Need to upgrade the HDDs in our server, it currently has 5 x 1TB HDDs (RAID 5 Array) in it running 4 server VMs. It's running out of space so I was thinking of upgrading to 2Tb SSDs but there's a huge difference in prices:

• Samsung SSD 860 Pro MZ-76P2T0B/EU £421
• Kingston A400 SSD SA400S37/1920G £155
• Crucial MX500 CT2000MX500SSD1 £186

Some are way more expensive than the Samsung, but I thought the point of SSDs was that they were inherently quicker and more reliable than an HDD so why such huge price variation? And what should I be looking to buy?

I've always had good luck with the Samsung drives.

I seem to recall that there can be some issues with using SSDs in full-on servers though...

Short answer: just buy the Crucial drives and stop fussing.

Long answer: Samsung has been spending tons of cash marketing their drives as "secure" and "reliable". They probably are, too. But the Crucial ones have been doing this longer, and have always had the "safe" reputation that Samsung craves, at a fraction of the price. Crucial is actually Micron, and makes their own flash chips (as does Samsung).

Kingston for me has always been more of a commodity company, sourcing cheap components, and changing the innards frequently without relabelling or changing model names/numbers. I have a couple of Kingston drives here, but don't trust them enough for important uses. They're probably fine, but who knows.. ?

To this date, the only 100% failure of an SSD that I have had was a Kingston drive. But a friend of mine had a 100% death of an ADATA drive. I suppose this sort of thing is bound to happen once in a while, but far FAR less often than with mechanical drives.

If you really want quality, then hunt out brands/models with "powerfail capacitors". Manufacturers can include these for an extra quid or two if they choose, and IMHO those can make all of the difference for reliability. These basically give the drive's firmware a tiny time buffer in which to ensure things are in a consistent state upon sudden loss of power. That's important when a drive could be moving data and rewriting flash blocks in the background ("garbage collection", anyone?).

Micron/Crucial used to put "powerfail capacitors" in all of their drives, which made them my fav brand years ago. Nowadays some models lack them. TomsHardware reviews generally show which have them and which don't. Some Samsungs also now have powerfail capacitors. And most brands seem to include them in their insanely priced "enterprise" models.

Another consideration is whether or not companies are known for releasing firmware updates to fix bugs discovered over time. Samsung and Micron/Crucial do so reasonably often.

Those are my general opinions on things. Here, my own server currently has these installed:

2TB [0:0:0:0] disk ATA Crucial_CT2050MX R060 /dev/sda
2TB [1:0:0:0] disk ATA Micron_1100_MTFD U031 /dev/sdb
1TB [4:0:0:0] disk ATA Crucial_CT960M50 MU05 /dev/sdc

I have another 2TB Micron drive in reserve as well -- just haven't gotten around to deploying it yet.

Thanks guys, I did think Crucial would be a good bet

For "servers", it really depends on what stress the server will be under. For a shared corporate email server with dozens or more users, one should really be looking at the more expensive enterprise SSDs. These typically have much longer life expectancies, as expressed by the manufacturer's "endurance" rating.

Endurance is improved by lowering the "write amplification" factor, and by preferring FLASH memory that uses fewer bits per cell than what might be found in consumer level drives. Fewer bits per cell means more chips are needed for similar capacities, driving up the price substantially.

For home use, like me here, any modern drive of (say) 1TB or larger will have enough wear-levelling capacity that I should never need be concerned about such stuff. So I no longer pay attention to it for my own server, or even for the 500GB drive in my main work machine.

I do buy "burner" drives for non-critical uses, generally CAD$25 for 128GB of SSD with no cache and no powerfail capacitors. These make great/fast substitutes for otherwise slow USB sticks for video files and the like, as well as for temporary operating system installs on my test machines.

One extra note: take a closer look at the drive you choose, and ensure that it does have some onboard RAM cache. The lowest end Crucial ones might lack that, and for 20 quid more there'll be a very similar model which does have some RAM cache.

You WANT/NEED some RAM cache on the drives, otherwise writes can be painfully slow and endurance will suffer.

