Is it safe to mix new and used SSDs to make a RAID 1?

[geno888]

Hi

I have a Samsung 850 PRO drive (currently in use in my computer). Because of hardware problems, I'm planning to buy a new motherboard and I'm exploring various scenarios.

One thing that I'm considering is to make a RAID 1 buying another 850 PRO drive.

Is it a good idea to use a brand new disk with one that already has 2.02 TB of total bytes written (according to the Samsung magician software)?

Maybe is better to use two brand new disks?

Moreover, I'm a bit concerned about that "2.02 TB of total bytes written": that disk only contains operative system and programs, but no data at all. I wonder where all these written data comes from and why it seems that this number increase on an almost daily base of about 0.01 TB (I mean that yesterday that value was 2.01 TB)

Maybe is time to get anyway a new disk before this one fails?

Thanks for help

 

[BonzaiDuck]

Hi

I have a Samsung 850 PRO drive (currently in use in my computer). Because of hardware problems, I'm planning to buy a new motherboard and I'm exploring various scenarios.

One thing that I'm considering is to make a RAID 1 buying another 850 PRO drive.

Is it a good idea to use a brand new disk with one that already has 2.02 TB of total bytes written (according to the Samsung magician software)?

Maybe is better to use two brand new disks?

Moreover, I'm a bit concerned about that "2.02 TB of total bytes written": that disk only contains operative system and programs, but no data at all. I wonder where all these written data comes from and why it seems that this number increase on an almost daily base of about 0.01 TB (I mean that yesterday that value was 2.01 TB)

Maybe is time to get anyway a new disk before this one fails?

Thanks for help

I think you're imagining things. Over several months to a year's time, 2.02TB would seem normal. Those disks begin to fail between something around 250 TB and 500 TB. They may last longer.

You can reduce the size of your swapfile/page-file to reduce the daily wear and tear. But there's a lot of life left yet in that "old" 850 pro.

I could suggest that you carefully "Secure Erase" the old drive, than pop in the new one for a RAID 1, initialize the array and load the OS.

 

[VirtualLarry]

I think you're imagining things. Over several months to a year's time, 2.02TB would seem normal. Those disks begin to fail between something around 250 TB and 500 TB. They may last longer.

850 Pro is MLC. I would say those lifespans seem conservative. Maybe they would be correct for the smaller drive sizes.

I could suggest that you carefully "Secure Erase" the old drive, than pop in the new one for a RAID 1, initialize the array and load the OS.

That's a good idea.

 

[VirtualLarry]

sorry, dup.

 

[geno888]

Thanks for your answers

This is my first SSD (it's the 250 GB model), so I don't know exactly the numbers (I'm still in learning mode). I'm happy to know that the disk is still in the safe zone

 

[Billb2]

Thanks for your answers

This is my first SSD (it's the 250 GB model), so I don't know exactly the numbers (I'm still in learning mode). I'm happy to know that the disk is still in the safe zone

You can (should) move the page file and swap file to a non SSD.

Also there are may backup options other than RAID 1. Have you considered anything else?

 

[UsandThem]

One thing to consider: Using more than one PCIe M.2 drive on many motherboards drops the performance of the drives, disables some SATA ports, and can disable the speed of a PCIe slot, or disable it all together.

Since each motherboard is different, you'll have to refer to your user manual.

 

[corkyg]

BTW, RAID1 is not a good backup. Any data corruption or malware will be present on both drives. It does provide for system redundancy in case of hardware failure of one of the drives, but you should also have another backup for any critical data.

 

[VirtualLarry]

You can (should) move the page file and swap file to a non SSD.

Why would you do anything so silly? That negates much of the performance of an SSD, quite frankly.

Also there are may backup options other than RAID 1. Have you considered anything else?

RAID is not a backup!

 

[BonzaiDuck]

850 Pro is MLC. I would say those lifespans seem conservative. Maybe they would be correct for the smaller drive sizes.

That's a good idea.

I just remember vague ranges reported in a TBW comparison review of several drives. There were actually cases or models where the expected life in TBW was around a petabyte.

Since there are several decent SSDs at lower prices now available, I also dabble in the low end, even for buying an occasional "Pro" drive. So it may be more of a case that I'M being conservative . . .

It's just comforting that the various SSD-makers provide a utility to let you glimpse this "aging-through-usage" process. It also occurs to me that any vestigial page-file or much larger hiberfil.sys could be located on traditional media, and there are some tweaks you can make to reduce the number of writes through restart or shutdown which aren't necessary. I'd need to go back and review yesterday's notes just to make it look like I know what I'm talking about.

[later . . ] OOPS! I think the hiberfil.sys has to be on the boot-disk, no matter what.

