In my limited experience RAID is not worth it for backup drives. You need to carefully consider the configuration, which is not straightforward. Also I found out the hard way, that if you get this wrong and one drive fails then the second drive, in effect, has the data encrypted so it cannot be accessed. But I must admit, I didn’t really understand the technical side - it just came configured for RAID (Lacie I think) and I just plugged it in!
There are many different kinds (“levels”) of RAID, with different capabilities. You obviously need to select the one that best suits your needs, and you need to be aware of each level’s benefits and limitations.
I’m not sure what you were using, but if you were buying a high-capacity drive using RAID internally to make two drives look like one large drive, then it was probably RAID 0, which has no redundancy.
For those who are curious, here are the most commonly used RAID levels:
RAID 0 is striping. Your data is spread across two or more discs with no redundancy. If one drive fails, the entire array is dead. This is not technically RAID because there is no redundancy. It can give you high speed and large capacity, but it is less reliable than formatting each drive as a separate volume.
RAID 1 is mirroring. The array consists of two drives, each being an exact duplicate of the other. The capacity of the array is the capacity of a single drive. It is the only possible RAID level you can use on an array consisting of only two drives. If one drive fails, the array can keep working as long as the other drive doesn’t also fail.
RAID 5 involves striping data across multiple drives. Parity data (which can be used to reconstruct data from a failed drive) is also striped across the array. This requires a minimum of 3 drives. The capacity of the array is the capacity of all but one of the drives (e.g. a 5-drive array has the capacity of 4 drives). Any single drive can fail without taking down the array, but a second failure before the first has been replaced and rebuilt will kill it.
RAID 6 is similar to RAID 5, but it uses two different parity mechansims, which are striped across the array. This requires a minimum of four drives. The capacity of the array is the capacity of all but two of the drives (e.g. a 5-drive array has the capacity of 3 drives). Two drives can fail without taking down the array, but a third failure before at least one of the failures has been replaced and rebuilt will kill it.
RAID 10 (aka RAID 1+0) is a hybrid of RAID levels 0 and 1. An even number of drives is required. The drives are configured as mirrored pairs (RAID 1), and data is striped across these sub-arrays (RAID 0). This gives you the improved performance and capacity of RAID 0, with the redundancy of RAID 1.
RAID 50 (aka RAID 5+0) is a system where two (or more) RAID 5 arrays are created, and data is striped across them. This gives better performance than RAID 5, but retaining its resiliency.
RAID 60 (aka RAID 6+0) is similar. Two (or more) RAID 6 arrays are created, and data is striped across them.
There are also RAID levels 2, 3 and 4, but they are not used these days. But they are interesting to read about.
RAID systems often also support hot and cold spare drives.
A hot-spare is a drive that is powered-on but is not used to store data. When the array detects a drive failure, the spare is immediately brought into service to replace the failed drive.
A cold-spare drive is kept powered-off until it is needed to recover from a failure. It is also used to refer to spare drives that you keep disconnected so you can quickly swap it into service when a failure occurs (vs. needing to buy a replacement at that time).
Needless to say, you need a lot of drives for RAID 5 and 6 to be practical, especially if you also want some hot spares. This typically means an enclosure that can hold them all and a controller board (or a dedicated CPU) to manage the array and make it appear as a single drive (or file server).
For example, one server at a company I worked for had 16 drives configured as two arrays. Each array was 7 drives (configured as RAID 6) and a hot spare drive. We didn’t stripe across the two arrays - the server was configured for two network volumes - one for each array. The server was configured to send alerts to our administrator when a drive failure is detected, so he could remove the bad drive and replace it with a new one (which would automatically become the new hot-spare). Of course, this was not a cheap device, but it was very reliable.
This is several years ago when I bought a dual drive system. Something like this:
I assume the drive was set up for RAID 1 (mirrored drives). My low-tech experience was that one drive failed and when I tried to access the data on the other drive it couldn’t be read. Probably some set-up mistake on my part but it is a word of caution about RAID.
