SSD Reliability and Other Questions

[Vesa]

TL;DR:

Hello World.

• Is there anywhere online that allows one to sort SSDs by warranty limitations?
• Is there an SSD with both a similar warranty and a similar price point to the Samsung 850 Pro series? (10 Years or 150TBW)
• Is there a drive with a lengthy warranty that comes in capacities below 128 GB?
• Is there a highly reliable drive in the 20GB-64GB space?
• Is there a site that can filter by or sort by warranty?
• Is there a better approach to redundancy than RAID in a Windows (7) system?
• Is there a better approach than a UPS to combat the issue of sudden data loss in an SSD?

First(ly), Salutations to All.


I have recently been tasked with replacing the storage drives in 3 circa 2010 desktop systems. One of these acts as a server for a database, and the other two run a windows-based PoS or entry software IIRC. All machines are Windows 7. To date, I have not installed an SSD in a business environment wherein reliability matters. I have reviewed several articles on the status quo of solid state drives related to price trends and reliability. Most notably, the extremetech articles which are referenced at [H] in several threads in this sub-forum. Specifically, [1] which references the study [2] and [3] which references the study [4].

Given this information, it appears that most any SSD will be reliable well past the limits of the data that will be written to and read from them in this use case. The DB on the "server" is less than 20 MB IIRC. The entry machines mostly act as the front end of the software that access the db and are occasionally used for printing reports.

I have decided on the Samsung 850 Pro for the server (and installed it [and wow]) as it has the best warranty that I could find at 10 Years or 150TBW barring Enterprise class products (and enterprise class prices). However, the smallest storage space offered by the Samsung Pro series, 128 GB, is overkill for these tasks, which would be well served by 32GB or 64GB drives.

*Is there a site that allows one to filter search results by or sort by warranty length? Wading through warranty information has been the most painstaking part of the process.

*Is there a similarly warrantied drive that is smaller in capacity, similar in performance to, similar in reliability to, or that would fit this purpose well that I may have overlooked?

*Is there another SSD that you would recommend? I feel like money is being wasted on the capacity.

Redundancy for the "server" would be nice. The OS will be Windows 7. The current plan is to have an offsite backup of the database only as well as a local backup to a usb drive each day (with Cobian or Windows Backup and Restore). Ideally, what I would like to achieve is a system that would require minimal maintenance and service such that a drive replacement/reloading could be performed in my absence, preferably without the end user having to install or swap drives. I have considered RAID, and it appears that it would not be the best or even a decent solution. The first problem with RAID is the physical size of the card, all machines are slim. The second problem discovered during my research is that RAID with SSDs causes them to be more prone with failure. Does anyone have any experience with this or a similar setup? Or any counter arguments that I may not have considered to sway my conviction?

What system or method would you recommend for backing up or ensuring redundancy in the event of a drive failure? If more information is required I am more than happy to provide it.

The total-loss potential in the event of sudden power loss appears to be negated with a good UPS. Is there a better approach, or am I miguided?



References:

[1]
"How long do modern consumer SSDs actually last? Longer than you'd expect!"
http://www.extremetech.com/computin...er-ssds-actually-last-longer-than-youd-expect

[2]
"The SSD Endurance Experiment: Casualties on the way to a petabyte"
http://techreport.com/review/26523/the-ssd-endurance-experiment-casualties-on-the-way-to-a-petabyte

[3]
"The mysteriously disappearing drive: Are power outages killing your SSDs?",
http://www.extremetech.com/computin...ing-drive-are-power-outages-killing-your-ssds

[4]
"Understanding the Robustness of SSDs under Power Fault"
https://www.usenix.org/system/files/conference/fast13/fast13-final80.pdf


Sorry for wall of text and thank you for your response.

 

[wtourist]

The less storage (and greater size) on an SSD, the longer they are likely to last.
If you are willing to spend enough for 10 year warranty SSDs, They will outlast your hardware.
In your case, given the normal costs of 32 and 64 GB drives, 128 is the way to go.
With Samsung 840 EVO or Crucial MX100 you should be fine, to cut costs.
Most if not all 32/64 GB drives are old tech. I see 32 GB $34-45 from 2 generations back, 64 GB are from $55-80. (Toshiba has a 32 GB cache only drive for $30 or so).

I believe Sandisk does or is going to offer a 10 year warranty drive, as is/does Intel

5 year warranty
http://www.amazon.com/Intel-Solid-S...id=1409381512&sr=8-2&keywords=intel+120gb+ssd

850 PRO 120 GB is $120 on amazon, probably worth it over $70-80 drives.
.

 

[drescherjm]

Is there a highly reliable drive in the 20GB-64GB space?

I would spend a little over $100 US and get a Crucial MX100 256GB SSD which would have a much longer lifespan because it has many more cells to wear level. Bottom line is small drives are not great performers (either use older slower controllers and 2 year old nand or very few current nand chips in raid) and do not have the same write endurance of a ~256GB or larger drive.

