Hello all,
Please, I’m looking for a good advice about what would be better, creating a raid0/1 using the controller that came with my Asus ROG Laptop - (Advanced Intel Rapid Storage Technology (VMD driver)) or software raid by mdadm?
Much appreciated your answers.
Thanks!
I never use RAID0. That is striped across devices, and each device is a single point of failure that wipes out the entire file system. Raid 0 was introduced as a way to have more space when the physical storage devices were MUCH smaller than they are today. A single device failure with raid0 wipes out everything on the entire raid array.
In the past fedora was unable to even see sata devices when the intel software raid was enabled. Things have improved but I still would never use it (until it has a history of successful usage).
I personally use mdadm, but also only with either raid 5 or 6 so that I have additional redundancy in case of a drive failure.
Advanced Intel Rapid Storage IS software RAID and fully supported by the linux kernel and the mdadm tools.
I use this on my file server and it works very well.
yeah, I know all that.
I am interested about the performance of the hardware raid offered by the laptop or better stick with mdadm software raid.
thanks …
so there isint actually a hardware controller ?
Correct.
so why using it and not use the mdadm or btrfs native raid?
The UEFI BIOS can access data on the disks.
If I mirror a pair of drives then I can boot from the disk as the UEFI files are accessible to the BIOS.
But I have not used the RAID to join disks together for the reasons that @computersavvy explained. One disk failure means ALL your data is lost.
Better to mount the second disk as it’s own file system.
yeah, I know the risks, I just wanted to clarify and to find out which solution provides a better performance: BIOS or raid made by Fedora 
It is the same thing.
When you are booting up the UEFI BIOS handles the RAID until the OS is running then the OS handles the RAID and the UEFI BIOS is nolonger involved.
then I guess I’m gonna stick with raid made by Linux.
thanks!
I’ve heard that Btrfs RAID is currently unstable/unreliable.
A downside to MD RAID1 is that it doesn’t do checksums to protect against bit rot. (If a whole disk fails, then there is redundancy, but if individual sectors are corrupted, MD RAID1 will just pick one or the other copy at random, it won’t necessarily be able to determine which copy is good.)
ZFS is very reliable and supports full checksums of all the data, but it is not “supported” by Fedora Linux.
Btrfs on top of MD RAID1 is another option that does provide both redundancy and checksums, but it might be slightly more complicated to configure.
Just FYI.
i’d stay away from BTRFS although i love the snapshots idea.
Last week I only did “balance” on a raid and it wiped out the whole GPT table.
On reboot everything was gone and the only option on the screen was to boot from network
I had a bad experience with Btrfs as well, but it was many years ago (the first time Fedora tried to switch to Btrfs). I subsequently switched to ZFS and I’ve been happy with that ever since. I don’t use Btrfs on anything these days. (ZFS does snapshots as well and I use them extensively.)
I also want a fast and reliable filesystem.
What would be that one, in your opinion ?
I’m happy with ZFS’s performance, but it does use a lot of RAM[1] and, depending on settings (which checksum and compression algorithms you are using and how well they match your hardware’s abilities), it can be a little slower. I hear XFS is suppose to be good performance wise, but I don’t use it personally.
That is to say, for good performance, you will want to have plenty of RAM for ZFS to take advantage of. ↩︎
like how much?
this is a development station and day by day normal usage, with 32 GB ram and a i9 CPU
ZFS would perform very well on that system. By default, it will try to allocate half the RAM for “caching”. (But it won’t be “reserved”. If any other process on the system needs the RAM, ZFS will immediately flush/drop some of the cache and yield the RAM to whatever needs it.) Also, ZFS is quite good at taking advantage of multiple cores, so that i9 will also help. Basically, the more hardware you “throw” at ZFS, the better it will perform. It also has impressive algorithms for distributing the reads and writes across the disks if you have several. Since ZFS is both the filesystem and the volume manager at the same time, there are several ways it can be clever about queuing and distributing the writes and reads so that it will get the max out of all the drives. (This isn’t true of, e.g., a Btrfs on MD RAID configuration – MD will be “blind” to some of the info that Btrfs will have in its data structures about where the next reads/writes are likely to happen and it won’t be able to optimize for those.)
yeah, playing with 32 cores and they are good friends
i could give it a try to see what’s about the ZFS
ZFS should be the standard Fedora FS. It’s amazing how the Proxmox team has a reliable ZFS installer, yet for decades now nobody else has been able to do that in the whole Linux ecosphere.
Sorry to be blunt, but BTRFS is nothing but a joke. It’s not a serious project. IMO it was a major mistake for Fedora to choose it.
Initially when switching from Debian to Fedora, i thought i’d just do BTRFS to avoid issues, after all, it must have snapshots and rollback support atleast, right?
LOL was i wrong, i read some article on how to take a snapshot, boot a Fedora System from that snapshot, or do a rollback, and it was just ridiculous, messing with readonly crap, ridiculously complicated GRUB issues and on and on. What a joke.
zfsbootmenu has zero issues., and ZFS should be the default for any serious project. When i take a snapshot it means /boot as well.
Partitioning a HDD is something ppl did 20 years ago, it should not be necessary anymore. That’s why we have ZFS.
About Raid0. It makes perfect sense to run your OS from 2x 2TB SSD drives in a stripe. You get 4TB continuous space, and maybe even a bit more speed.
There’s no point in throwing away half your storage space, and or pay DOUBLE for your SSD’s, just incase one drive might or might not fail in the next 10 years, when statistically it is extremely unlikely. That’s what backups are for.
So run Fedora on ZFS root, use zfsbootmenu so you can snapshot, clone and rollback your OS simply and quickly, without needing to read hundred page guide’s and spend days on forums.
Stripe your SSD’s to double your continuous storagespace, and keep periodic backups of the important stuff, should your drive fail, which is EXTREMELY unlikely.
Live alittle.
And shame on the Fedora team for choosing BTRFS, They will come to regret it. BTRFS is nothing but a bad joke.