Thanks Michael. My intention is to store all my data on the RAID drive (>16TB) and not on the Mac Mini. I will also need a backup solution which I guess does not need to be RAID.
I set Time Machine so that it would only run when the computer was plugged in, which I think is the default setting, and in Energy Saver, I enabled Power Nap so Time Machine could run when the computer was sleeping. (On my M1 MBP, this setting doesn’t exist.) At home, I don’t typically use my computer plugged in, so without Power Nap enabled, backups would have rarely run.
Coincidentally, this video was posted this month:
David Plummer is a retired Microsoft engineer. His YouTube channel is full of really interesting content.
I have a Synology NAS 218j. It no longer supports Plex, but Synology provides just fine media service with its DS apps. If you insist on Plex, you would have to pay more for the NAS or go with the Mini. I have had no end of problems with Time Machine over a network even with an AirPort when they existed (which is probably why they don’t anymore). Get a dedicated 1 TB SSD. They are cheap enough.
I have many systems in a home lab. I’ve cobbled together several differing solutions to this problem over the years. The best solution on the market right now is Synology. It’s an appliance, you plug it in, add your drives and optional SSD cache drives and you configure it via a web interface. It couldn’t be simpler. Then you forget about it and it just chugs along doing it’s thing like a good appliance should. The Synology hits every single one of your needs out of the box. Add a second one as a backup because RAID is not a backup solution. If you have a lot of critical data, you should back it up in a myriad of ways including something like Backblaze or at least multiple copies spread across systems. The Synology’s are small, quiet and very low power.
Most all other alternatives are overkill on compute resources or are so incredibly complex that it becomes a sort of hobby to keep it up and going. TrueNAS isn’t nearly as user friendly. But it is more powerful. Linux w/ZFS and LXD containers works great but it requires a lot of Linux command line experience, there are GUI’s but you have to install them and get them working and the command line is still king of the hill.
My MacStudio to contain only the bare essential data on it since I didn’t want to take it apart if the computer had problems or because of security reasons if I needed to take my computer to an Apple store to be repaired.
The system must be able to have handle many TBs of images from my high resolution cameras.
All my data and back up systems needed to be in one place except for the offsite back-up and not be involved with a lot of cables.
The data storage system must be quiet, available for all my computers, and back itself up automatically.
One of my good friends who is the military network administrator for a couple of hundred computers at an R&D facility set me up with a Synology 1821+. Synology comes with all the software one needs.
It contains two RAID 5 setups each having 4 eight TB drives. One backs up the other at three in the morning. TimeMachine is also implemented. If one hard drive goes in either RAID, you insert a new drive.
It is a quiet machine and made more so because it resides in a far away closet and is connected using ethernet. A quiet computer room is good. The unit has a dedicated APC UPS connected to it. Currently I have two large hard drives used for offsite back-ups that are alternately rotated every two weeks to a friend’s house.
Nice. Far from a cheap solution, but that should work great for a very long time.
And if 24TB per RAID volume isn’t enough, the compatibility table seems to indicate that it supports the largest drives currently manufactured. An array of 16TB drives in this configuration would translate to 48TB per volume and if you find yourself needing more than that, then you’re officially in the realm of needing professional server gear.
I am, however, curious about your backup policy:
It appears (from your description) that the second RAID is designed to be a clone of the first RAID.
Does it maintain historic backups (e.g. via snapshots or some other similar mechanism) or is it strictly a clone of the first RAID?
If it is strictly a clone, then why have the 3:00am backup instead of a single larger array? If the goal is to protect against the possibility of two drive failures, you might want to consider putting all 8 drives in a RAID6 configuration. This would give you similar reliability (able to tolerate two failures) and higher storage capacity (one 48TB volume vs. two mirrored 24TB volumes).
Of course, since you’ve got an IT expert helping you configure all this, I would defer to his expertise, but I am curious about this particular configuration choice.