 

[Billb2]

Why would you do anything so silly? That negates much of the performance of an SSD, quite frankly.

RAID is not a backup!

With 16 gigs of memory there is little use of a page file nor swap file. Yeah there a performance hit having then on a spinner (which I do) but the performance loss to the slower spinner is offset by the ability to access an SSD and page / swap file simultaneously.

A RAID 1 is a copy - of a corrupted OS, malware, viruses or a compromised computer. It is not necessarily a clean backup of data.

 

[BonzaiDuck]

With 16 gigs of memory there is little use of a page file nor swap file. Yeah there a performance hit having then on a spinner (which I do) but the performance loss to the slower spinner is offset by the ability to access an SSD and page / swap file simultaneously.

A RAID 1 is a copy - of a corrupted OS, malware, viruses or a compromised computer. It is not necessarily a clean backup of data.

My understanding was that a minimally-sized swap-file -- 700MB to 2GB -- assured that the volume shadow-copy service worked correctly, which in turn guarantees that my nightly server backups of our workstations always complete successfully. And usually when I had problems with that backup process, it was because of problems with VSS.

But I can certainly put the swap file or page-file on the spinner. I can't understand why I haven't done it, because I knew all about this before. Something I just overlooked, I guess . . .

 

[Chaotic42]

As long as you're doing a proper backup, I think you'll be fine. Just remember like VirtualLarry said, RAID is not a backup - it's a safety net. Together with a good power supply, a solid UPS, and regular backups (offline, preferably off-site or at least in a fire- and waterproof safe), you'll add in a layer of security.

 

[BonzaiDuck]

As long as you're doing a proper backup, I think you'll be fine. Just remember like VirtualLarry said, RAID is not a backup - it's a safety net. Together with a good power supply, a solid UPS, and regular backups (offline, preferably off-site or at least in a fire- and waterproof safe), you'll add in a layer of security.

Assuming you're addressing the OP, I'll add to the topic-at-hand to note that I use a drivepool (Stablebit) on the server. I can duplicate data at the file and folder level, leaving less essential files at risk without duplication with less demand for server "virtual drive/disk" capacity. My workstation backups are all duplicated.

From the server, I back up my personal data files to one disk, and (occasionally) back up the workstation backups to another disk.

So if you're going to make your system redundant, you still need to back up.

 

[CiPHER]

You should be aware that consumer-grade Samsung SSDs are unsuitable for anything other than standalone desktop usage, meaning no RAID or other form of complex storage.

Samsung SSDs use PoR recovery mechanism that allows the SSD to go back in time after unexpected power loss. Other SSDs use capacitors or other means to protect the SSD against corruption. Samsung uses journalling on the mapping tables to achieve this task. This works well for regular desktops, but can cause havoc when the SSD is used for complex storage, like RAID.

Imagine you have two RAID1 and one SSD decides to go back in time. You can sync, but which SSD has the correct data? And, will the RAID engine actually detect the disparity in time? The problem is even worse for other RAID levels like RAID0 and RAID5.

Basically, I recommend against using consumer-grade Samsung SSDs for anything other than regular desktop usage. The enterprise datacenter-grade Samsung SSDs are exempt from this problem, since they have proper PLP.

 

[BonzaiDuck]

You should be aware that consumer-grade Samsung SSDs are unsuitable for anything other than standalone desktop usage, meaning no RAID or other form of complex storage.

Samsung SSDs use PoR recovery mechanism that allows the SSD to go back in time after unexpected power loss. Other SSDs use capacitors or other means to protect the SSD against corruption. Samsung uses journalling on the mapping tables to achieve this task. This works well for regular desktops, but can cause havoc when the SSD is used for complex storage, like RAID.

Imagine you have two RAID1 and one SSD decides to go back in time. You can sync, but which SSD has the correct data? And, will the RAID engine actually detect the disparity in time? The problem is even worse for other RAID levels like RAID0 and RAID5.

Basically, I recommend against using consumer-grade Samsung SSDs for anything other than regular desktop usage. The enterprise datacenter-grade Samsung SSDs are exempt from this problem, since they have proper PLP.

And that's why, sometime around 2011, I abjured further use of add-in hardware RAID controllers and RAID-mode BIOS settings. You'd be best to use those settings for NAS or Enterprise disks, which include a fault-tolerant feature -- I think it's called TLSR. correct me if wrong.

Like I described, there are other ways to create redundant storage, and there are many ways to create backups -- which shouldn't be dismissed just for the redundancy.

We've opened up a real bottleneck with SSD technology, and even much more so with NVMe M.2 or PCI-E SSDs. There's no need anymore to add complexity to the configuration or spend money on an expensive controller just to add speed. Speed is no longer a priority enough to lead you toward RAID as a solution for anything unless you're building a server or NAS.