Question Are these 4TB SSDs fine and reliable enough for a Raid 10 array in a NAS

[Wolverine2349]

I am trying to get 8TB capacity in RAID 10 with some fault tolerance and better performance to get closer to utilizing 10GBe network locally on laptop and desktop whcih have 10Gbe connections coming.



I want to try to keep budget as low as possible and it adds up fast with SSD drives.



https://www.newegg.com/p/pl?N=100011693 600545605 600038519&Order=1



That link above should give list of SATA SSDs 4TB sorted by lowest price.



Are the KingSpec, Silicon Power, and TeamGroup ones fine and reliable for $143 and lower, or are they cheap for a reason meaning they have bad reliability compared to Samsung, WD, Crucial and SanDisk and others?



Or are they just cheaper for having less IOPS and sustained and peak MB/s read and write performance than the name brands, but still should have fine reliability?



Not concerned about less performance as the performance will still be tons multiple times better than the best HDD and buying 4 of them for RAID 10 I want a good deal. Unless performance tanks so much that you get way under 350MB/s which is unlikely as it seems most SSDs even low end ones are often in the 400s MB/s and high end ones upper 500s with much better IOPs

 

[zir_blazer]

I didn't checked these SSDs specs, but avoid bottom of the barrel drives with no DRAM because those SSDs can actually stall at almost 0 Bytes for multiple seconds, making them just adequated for cold storage drivers.

 

[OlyAR15]

I have a mixture of the Sandisk and WD Blue SSDs in my NAS. So far no issues. I’d like to eventually replace them with Samsung 870QVO 8tb drives when the prices get low enough.

 

[mxnerd]

The reality is that you can't trust any vendors, even ones with "good" reputation.

And RAID is not backup, it's only for up time, it's been told many times and you can find tons of articles explaining why it's not backup.
You still need backup your raid 10 using other methods/media

 

[Wolverine2349]

The reality is that you can't trust any vendors, even ones with "good" reputation.

And RAID is not backup, it's only for up time, it's been told many times and you can find tons of articles explaining why it's not backup.
You still need backup your raid 10 using other methods/media

yes I plan on doing a regular backup as well.

Though is RAID 0 just as safe as long as you do backups with the only risk being more down time as 1 drive failure ruins it all? Or are there any other risks of RAID 0 because RAID 0, I could go more HDDs and get faster speeds for much cheaper.

 

[mxnerd]

Although I did say you can't trust any vendors, I'll still go with at least Crucial MX500 or Samsung EVO 870 with 5 years warranty. Both brands use their own chips.
There is a reason for Cheap SSD with only 3 year warranty. They usually have worse chips and as mentioned by others, they usually don't have dram. When you are copying a lot of files or huge files, the speed can drop a lot.

Yes, you can do raid 0 as long as you have backups. I'll buy at lest one 8TB HDD to backup your raid 0. Or you can do 3 disk raid 5 and also back it up to 8TB HDD.

 

[Wolverine2349]

Well How about a middle ground?

I heard PNY drives are budget, but are they budget reliable just less performing than the Samsung Evo 870 and Crucial MX500.

I know PNY is a well known brand

Like there is this:

Amazon.com

Is that a reliable drive that still would curb stomp any HDDs even if its performance is definitely less than the Crucial MX500 and Samsung drives?

It costs $30 more than the cheap no name Silicon Power and KingSpec and such. But is $41 less than high end Crucial MX500.

Good compromise or still a knock off that is taking a huge risk??

All the ones priced $200 or higher seem to be high end SATA SSDs per the Google SSD spreadsheet.

 

[In2Photos]

Well How about a middle ground?

I heard PNY drives are budget, but are they budget reliable just less performing than the Samsung Evo 870 and Crucial MX500.

I know PNY is a well known brand

Like there is this:

Amazon.com

Is that a reliable drive that still would curb stomp any HDDs even if its performance is definitely less than the Crucial MX500 and Samsung drives?

It costs $30 more than the cheap no name Silicon Power and KingSpec and such. But is $41 less than high end Crucial MX500.

Good compromise or still a knock off that is taking a huge risk??

All the ones priced $200 or higher seem to be high end SATA SSDs per the Google SSD spreadsheet.

I've used the PNY CS900 on several laptops (240GB and 500GB models) to replace the original spinning HDD. All of them are still in use, one is approaching 4 years old. I know one of them is used daily. I've also had one Crucial MX500. That drive failed after a few months and was replaced under warranty. I still believe Crucial makes a good drive, that's just the only experience I happen to have had with it.