Hi David. I asked my friend who is a network administrator to respond to you. He is what he said:
His statements are accurate. However, you are not using the second volume as a clone of the first—it is a destination for Hyperbackup, which does maintain a history of all files (until it runs out of space then deletes old ones, but can take a very long time depending on storage and usage). If you were only interested in data integrity, then his comment on using RAID 6 on the entire unit makes sense—that is how I have our set up at work. However, as he implies that does not give the ability to restore “mistakes,” only recover what you have. It also does not provide a backup in case of a full unit failure (such as motherboard or power supply). Of course yours does not protect against the latter either. At work I have ours back up using Hyperbackup to an identically configured unit in another physical location, so it is extremely fault tolerant and we can recover old files as needed.
Time Machine is Apple-centric, and has nothing to do with the NAS other than the NAS being the destination. The NAS treats the Time Machine backup (which is a single file—sparsebundle) like any other file.
Thanks. Now it all makes sense.
I have had lots of useful responses, thanks again. I am leaning towards a Mac Mini (M2 when it comes out) with 512GB storage (can I get away with 256GB?) and 16GB memory and attach to a RAID drive for my data. Are OWC as good as anything on the market (performance and reliability)? I was thinking of the ThunderBay 4 with USB-C. I plan to store Time Machine data (for this and my partners machine) on the attached RAID (should I create a separate partition for Time Machine?).
As a backup solution I was thinking of a NAS (Synology DS920+). Any thoughts on a backup strategy, is SuperDuper as good as anything (to backup the Mac mini and attached RAID)?
Out of curiosity, if a RAID drive fails (e.g. power supply or motherboard), can you simply insert the drives into another RAID enclosure and recover the data (if so I assume you have to use an enclosure made by the same manufacturer).
You are having fun.
Isn’t the common wisdom to have an off-site backup too (hence Backblaze in my case)?
I put an OWC RAID on my Studio for everything but the current year’s images (Nikon Z9 so large files) and the Lightroom catalog. Those reside on an OWC TB SSD so that I can carry the catalog with the laptop on travel.
The RAID gets backed up to an assortment of USB Seagate 4TB drives for local backup with CCC and it’s versioning…and it also gets backed up to BackBlaze.
The Studio boots from the internal drive and is always running as the house file server and one of 2 CCC Time Machine like backups since TM to network drives refuses to work properly, reliably, and consistently…which is irksome for a former Windows Sysadmin and Mac user and consultant since about 1990 despite redoing the setup numerous times. All the data lives on the RAID and/or external SSD…and the latter gets backed up to the RAID and internal drive as well.
There’s nothing wrong with doing it either way…NAS or mini…and costwise both are in the same general neighborhood…and with many TBs of images you’ve obviously spent a lot on camera gear and lenses…so I’m guessing that you’re not all that price sensitive. In that case…I would go with what you’re familiar or most comfortable with because in real terms a NAS is no different from a mini with a RAID on it. One never knows whether the NAS vendor will keep up whatever compatibility or apps it has to make it play nicely with macOS…see recent issues about Drobo basically abandoning the macOS market…and you can be sure that a mini will remain compatible. For me…any price savings with the NAS is balanced out by the noise and heat that the 3.5 drives make vs the 2.5s in the OWC RAID.
Off-site backup is part of the plan… not sure about fun, this is going to be a significant investment so need to get it right…
Did you find that when Apple switched from Intel processors it seemed that there were problems with SoftRAID and it took a couple of months to get it updated. Are you satisfied with a system using SoftRAID and have found that software works well with the last three MacOSes?
Hi Lynn, I asked this question in a separate post “SoftRAID Compatibility with macOS and Apple Silicon?”
Yeah…they had to fix it to work with Apple Silicon…and being a low level system sort of thing they took their time to make sure it was right…either that or it was a harder rewrite than they thought it would be I guess.
Very similar to my set up and I’ve found it works very well from the hardware and shareware aspects. I have had a couple of HDDs fail, but it was quite simple to replace those drives with new HDDs and was back in business almost immediately. A RAID configuration of drives is essential for a data store as well as a backup store. Offsite backup is also essential. Another thought - keep it simple and documented.