Raid 5 is becoming less viable due to the increasing rebuild times, necessitating raid 1 instead. But new drives have better iops too so maybe not as severe as predicted.
Yeah I would not touch RAID 5 in this day and age, it’s just not safe enough and there’s not much of an upside to it when SSDs of large capacity exist. RAID 1 mirror is fast enough with SSDs now, or you could go RAID 10 to amplify speed.
When setting up RAID1 instead of RAID5 means an extra few thousand dollars of cost, RAID5 is fine thank you very much. Also SSDs in the size many people need are not cheap, and not even a thing at a consumer level.
5x10TB WD Reds here. SSD isn’t an option, neither is RAID1. My ISP is going to hate me for the next few months after I set up backblaze haha
But have you had to deal with the rebuild of one of those when a drive fails? It sucks waiting for a really long time wondering if another drive is going to fail causing complete data loss.
Not a 10TB one yet, thankfully, but did a 4TB in my old NAS recently after it started giving warnings. It was a few days iirc. Not ideal but better than the thousands of dollars it would cost to go to RAID1. I’d love RAID1, but until we get 50TB consumer drives for < $1k it’s not happening.
tbf all the big storage clusters use either mirroring or erasure coding these days. For bulk storage, 4+2 or 8+2 erasure coding is pretty fast, but for databases you should always use mirroring to speed up small writes. but yeah for home use, just use LVM or zfs mirrors.
depends on a lot of factors. If you only need ~30TB of storage and two spare RAID disks, 3x 40TB disks will be much more costly than 6x 10TB disks, or even 4x 20TB disks.
The main issue I see is that the gulf between capacity and transfer speed is now so vast with mechanical drives that restoring the array after drive failure and replacement is unreasonably long. I feel like you’d need at least two parity drives, not just one, because letting the array be in a degraded state for multiple days while waiting for the data to finish copying back over would be an unacceptable risk.
I upgraded my 7 year old 4tb drives with 14tb drives (both setups raid1). A week later, one of the 14tb drives failed. It was a tense time waiting for a new drive and the 24 hours or so for resilvering. No issues since, but boy was that an experience. I’ve since added some automated backup processes.
Yes this and also scrubs and smart tests. I have 6 14TB spinning drives and a long smart test takes roughly a week, so running 2 at a time takes close to a month to do all 6 and then it all starts over again, so for half to 75% of the time, 2 of my drives are doing smart tests. Then there’s scrubs which I do monthly. I would consider larger drives if it didn’t mean that my smart/scrub schedule would take more than a month. Rebuilds aren’t too bad, and I have double redundancy for extra peace of mind but I also wouldn’t want that taking much longer either
I guess the idea is you’d still do that, but have more data in each array. It does raise the risk of losing a lot of data, but that can be mitigated by sensible RAID design and backups. And then you save power for the same amount of storage.
I deal with large data chunks and 40TB drives are an interesting idea… until you consider one failing
raids and arrays for these large data sets still makes more sense then all the eggs in smaller baskets
You’d still put the 40TB drives in a raid? But eventually you’ll be limited by the number of bays, so larger size is better.
They’re also ignoring how many times this conversation has been had…
We never stopped raid at any other increase in drive density, there’s no reason to pick this as the time to stop.
Raid 5 is becoming less viable due to the increasing rebuild times, necessitating raid 1 instead. But new drives have better iops too so maybe not as severe as predicted.
Yeah I would not touch RAID 5 in this day and age, it’s just not safe enough and there’s not much of an upside to it when SSDs of large capacity exist. RAID 1 mirror is fast enough with SSDs now, or you could go RAID 10 to amplify speed.
When setting up RAID1 instead of RAID5 means an extra few thousand dollars of cost, RAID5 is fine thank you very much. Also SSDs in the size many people need are not cheap, and not even a thing at a consumer level.
5x10TB WD Reds here. SSD isn’t an option, neither is RAID1. My ISP is going to hate me for the next few months after I set up backblaze haha
But have you had to deal with the rebuild of one of those when a drive fails? It sucks waiting for a really long time wondering if another drive is going to fail causing complete data loss.
Not a 10TB one yet, thankfully, but did a 4TB in my old NAS recently after it started giving warnings. It was a few days iirc. Not ideal but better than the thousands of dollars it would cost to go to RAID1. I’d love RAID1, but until we get 50TB consumer drives for < $1k it’s not happening.
tbf all the big storage clusters use either mirroring or erasure coding these days. For bulk storage, 4+2 or 8+2 erasure coding is pretty fast, but for databases you should always use mirroring to speed up small writes. but yeah for home use, just use LVM or zfs mirrors.
Of course, because you don’t want to lose the data if one of the drives dies. And backing up that much data is painful.
depends on a lot of factors. If you only need ~30TB of storage and two spare RAID disks, 3x 40TB disks will be much more costly than 6x 10TB disks, or even 4x 20TB disks.
The main issue I see is that the gulf between capacity and transfer speed is now so vast with mechanical drives that restoring the array after drive failure and replacement is unreasonably long. I feel like you’d need at least two parity drives, not just one, because letting the array be in a degraded state for multiple days while waiting for the data to finish copying back over would be an unacceptable risk.
I upgraded my 7 year old 4tb drives with 14tb drives (both setups raid1). A week later, one of the 14tb drives failed. It was a tense time waiting for a new drive and the 24 hours or so for resilvering. No issues since, but boy was that an experience. I’ve since added some automated backup processes.
Yes this and also scrubs and smart tests. I have 6 14TB spinning drives and a long smart test takes roughly a week, so running 2 at a time takes close to a month to do all 6 and then it all starts over again, so for half to 75% of the time, 2 of my drives are doing smart tests. Then there’s scrubs which I do monthly. I would consider larger drives if it didn’t mean that my smart/scrub schedule would take more than a month. Rebuilds aren’t too bad, and I have double redundancy for extra peace of mind but I also wouldn’t want that taking much longer either
I guess the idea is you’d still do that, but have more data in each array. It does raise the risk of losing a lot of data, but that can be mitigated by sensible RAID design and backups. And then you save power for the same amount of storage.
These are literally only sold by the rack to data centers.
What are you going on about?