchronization or checking of the backup in real-time that would slow down the RAID array. Since it's incremental, it may not be much in reality. Acronis also has a powerful scheduler to make the backup at some time when you are not on the computer. This would not interfere with the computer while it's in use, but the risk is losing any data created since the last backup is run.

Acronis is not the only backup program to consider. Macrium (Reflect) and Paragon (Hard Disk Manager) are also quite popular. At various times, various releases of one or the other products have had issues. I currently use CrashPlan, which could be used for a backup scheme like the one above, but it really meant for backing up one computer's files to another (which could be a local PC, a friend's PC over the Internet, or even CrashPlan's own servers).

From a build point of view, for hardware RAID, we need a motherboard that supports RAID or we need to buy a RAID controller add-in card as before. For software RAID, we need an operating system that supports that feature (also as for RAID 1). The second thing we need is a matched pair of disks and a third disk that's approximately the size of the RAID array. The disks to be used in the array don't have to be the same, but the array is generally more stable if they are. If a 1 TB drive is paired with a  1.5 TB drive, the resulting RAID 0 array will be twice the size of the 1 TB drive or  2 TB with the 500 GB of extra space on the 1.5 TB drive wasted. If the two drives have significantly different write times, the effective write time is that of the slower drive. This would waste the potential speed of the (presumably more expensive) faster drive.

As for the backup drive, it may either be external or internal. If using an external, try to use USB 3.0 (or above) or an eSATA connection. Disks are installed using the same rules of thumb as with RAID 1. Leave airspace between the drives if possible and use or install fans for the drive cage, if possible. The power and data cables should be kept out of the way if possible. The picture below shows a pair of drives used for RAID 0 and a third, larger drive used for backup. (The case is an Antec Nine Hundred.)

RAID 0 drives and backup drive installed - image 1 0f 1 thumb

As before, rather than giving instructions for setting up a RAID system here, please refer to the list of guides found on the Internet such as those listed in the first section.

Craig Prall

RAID 10 (1+0) - Mirroring and Striping

RAID 10 (1+0) - Mirroring and Striping

In the RAID 1 (mirroring) discussion, it was mentioned that the write speed of such an array is slightly slower than writing to a single disk. On the other hand, RAID 0 yields a write speed that is nearly double that of a single disk (because it involves writing simultaneously to two disks). What would be ideal is to have the continuous backup that RAID 1 affords while somehow keeping the write speed of a RAID 0 array.

Somewhere along the line, somebody realized that if we had a pair of identical RAID 1 arrays, we could write to those using the striping technique of RAID 0 and let the RAID array perform the mirroring as time permits. This is the essence of the solution provided by RAID 1+0, which is also known as RAID 10. It is almost always written as RAID 10. Essentially a pair of RAID 1 arrays is used in a RAID 0 array, which makes it a stripe of mirrors. This gives us (nearly) the speed of a RAID 0 array with the data redundancy and security of a RAID 1 array. A RAID 10 array requires (a minimum of) four hard disks and like a RAID 1 array, only half the total disk space is accessible. However, the write speed to that array is essentially the same as that of a RAID 0 array. Even better, the read speed is double that of a RAID 0 array. For a gaming system, which tends to do a lot of burst reading (when switching levels or areas for example), this can make quite a difference.

The first and major plus of RAID 10, which was already mentioned, is the combination of the advantages of RAID 0 speed and RAID 1 security. If any single disk fails, no data is lost. The failed disk can be replaced and the RAID rebuilt from the remaining disks. In theory, two drives can fail so long as they do not both contain the same mirrored stripe. In general, we are trying to guard against single disk failure while maintaining RAID 0 (or better) performance.

From a build point of view, we need the motherboard support for RAID 10 or an add-in card that supports RAID 10. This time, however, we need a set of four matched disks. The same caveats about disk size and speed apply as in RAID 0 and RAID 1. One additional factor in the mix is that you need a case that holds four internal hard disks and still has decent airflow. Leave airspace between the drives if possible and use or install fans for the drive cage, if possible. The power and data cables should be kept out of the way. The picture below shows four drives in RAID 10.