 

[Wolverine2349]

I've used the PNY CS900 on several laptops (240GB and 500GB models) to replace the original spinning HDD. All of them are still in use, one is approaching 4 years old. I know one of them is used daily. I've also had one Crucial MX500. That drive failed after a few months and was replaced under warranty. I still believe Crucial makes a good drive, that's just the only experience I happen to have had with it.

Thanks very much for your reply as that is helpful.

Would you say PNY is much more reliable and trustworthy than Silicon Power, Team Group, and KingSpec, Leven, or Fanxiang? Have you had any experience with these brands that have prices of $149 or lower for 4TB SATA and maybe they are that cheap and knockoff for a reason.

 

[In2Photos]

Thanks very much for your reply as that is helpful.

Would you say PNY is much more reliable and trustworthy than Silicon Power, Team Group, and KingSpec, Leven, or Fanxiang? Have you had any experience with these brands that have prices of $149 or lower for 4TB SATA and maybe they are that cheap and knockoff for a reason.

I haven't used any of those other brands. Team Group and Silicon Power seem to be brands that get used in budget builds from popular YouTube PC channels. But you never hear any long term reviews on them.

 

[Wolverine2349]

I haven't used any of those other brands. Team Group and Silicon Power seem to be brands that get used in budget builds from popular YouTube PC channels. But you never hear any long term reviews on them.

How about PNY., I know you stated you had no problems with yours, but are they generally used in budget build or more mid range builds?

 

[In2Photos]

How about PNY., I know you stated you had no problems with yours, but are they generally used in budget build or more mid range builds?

I've honestly not seen PNY used in most builds on YouTube. But I think some of that is because they don't go after influencers like Crucial does. the MX500 has almost 100,000 reviews with 4.8 stars. The CS900 has almost 20,000 with 4.7. Some of the others you listed also have good reviews. I think in this case I would just trust the more well known companies.

 

[Wolverine2349]

I've honestly not seen PNY used in most builds on YouTube. But I think some of that is because they don't go after influencers like Crucial does. the MX500 has almost 100,000 reviews with 4.8 stars. The CS900 has almost 20,000 with 4.7. Some of the others you listed also have good reviews. I think in this case I would just trust the more well known companies.

So in that case would you say PNY not well known cause not used in most builds on YouTube?

Which would you consider most well known companies. PNY and Crucial and Samsung , WD, Sandisk,

Where as the others I listed are not near well known enough?

 

[In2Photos]

So in that case would you say PNY not well known cause not used in most builds on YouTube?

Which would you consider most well known companies. PNY and Crucial and Samsung , WD, Sandisk,

Where as the others I listed are not near well known enough?

PNY has been around a long time. They just don't seem to go after the influencers so they don't appear in videos, but they are still well known in the industry.

 

[aigomorla]

Unless performance tanks so much that you get way under 350MB/s which is unlikely as it seems most SSDs even low end ones are often in the 400s MB/s and high end ones upper 500s with much better IOPs

Raid10 -> Raid-Z2 over 10gbe.... on HGST He drives...

Point being you don't need SSD's for this kind of speed...

Also i do not really trust consumer grade SSD's for a NAS, unless your doing it for archival reasons, and you will not have many writes. Ie.. a movie dump drive, or some kind of Media Vault, where you move data into it once, and then read many many times.

Also you will capped at the controller end.
SATA6G can only go so fast, and you need a dedicated controller for SATA12G.

 

[Wolverine2349]

Raid10 -> Raid-Z2 over 10gbe.... on HGST He drives...

Point being you don't need SSD's for this kind of speed...

Also i do not really trust consumer grade SSD's for a NAS, unless your doing it for archival reasons, and you will not have many writes. Ie.. a movie dump drive, or some kind of Media Vault, where you move data into it once, and then read many many times.

Also you will capped at the controller end.
SATA6G can only go so fast, and you need a dedicated controller for SATA12G.

Well I was looking at controller cards and would have needed an M.2 PCIe one if indeed I was to make using lower capacity drives an option with expansion in the future. Instead using 2TB drives 4 of them in RAID5 instead oif 4X4TB in RAID 10 which is much more expensive.

