The pros and cons of RAID 1

Topic

New Reply

ISSUE 20.18 • 2023-05-01 HARDWARE By Will Fastie Revisiting an old friend, it’s time to update our thinking about RAID 1. In our all-too-brief time wo
[See the full post at: The pros and cons of RAID 1]

8 users thanked author for this post.

glnz, ibe98765, fisher75, teuhasn2, TechTango, AlexEiffel, DLivesInTexas, wdburt1

Reply | Quote

Replies

Very interesting post for me.  I have some stuff stored on a Western Digital MyBook Duo, which is set up for RAID 1.  I understood from the start that a RAID setup does not substitute for a backup, preferably both local and to an online service.  I did not understand until later that if the RAID enclosure controller fails, there might be a problem.  I can’t simply take one of the drives out of the Duo and put it in another enclosure, because the Duo encrypts the drives.  So I am backing up the Duo to another drive, daily.

More broadly, I think all that mission-critical hype has little relevance to me, and when the Duo is replaced it will be with a regular external hard drive.

1 user thanked author for this post.

Will Fastie

Reply | Quote

When I built my DIY NAS, I toyed with RAID 1, but settled on RAID 10 for the additional redundancy.  Both disks in a pair have to fail for the array to fail.  I started with 3TB Seagate NAS drives, and upgraded a couple of years ago to 4TB drives.  I’m considering upgrading to 6TB drives.

I use full format on all drives before I put them into service, and the upgrade is simple and straight forward.  Replace a drive, wait for the array to rebuild, replace another drive, wait for the array to rebuild, etc.  Once all four larger capacity drives have been installed, the RAID 10 array is automatically reconfigured by Intel RST to reflect the increased capacity.  All I had to do was format/swap drives.

Always create a fresh drive image before making system changes/Windows updates; you may need to start over!

We all have our own reasons for doing the things that we do with our systems; we don't need anyone's approval, and we don't all have to do the same things.

We were all once "Average Users".

Computer Specs

1 user thanked author for this post.

Will Fastie

Reply | Quote

Thank you for discussing this.

As with all complicated technology, understanding RAID 1’s nuances is key. Within a narrow range of requirements, and used in conjunction with other backup methods, RAID 1 can provide a great advantage to a small business. I had several small business clients for which RAID 1 provided nearly 100% uptime over years of operation during business hours, which is when those businesses needed that uptime. When one mirrored drive failed physically, which happened occasionally (although less and less as time went on) no one at the business ever knew about it. They continued to work as usual. I could then go in after hours, swap the drive, and let the mirror rebuild itself.

Mr. Langa’s two objections–that both drives could both fail simultaneously and that a software glitch or virus are immediately carried over–are true, but

1) as you say, the chances of both drives simultaneously (or near simultaneously) failing physically are tiny, and that very possibility requires an admin to replace a failed drive ASAP. To limit even that limited possibility, I replaced operating RAID 1 drives alternately every couple of years and kept the original drive as spares (that occasionally were useful.) Additionally, Mr. Langa’s catastrophic-failure objections, in this case both members of a RAID 1 configuration failing, is only one possible catastrophic-failure scenario an admin must foresee and guard against.

Which is why 2) a RAID 1 setup should not be the only backup.

GaryK

2 users thanked author for this post.

teuhasn2, Will Fastie

Reply | Quote

RAID ISN’T A BACKUP!  Just that he was arguing that, I’d consider them not a good information source.

RAID is to prevent downtime.  RAID prevents time loss.  Backups prevent data loss.

RAID used to be to expand capacity and speed over a single drive but that’s now becoming a moot point with 8TB+ NVMe enterprise SSDs.

Reply | Quote

Ben Shoemaker wrote:

RAID used to be to expand capacity and speed over a single drive but that’s now becoming a moot point with 8TB+ NVMe enterprise SSDs.

At extreme expense. A typical NVMe SSD at 8TB is over $1,000. And as I point out, that can fail, too.

1 user thanked author for this post.

teuhasn2

Reply | Quote

RAID 5 is not a good choice if you are using multi-terabyte drives. RAID 5 can withstand a single drive failure. When a RAID 5 array is rebuilding, the drives are subject to very high I/O and mechanical stresses (spindled drives). Because of the extended rebuild times (1+ days) on multi-terabyte drives, a 2nd drive failure is highly probably. When this happens, all data is lost.