RAID 10 drives installed - image 1 0f 1 thumb

Note that I did not talk about software RAID in the last paragraph. As far as I can tell, Windows 7 does not support RAID 10 directly, so a hardware RAID is the only option. Windows 8 and Windows 8.1 support a newer concept called Storage Spaces. Storage Spaces don't provide an exact equivalent to RAID 10, although they do have support configurations with similar goals. However, that's not something I have used (and a bit beyond the scope I was trying to cover here).

Also, note that I only discussed RAID 10 and not RAID 01 (also known as RAID 0+1). While RAID 10 is known as a stripe of mirrors, RAID 01 is known as a mirror of stripes. Most motherboards only support RAID 10 and not RAID 01. RAID 10 and RAID 01 require the same resources and yield the same performance benefits. The devil in the details is more involved than I want to get into here, but suffice it to say that RAID 10 is inherently more stable than RAID 01 when a drive failure occurs. A good article on why this is so can be found at The Geek Stuff.

As before, rather than giving instructions for setting up a RAID system here, please refer to the list of guides found on the Internet such as those listed in the first section.

Craig Prall

My Current Backup Approach

My Current Backup Approach

Although I have used RAID 1 and RAID 10 in the past - even occasionally RAID 0, I am not currently using any of those approaches to maintain my backups. In fact, I don't even have a RAID setup in my current system even though it's perfectly capable of supporting one. I even have a pair of matched hard drives.

In my last several builds, I have switched from rotating hard disk drives (HDDs), to a combination of Solid State Drives (SSDs) and HDDs, and finally to just SSDs. In my current build, I'm using a pair of 2 TB NVMe SSDs that install directly on the motherboard. (The speeds on the two SSDs differ.) Even though the prices have come down dramatically, the cost of SSDs makes setting up a RAID a bit prohibitive. Instead, I have switched my desktop, a laptop, and a file server to using offsite backup over the Internet. I have used a couple of services to do that.

Nasty Note: 22 August 2017

CrashPlan Kisses Consumers Goodbye image 1 0f 1 thumb

CrashPlan has decided they no longer wish to service consumers. After five years of giving them more than I thought they were worth ... I am not pleased.

I very recently switched from CrashPlan Home (Family Plan) to IDrive. I didn't switch because I had any issues with CrashPlan. I switched because CrashPlan decided they no longer want to bother with consumers. Like CrashPlan, IDrive lets me pick the files and folders I want to back up. IDrive doesn't have unlimited storage as CrashPlan did. However, their $70/year plan - which is less than half what I was paying CrashPlan per year - includes up to 2TB of storage for an unlimited number of devices (where the list of devices includes iPhones and Android phones and tablets in addition to computers). That's considerably more space than I actually used with CrashPlan.

CrashPlan Home was available as both a free version and a subscription version. The free version of the software supported backing up to an external disk, another computer in your house or even a friend's computer. The free version did not support backing up to the Internet. With a yearly subscription on the CrashPlan for Home Family Plan, I could back up any 10 devices to both a local and CrashPlan's Internet-based servers (in parallel). None of the other online backup services (including IDrive) seems to support that parallel local/online backup. IDrive's (significantly more expensive) business plans offer Hybrid backup, which does support both.

IDrive allows anyone to create a free, small (5GB limit) online backup using their software. It also seems to support creating a local backup to an external or network drive (I haven't tried it.) IDrive does let me use my own encryption key and has 256-bit AES encryption. I can get at my backed-up files from anywhere (so long as I have the key with me). IDrive keeps files backed up to it forever or until cleaned up (i.e., manually deleted, but it has tools to help with that) and keeps the last 10 updates to any file. I'm not sure how IDrive tallies up disk usage, so it remains to be seen if this will become an issue or not.

Since the very first backup(s) can take excruciatingly long over slower Internet connections if the amount of data is moderately large, so IDrive offers IDrive Express that allows the user to use the IDrive software to backup to an external drive (which IDrive supplies). After that backup completes, the user ships the drive to IDrive (IDrive pays for the shipping), and they load the data onto their servers. This is known as a Seeded Backup. From that point, the users only need to do incremental backups. IDrive customers can use this service to bulk upload their data once per year for free. It's $60 for additional uploads. IDrive also supports loading backed-up data onto an external drive from their servers and shipping that to you in the event of a major loss of data. That costs $100 for each request. (CrashPlan originally offered the same services for home users but discontinued them.)

Craig Prall