But if I wanted more SATA ports:

. And all controller cards have only PCIe Gen 3 X2 not the full X4 M.2 slots can use so max out at 2GB/s which could only deal with not even 4 SATA ports

But doesn't each port on the motherboard get a dedicated 6Gbits per second bandwidth as lung as the chipset has enough which Intel LGA 1700 chipsets do for 4 SATA ports unless the chipset is also capable of handling it?

https://b2c-contenthub.com/wp-content/uploads/2022/01/Intel-Alder-Lake-chipsets.png?w=1200

I mean H610 chipset provides 8 PCIe 3.0 lanes and don't all 4 on board SATA ports on those motherboards have full access to that though it is shared with other chipset things but 4 SATA ports is 2400MB/s (600MB/s per port X 4) and there is still another 5600 MB/s as PCIe Gen 3 is 1GB per lane and 8 lanes is 8GB/s.

Every controller card I have looked at like this: https://www.amazon.com/Trotwei-Blue...ords=m2+to+sata+adapter&qid=1697251527&sr=8-6

Maxes out at 2GB/S which is not even enough for 4 ports let alone 5 or 6 as they are only PCIE Gen 3 X2 AsMedia 1166 maxes out at that. Where as on board SATA ports have access to 4X as much bandwidth even though its shared with other stuff as most other things are off and not using it.

 

[aigomorla]

. And all controller cards have only PCIe Gen 3 X2 not the full X4 M.2 slots can use so max out at 2GB/s which could only deal with not even 4 SATA ports

well 1 gen3-4 nvme will stil lbe faster then 4 x 4TB in R10.
I have 4x4TB in Raid-0 for my gaming drive and the nVME i have is still faster because of the SATA6G limitation.

Also this is the fastest ive gotten on 10gbe.

This was done on 2 RAM drives.... both the the destination and source.
And well RAM drives are ridiculously fast, its its volatile storage.

I think iperf will probably say i can go a bit faster if i did use it, but in real world testing, i think 1.57GB/s will be the realistic cap, which is still below a nVME transfer rate.

Hence i think it might be better for you to go a single 8TB nVME over 4 x 4TB Sata SSD's in R0 or better yet if your board supports bifurication, then possibly 2 x 2TB nVME's and bifur it.

Eitherway you will still hit that 1.57GB/s wall on 10gbe unless you go bat crazy like @Fallen Kell and go 40GB/e like he has, maybe he can tell you what speeds he has seen.

I don't even know what speeds you get there, but even to test it, it would be majorly expensive.

 

[Wolverine2349]

well 1 gen3-4 nvme will stil lbe faster then 4 x 4TB in R10.
I have 4x4TB in Raid-0 for my gaming drive and the nVME i have is still faster because of the SATA6G limitation.

Also this is the fastest ive gotten on 10gbe.

This was done on 2 RAM drives.... both the the destination and source.
And well RAM drives are ridiculously fast, its its volatile storage.

I think iperf will probably say i can go a bit faster if i did use it, but in real world testing, i think 1.57GB/s will be the realistic cap, which is still below a nVME transfer rate.

Hence i think it might be better for you to go a single 8TB nVME over 4 x 4TB Sata SSD's in R0 or better yet if your board supports bifurication, then possibly 2 x 2TB nVME's and bifur it.

Eitherway you will still hit that 1.57GB/s wall on 10gbe unless you go bat crazy like @Fallen Kell and go 40GB/e like he has.
I don't know what speeds you get there, but even to test it, it would be majorly expensive.

Yes of course NVME will be much faster evne single drive.

Issue is no data redundancy. And 10Gbe would be wasted with NVME as it goes faster than 10Gbe could do, so that is kind of why I liked 4 SATA RAID 10 or now doing RAID 5 as it will get close to 10Gbe saturation but not quite. Rather than waste it and have to go to 20Gbe or more which is so expensive where as 10Gbe can be somewhat affordable now.

How were you able to get 1.57GB/s on 10Gbe. That seems much faster than it could max out at.

10Gbe is 10000Mbits per second and 10000/8 is 1250 so maximum would be 1250 Megabytes per second. And in reality overhead so they say divide it by 10 so it would be 1000MB/s or 1GBs for 10Gbe one way.

I know its full duplex but copying a file is a one way thing so how could it go faster than 1250MB/s peak.

 

[Shmee]

Are you planning on doing hardware RAID, or software with ZFS on a Truenas or similar? Consumer SSDs should be fine IMO, at least in groups of mirrors, as long as they are the good ones and not lower tier drives (DRAMless or using QLC).