EDIT: Ah. Yes, the MX500 is the one to buy. They also have the BX500 which is the stripped-down version: avoid for server or main machine uses.

SSD drives are either read intensive, write intensive or mix use. The same hardware can be used for all three, but the difference will be in the firmware, and will result in different capacity sizes. The hidden capacity is used to replace cells as cells wear out, which will occur more frequently with a write intensive drive. This is why a 1TB write intensive drive will cost more then a read intensive.

On a related note; getting exposed to some Intel Persistent Memory at work, which is non-volatile memory, which can be used as RAM or storage. It is a little slower that RAM, but a lot cheaper, therefore a hell of a lot faster then any storage, but more expensive.

If you have spare PCIe slots in your server, and it runs Linux or anything else newish, and you don't care about hotswap, then the things that let you stick two or four NVME SSDs in a PCIe slot will give you much better performance than SATA SSDs.

Peter

Not sure it does Peter, I'll take a look. It was bought in 2016

No question that native PCIe solutions like NVMe will be faster. But they also suffer from lack of diagnostic tools and the like. SATA has been around now for a long time, and there are a lot of tools for it on most operating systems that understand how to access, format, read diagnostics, upgrade firmware, etc..

Not so much for NVMe. The NVMe drive vendor will have their own basic tools to update firmware and (in theory) test for defects etc. But if WD is any example, they are not always trustworthy.

So the point here being: for a (home) server (especially!), the ability to swap drives around, diagnose and recover from failures, is very high on my personal priority list. So while I adore the lighting quick NVMe drive in my ultrabook, the server here instead gets SATA2 SSDs.

They are still leagues faster than the GigE network connection, which itself is currently still 3X faster than the internet link, so not really a bottleneck for anything here.

Cheers

PCIe shouldn't be a problem on a 2016 machine for just storage. Usually the problem lies in *booting* from NVMe drives if anything.

You mention RAID5 though. Is that hardware or software RAID? If hardware that will likely rule out anything NVMe/PCIe as you'll need/want(?) to continue to use the SATA connections.

Yeah, hardware RAID so I'll be staying with SATA

My 2014-ish desktop boots from a SATA-connected SSD, and I have /home on a PCIe-connected NVMe (which is not as quick as the one in my NUC, but it's still very quick). It's connected to a Lycom DT-120, fwiw. Other adapters are available.

One of the MX500 disks appears to have failed, the array was showing as degraded and one of the drives was in error so I swapped it out. It's still under warranty so I could just send it back but I'm surprised it would have failed, and so quickly. Is there anything I can do to check the drive?

Read out the S.M.A.R.T. data to see if the drive had issues, versus the cabling or motherboard.

With Linux: smartctl -a /dev/sdX

I would also check to ensure the drive has up-to-date firmware. There is often perhaps one extra version after initial release, to fix the odd bug discovered once mass deployment (sales) began.

If there is newer firmware available, install it. Here, I might then just reformat the drive, aka "SECURE ERASE", and then restore from back-up. It'll probably last another decade or more -- but the S.M.A.R.T. data will give a better indication of that.

Note that I have never had to restore a retail-purchased SSD from backup. They have been super reliable for me.

Yeah, I really can't see why it would have failed seeing as none of the previous SATA disks had, that's why I was tempted to investigate. Will give it a go as soon as I get a chance

I can't see any issues in SMART, the Crucial utility says there's new firmware available so I'll apply that.

It doesn't seem to be reporting any issues. I had (have) a Crucial drive like that, random failures from time to time but "perfect" SMART data.

Turned out to be a loose SATA connector on the motherboard.

Cheers

Look like its the battery on the RAID card, it's a Fujitsu version of LSI MegaRAID SAS 2208 PCIe 3.0 x8.

Now how do I find a replacement battery? No longer supported by Fujitsu.

I think it's this one, I've ordered it.

Did you get the last one?

There was a choice of two resellers on Prime, not sure which that was but it's due for delivery today. Then I've got to open the server up....