At my work we use RAID 1 when write performance isn’t a concern. Everything else is RAID 10. Of course, RAID 10 is too expensive for most home users….if their hardware even supports it. Assuming you have a good backup solution (3+ copies of your data that gets regularly backed up), RAID 1 is a perfectly viable disk redundancy solution.

Reply | Quote

LizKos wrote:

a 2nd drive failure is highly probable

You’re right about the additional overhead when rebuilding RAID 5 because to restore one block on the replaced drive, it’s necessary to read one block from every other drive in the array. It’s one reason to consider RAID 6 because it can survive two failures at the same time.

Nonetheless, RAID 5 is widely used and has proven itself reliable, even for large drive capacities. And there’s the matter of economy. A good 2TB drive costs about $60 ($30/TB), but a good 8TB drive costs $140 ($18/TB). That leaves more room in a budget for replacement drives.

Reply | Quote

There is a perverse drawback to increased reliability: Drives and drive arrays now last so long that one often finds that new,  similar replacement drives are no longer available when needed. This is rarely an insurmountable problem, but it does introduce complications that must be considered when replacing a dissimilar drive in an array. It’s much simpler simply to slip in a similar drive; for that reason, I always bought spares when setting up RAID arrays.

GaryK

Reply | Quote

I always bought spares for future use

You’re right about drives lasting a long time, although I stand by my five-year rule. But if you buy spares at the time, they will be old by the time they are needed.

SATA isn’t going away anytime soon.

Reply | Quote

Will Fastie wrote:

But if you buy spares at the time, they will be old by the time they are needed.

Old is relative. Back in the day, first and second generation 5, 10, and 20mb hard drives that sat idle could develop frozen spindles. (Frozen spindles could sometimes be freed up by tapping them with a screwdriver; ahh, the good old days.) I haven’t experienced, not to mention heard of, an idle drive developing a frozen spindle in a long while. Modern spares bought when the system was bought are exactly as old as the drives in the system but they are unused, and in my experience with hundreds of modern drives over the past 20 years, they don’t deteriorate sitting idle. Over those years I’ve used many several-year-old, but otherwise new, drives of different technologies–I’ll wager you have occasionally used several-year-old, but otherwise new drives too–and never had one that didn’t perform as well as a brand new drive for as long as a brand new drive performed well.

GaryK

Reply | Quote

I am definitively team Will, sorry Fred. I think RAID 1 is great.

It is not a substitute for backups, but it does serve this function in part (and I say in part) if you have people who don’t do very regular backups like almost everybody I help. Yes, you can get hit by a virus, but if you just suffer from a hardware failure and you don’t loose any minute of work, just replacing the hard disk, it is quite nice. In my experience, I have seen much more hardware failures than data loss due to a virus, oddly.

I think it is a great addition to a regular backup, even more if your backups are monthly or less frequent. I probably saved more data using RAID 1 than anything. The rescues I had to rely on backups were computers without RAID and often, people didn’t do backups close to often enough. Not everybody gets hit by viruses.

On a server level, RAID 10 is better for performance. It combines mirror and striping so you can read even faster than with a mirror only. I don’t like RAID 5. When you spend that much money, use RAID 10 or even RAID 1 to get the added performance benefit of being able to read two drives at the same time, plus maximum reliability. Over the lifetime of the computer, it is well worth it in my opinion and the performance boost is good, as well as the added reliability due to what gmileon said. RAID 5 is a compromise when data storage cost is too expensive. I don’t have huge amount of data even at the server level so that is why I don’t like RAID 5 at all for mission-critical work. Performance of both RAID 5 and 6 is lower than a single drive, I believe, also.

In theory, even on a regular PC, RAID 1 should be faster on read, but like Will, for the scenario I use it for at the PC level, storing data, I don’t care if it is even slower with a certain implementation. Reliability and safety of the data is the priority here.

Instead of Intel RAID, I switched to native Windows ReFS software mirrored drives a while back. Intel RAID had issues with WD Black drives taking many seconds to get out of sleep. With ReFS, in theory, you also don’t have to consider the different hardware controllers issue. Also, it is very easy to configure in Windows.