On my Truenas I have a pool of 2 mirrors, each mirror is 2x2TB Crucial MX500. These are connected to an LSI card of some sort I think, flashed to IT mode for use with ZFS. Got it from a member here in FS/T.

 

[Hail The Brain Slug]

Yes of course NVME will be much faster evne single drive.

Issue is no data redundancy. And 10Gbe would be wasted with NVME as it goes faster than 10Gbe could do, so that is kind of why I liked 4 SATA RAID 10 or now doing RAID 5 as it will get close to 10Gbe saturation but not quite. Rather than waste it and have to go to 20Gbe or more which is so expensive where as 10Gbe can be somewhat affordable now.

How were you able to get 1.57GB/s on 10Gbe. That seems much faster than it could max out at.

10Gbe is 10000Mbits per second and 10000/8 is 1250 so maximum would be 1250 Megabytes per second. And in reality overhead so they say divide it by 10 so it would be 1000MB/s or 1GBs for 10Gbe one way.

I know its full duplex but copying a file is a one way thing so how could it go faster than 1250MB/s peak.

I think the issue here is windows GUI file transfer lies.

 

[aigomorla]

10Gbe is 10000Mbits per second and 10000/8 is 1250 so maximum would be 1250 Megabytes per second. And in reality overhead so they say divide it by 10 so it would be 1000MB/s or 1GBs for 10Gbe one way.

I know its full duplex but copying a file is a one way thing so how could it go faster than 1250MB/s peak.

well as i said its theoretical.... and it was a direct peer 2 peer meaning no switch overhead.
Realistic, i showed at arround 500-600MB/s.

But yeah, don't expect to ever get that high, especially over a switch.

 

[Hail The Brain Slug]

well as i said its theoretical.... and it was a direct peer 2 peer meaning no switch overhead.
Realistic, i showed at arround 500-600MB/s.

But yeah, don't expect to ever get that high, especially over a switch.

...... The theoretical throughput is the 1250MiB/s that was mentioned before. 1.57GB/s is literally impossible across a 10GbE connection, no matter what your setup is. Windows GUI file transfer gives all kinds of wrong readouts. It can't be trusted.

And your direct peer-to-peer connection isn't any faster than a competent 10GbE switch inbetween.

 

[Wolverine2349]

Are you planning on doing hardware RAID, or software with ZFS on a Truenas or similar? Consumer SSDs should be fine IMO, at least in groups of mirrors, as long as they are the good ones and not lower tier drives (DRAMless or using QLC).

On my Truenas I have a pool of 2 mirrors, each mirror is 2x2TB Crucial MX500. These are connected to an LSI card of some sort I think, flashed to IT mode for use with ZFS. Got it from a member here in FS/T.

Not planning any hardware RAID. Just software RAID.

Can software RAID cross 2 controllers if I need to add ports.

How many SATA SSDs will be able to saturate 10Gbe in RAID 5 or 6 without being too much faster than the 10Gbe connection so its not a waste.

I mean do 2 SATA SSDs in RAID0 saturate 10Gbe already?

RAID 5 and 6 add overhead but are safer. RAID 1 is redundant but wasteful.

RAID 10 can speed things up but also somewhat wasteful.

 

[Shmee]

The answer probably depends on the controllers and software/OS used. In general though I would think you could add drives across different controllers to the same software volume. As for saturating the 10Gbe link, again it depends, but 2 quality SATA SSDs in a striped config should get pretty close I would think. Performance in general is going to depend on various factors: system hardware, OS, software/RAID level used, and what the system is transferring to on the network.

 

[aigomorla]

Windows GUI file transfer gives all kinds of wrong readouts. It can't be trusted.

And your direct peer-to-peer connection isn't any faster than a competent 10GbE switch inbetween.

I completely agree windows can't be trusted, but its a funny number that i got when testing a direct peer 2 peer.

And the last part of the second comment also applies... a competent 10gbE switch is typically a enterprise class. I didn't start using those until Mikrotik came out with the quiet SFP+ switches which run passively, because as you know the good ones are ridiculously loud.

Also i think it might also have to do with the fact i was running 2 DAC's cable and not optical transceivers, so it was a direct connect without any overhead even from transceivers.

But im also in agreement that its most probably windows playing funny with transfer rates.