Is this an enterprise SSD?
- Thread starter promoxer
- Start date
chatgpt is usually best for these things, if you ask it with web search enabled on a model number it will check out the web and tell you yes or no and why it came to that conclusion.
Don't ever use the "R" variant, they are optimized for Reading only and have poor write performances, you have to use the "M" (mixed, read/write) variant, like DC500M or the newer model DC600M, a little more expensive but with optimal I/O performances and good durability (I use them)
From the Kingston datasheet (pdf):
I'd go for DC500M based on the 1.3 DWPD endurance.
But both variants can be considered "Enterprise".
Code:
Power Loss Protection: Tantalum Capacitors
> Endurance
DC500R:
480GB — 438TBW (0.5 DWPD)
960GB — 876TBW (0.5 DWPD)
1.92TB — 1752TBW (0.5 DWPD)
3.84TB — 3504TBW (0.5 DWPD)
DC500M:
480GB – 1139TBW (1.3 DWPD)
960GB – 2278TBW (1.3 DWPD)
1.92TB – 4555TBW (1.3 DWPD)
3.84TB – 9110TBW (1.3 DWPD)I'd go for DC500M based on the 1.3 DWPD endurance.
But both variants can be considered "Enterprise".
I can recommend the following website to quickly check and filter stuff based on properties.
English (UK): https://skinflint.co.uk/
German (AT,DE,CH): https://geizhals.at
So for this particular SSD: https://skinflint.co.uk/kingston-dc...ssd-0-5dwpd-480gb-sedc500r-480g-a2014002.html
If someone knows other websites that offer such nice and detailed filtering options and have the specs in the same format, please do let me know!
Thanks all for the inputs, no wonder the prices looked suss
I'm probably going for Intel 240G D3-S4510 ($25) or Intel 480G D3-S4510 ($55) instead
The Price tag
Seriously, the issue to understand with the SSD/NVMe, is the design for read vs write vs endurance and the use cases required.
"Enterprise" then typically becomes something that is designed for "Database & high file I/O" ie. high write percentage use cases writing typically a drive's capacity in the course of a day.
The Pro-sumer then becomes the "high burst write" versions that doesn't have a high write % percentage (Like gaming and photo/video editing where you need the high bursty writes, but mostly do mostly reads percentage wise)
The Consumer drives are meant for the cases where the user downloads emails/etc. and rarely writes over the course of the day, ie. you don't write more than like 10% of the drives capacity in a day for the "warranty period"
Yes, I'd include the PLP side of things for reliability reasons, not for the other classifications as much, but then that's what so nice of definitions, pick the one YOU like!
Well, tell that to everyone on this forum calling for the use of "enterprise SSD" with ZFS.The Price tag
Seriously, the issue to understand with the SSD/NVMe, is the design for read vs write vs endurance and the use cases required.
"Enterprise" then typically becomes something that is designed for "Database & high file I/O" ie. high write percentage use cases writing typically a drive's capacity in the course of a day.
The Pro-sumer then becomes the "high burst write" versions that doesn't have a high write % percentage (Like gaming and photo/video editing where you need the high bursty writes, but mostly do mostly reads percentage wise)
The Consumer drives are meant for the cases where the user downloads emails/etc. and rarely writes over the course of the day, ie. you don't write more than like 10% of the drives capacity in a day for the "warranty period"
Yes, I'd include the PLP side of things for reliability reasons, not for the other classifications as much, but then that's what so nice of definitions, pick the one YOU like!
I only thought of this question after seeing the 28% remaining lifespan on my consumer SSDs.
Data integrity (PLP) and high endurance is what I expect from an enterprise SSD. Something like this:
Plus the 1 Mio IOPS is nice to have for sure. 
I guess I would have replaced a consumer SSD three times already...
Code:
SMART/Health Information (NVMe Log 0x02)
Critical Warning: 0x00
Temperature: 40 Celsius
Available Spare: 100%
Available Spare Threshold: 10%
Percentage Used: 1%
Data Units Read: 237,949,946 [121 TB]
Data Units Written: 660,525,316 [338 TB]
Host Read Commands: 4,426,033,297
Host Write Commands: 6,435,863,236
Controller Busy Time: 1,249
Power Cycles: 13
Power On Hours: 18,464
Unsafe Shutdowns: 3
Media and Data Integrity Errors: 0
Error Information Log Entries: 0
Warning Comp. Temperature Time: 0
Critical Comp. Temperature Time: 0
Temperature Sensor 1: 40 Celsius
Temperature Sensor 2: 38 Celsius
Temperature Sensor 3: 38 CelsiusI guess I would have replaced a consumer SSD three times already...