Lastly, unlike Intel RAID 1, ReFS protects against bit rot, which Paul here doesn’t seem to believe in because he thinks CRCs already does enough. That is an interesting question, as I don’t know why we would have many filesystems on Linux too created to protect against bit rot if it is not an issue, but I am open to the idea that maybe people worked on all this for years for nothing. Something to research, I guess, but in the meantime, I will benefit from the mirrored effect even if bit rot isn’t an issue.

ReFS was affected by Windows feature updates in a totally unacceptable manner, but I didn’t suffer from it because I always delay feature updates by as long as I can. Also, Microsoft removed it from the PRO version to restrict it to a new “Workstation” version of Windows at some point, a version that not many people know of. This is not good in my opinion because it reduces even more the user base so we end up with stupid issues like when Microsoft broke it after a feature update and they had not been aware of it before releasing it. I guess they needed to put a few things together marketing-wise to justify creating a Workstation version for milking more money out of the professionals. Some of them must have been very angry if they lost data and time due to the feature update.

I think ReFS doesn’t get the love it deserves. For now, I have many PCs running on it since the days of Windows 8.1, but I didn’t have a single hard drive failure to report how it went so I can’t comment on that part. At least, it didn’t do like many of the PCs I managed did with Intel RAID where sometimes the array doesn’t get recognized anymore and the user starts running Windows with two identical disks together…

I think RAID 1 or mirrored drives are not used because people don’t know about it. Whenever I suggest it to someone I help, a lot of them will want it because they know they are not that good at backing up often enough. Knowing how hard it is for people to change, it is easier to set them up with mirrored drives and remind them periodically to do a backup than just not doing anything. We have to live in the real world that most people live in while trying to educate them to do better.

If you have a ReFS formatted drive, Windows will recognize it and use it even if it is not running the Workstation version. I think it can even update the version of ReFS on the disks when a new one is available. I did it on my Windows Pro version, having created the ReFS mirrored drive before ReFS got restricted to Workstation version, but upgrading the version after the restriction even on Pro version. ReFS is still actively developed in the servers versions of Windows too, but I don’t know if it is used much due to the fact that servers often have great hardware RAID controllers.

As for having two disks break at the same time, it almost happened to me in a server with a 2 days delay. Thanks to HP’s great 4h service window, the bad hard drive had already been hot swapped and reconstructed before the second one broke. I think having HDs break exactly at the same time is highly unlikely, so you might have time to replace the drive or do a backup if you are not current. Still, don’t rely on RAID for backups only, of course, but I will take the added peace of mind and convenience.

 

3 users thanked author for this post.

wavy, Will Fastie, gkarasik

Reply | Quote

Can I reinforce Will’s warning about 2 drives failing close to the same time, being in the same manufacturing batch. It happened to me on my (Netgear) NAS, after only around 2 years’ use and quite shocked me.  Forum responses helped me to understand how this might happen – along the lines Will describes.

I later moved to a Synology NAS and have made sure I always buy NAS grade disks and replace after 5 years. I take the point that I should stagger the replacements in future.  Also, upgrading the NAS (which uses Synology’s RAID implementation) was a piece of cake. See this post

#2358903

Chris
Win 10 Pro x64 Group A

1 user thanked author for this post.

Will Fastie

Reply | Quote

gkarasik wrote:

… they don’t deteriorate sitting idle

I keep all old hard drives when my PCs are decommissioned. They sit in a box (with a tiny bit of desiccant). Sooner or later, those drives won’t work. They either won’t spin up, can’t be detected, or won’t move the heads.

Age is age.

Reply | Quote

AlexEiffel wrote:

I think RAID 1 or mirrored drives are not used because people don’t know about it.

That’s a good point, but I also think people are scared, or at least hesitant, to try it.

1 user thanked author for this post.

AlexEiffel

Reply | Quote

So why do all your computer names start with the letter “O”? Is this like families who name all their kids with the same first letter?

Reply | Quote

ibe98765 wrote:

So why do all your computer names start with the letter “O”?

Can’t I have some fun?

It started because my Win10 machine is housed in a black case and the new Win11 box is housed in an ivory case. Because my backup PC is a Dell OptiPlex, I called that one Opti, and from that came Onyx and Opal. The Vostro is the odd PC out, thus Outlier.

1 user thanked author for this post.

PKCano

Reply | Quote