It's much more than that. A PLP drive will return ACK to the OS/app once that is written in the cache because PLP will allow that data to be written to the chips no mater what happens (OS panic, app crashes, power loss, hardware failure, etc). Of course, assuming that firmware isn't buggy.
That will allow a drive to coalesce writes, sending them in batches in a way that both reduces the amount of writes to the chips and does it faster than doing single writes. This is the real reason why an "enterprise drive" last much longer than a consumer one, not just the over provisioning of storage chips.
Imagine you have a DB, you constantly update a record of a table with a counter. Those little bytes get write amplified to at least 4k (typical sector size) per write op. As it's a db, those writes are sync writes: the DB will wait until the disk returns ACK. If you have PLP, the disk will return ACK as soon as the write data is in cache, then coalesce N write ops and send to the chips some of the writes with whatever the counter was at that op. If there's no PLP, every write must go to the chips, effectively overwriting data on the disk over an over just to keep the "last" value.
Multiply that with lots of VMs, different workloads, ZFS volblocksize, Ceph's rocksDB, etc, and you'll figure out why a consumer drive lacks endurance / performance for anything "production" (although I use them in my homelab and they are still fine).
It is about horses for courses.
I have ran consumer/pro-sumer SSDs for ~1year to get a business off the ground. Those would've lasted another 1.5-2years on the usage it had, but then again I expected that,the cost was manageable at that stage, while "Enterprise SSDs" was so totally out of the budget reach the project would've died before we even turned the first month!
SO yes, understand the tech more than the labels!
Do note the last sentence
Yes, I use(d) consumer SSDs like that for ZFS cache and ZIL/SLOG in that exact manner as the PLP to gain some performance boosts "needed"
You are right w.r.t. the coalescing/etc. but that also adds costs that is not always required... but when you are in that price/budget brackets, you will be happy for them... other times the performance vs budget vs reliability from an accountants perspective makes for the enterprise drives's costs to be prohibitive
Don't take me wrong, I'd like to have 8TB "Enterprise" SATA SSDs instead of 8TB HDDs in my storage servers, but the cost just doesn't make sense for the backup storage.... so I use "pro-sumer" NVMes on PCI adapters to front the HDDs to do the zil/slog stuff and caching the meta. Good enough performance, but the ~6TB DB that is on enterprise NVMes that does hammer the NVMes
I don't. I'd like to have multiple smaller SAS SSDs.
Unless you need to restore something fast of course
There is much more that makes an Enterprise SSD beyond the firmware and Power Loss Protection.
You can sustain the listed read/write IOPS, so little to no cache is used, unlike consumer SSD, depending on the amount of write you do at once, transactions are simply distributed over more chips. It is what makes DC SSD more expensive, just more chips (which translates in smaller chips so you don’t necessarily have to go to QLC but stay in TLC or use TLC as SLC), better firmware, faster controllers (allowing things like inline compression), better quality control, better monitoring, better cooling, easy warranty, more overprovisioning etc.
You can sustain the listed read/write IOPS, so little to no cache is used, unlike consumer SSD, depending on the amount of write you do at once, transactions are simply distributed over more chips. It is what makes DC SSD more expensive, just more chips (which translates in smaller chips so you don’t necessarily have to go to QLC but stay in TLC or use TLC as SLC), better firmware, faster controllers (allowing things like inline compression), better quality control, better monitoring, better cooling, easy warranty, more overprovisioning etc.
Last edited:
Glad you could pay for the Bugattis, I needed a take out a loan for the Toyotas ...Yeah, if the costs matter just calculate when my SSDs will be about to die. Unfortunately, I won't live to see it.
@hevisko: In reality, the desktop SSDs are a lot more like the Bugatti's, on paper they have great performance, in reality, you can only do that a few times before you get an expensive repair job.
What is the value of your data? That's why Desktop SSD are considered 'more expensive' in business. You won't see a Samsung EVO in any Dell or Apple hardware, they're all Kioxia, Intel, ... Apple's stuff is practically a datacenter SSD, even a $500 Dell OptiPlex has a relatively 'expensive' workstation SSD (that's why they're so small in capacity). Because if you go through warranty and replacement costs and bad reviews from your customers, that $50-100 savings suddenly becomes a $300 loss.
What is the value of your data? That's why Desktop SSD are considered 'more expensive' in business. You won't see a Samsung EVO in any Dell or Apple hardware, they're all Kioxia, Intel, ... Apple's stuff is practically a datacenter SSD, even a $500 Dell OptiPlex has a relatively 'expensive' workstation SSD (that's why they're so small in capacity). Because if you go through warranty and replacement costs and bad reviews from your customers, that $50-100 savings suddenly becomes a $300 loss.