 

[Concentric]

... However, the smallest storage space offered by the Samsung Pro series, 128 GB, is overkill for these tasks, which would be well served by 32GB or 64GB drives. ...

This is not an issue; 128GB is not a lot of storage. If anything, over-provisioning your storage capacity is a good thing, especially with SSDs. I think there are some Intel 3500 or 3700 drives around the 80GB or 100GB capacity points, if you really want something smaller.

From what you've said this sounds like a case where you should consider a whole new system from a tier 1 vendor, with a warranty, support etc? This is a business system at the end of the day. They can set up the RAID etc for you as standard and provide a replacement (perhaps on site service) if something fails.

 

[Nenu]

[*]Is there an SSD with both a similar warranty and a similar price point to the Samsung 850 Pro series? (10 Years or 150TBW)

No.
The new methods employed by Samsung have no competition.

 

[Vesa]

Thanks to all for replies.

The less storage (and greater size) on an SSD, the longer they are likely to last.
If you are willing to spend enough for 10 year warranty SSDs, They will outlast your hardware.
In your case, given the normal costs of 32 and 64 GB drives, 128 is the way to go.
With Samsung 840 EVO or Crucial MX100 you should be fine, to cut costs.
Most if not all 32/64 GB drives are old tech. I see 32 GB $34-45 from 2 generations back, 64 GB are from $55-80. (Toshiba has a 32 GB cache only drive for $30 or so).

I believe Sandisk does or is going to offer a 10 year warranty drive, as is/does Intel

5 year warranty
http://www.amazon.com/Intel-Solid-S...id=1409381512&sr=8-2&keywords=intel+120gb+ssd

850 PRO 120 GB is $120 on amazon, probably worth it over $70-80 drives.
.

The drive/storage outlasting the hardware is completely acceptable in this case, and how I got this recommendation. They were previously burned by data loss. Almost a month of data was lost and had to be reconstructed/rebuilt from books/accounting records. Reliable drives have been been requested as the prior loss cost them many man hours and much grief. Had the data been safe, they could have migrated the data/drive to a new machine or could have used a SATA->USB adapter or enclosure. The argument is, "Data isn't replaceable, but the system is."


Of the technologies it seems to be that:

MLC has better performance, but is more expensive.
TLC doesn't perform as well, but is cheaper.
V-NAND doesn't seem comparable to nor compared to the older technology. According to what I have read, V-NAND is a better technology overall. It appears more reliable and the drives are faster.

I completely looked over the 530.
The Intel 530 is currently ~$80, which is definitely cheaper than the $120 850 Pro. The difference is in IOPS.

• The random read on the Samsung is 100K, the Intel is 41K
  
• The random write on the 850 is 100K, the Intel is 90K

Do these numbers translate to a real world difference in anything other than benchmarking?

I wasn't aware it had been a while for a refresh on the lower capacity drives or that they used slower controllers. Using larger drives saves a potential headache and huge performance hit according to the numbers I'm seeing now.

I would spend a little over $100 US and get a Crucial MX100 256GB SSD which would have a much longer lifespan because it has many more cells to wear level. Bottom line is small drives are not great performers (either use older slower controllers and 2 year old nand or very few current nand chips in raid) and do not have the same write endurance of a ~256GB or larger drive.

Does the wear level on a larger drive make a huge difference over time? I assume the wear-leveling tech in SSDs keeps them from wearing the same sections of the drive over and over, but does this spread out the density and keep it low given small data and a large drive as well? E.g. the same 32GB won't be written and read in the same place on the 256 GB drive, even though it's a very small quantity compared to the size of the drive? I thought I understood it before, but I am confused as the 850 Pro has a 150TBW or 10 years, whichever is less, warranty on all the drives regardless of size.

I wasn't aware the smaller SSDs used slower and/or older controllers.

This is not an issue; 128GB is not a lot of storage. If anything, over-provisioning your storage capacity is a good thing, especially with SSDs. I think there are some Intel 3500 or 3700 drives around the 80GB or 100GB capacity points, if you really want something smaller.

From what you've said this sounds like a case where you should consider a whole new system from a tier 1 vendor, with a warranty, support etc? This is a business system at the end of the day. They can set up the RAID etc for you as standard and provide a replacement (perhaps on site service) if something fails.

From above, considering wear-leveling and Capacity vs Storage:

Does the wear level on a larger drive make a huge difference over time? I assume the wear-leveling tech in SSDs keeps them from wearing the same sections of the drive over and over, but does this spread out the density and keep it low given small data and a large drive as well? E.g. the same 32GB won't be written and read in the same place on the 256 GB drive, even though it's a very small quantity compared to the size of the drive? I thought I understood it before, but I am confused as the 850 Pro has a 150TBW or 10 years, whichever is less, warranty on all the drives regardless of size.

I would be more than happy to get them in contact with a vendor, however they want someone they know they can trust as they felt the last deal was, "an absolute nightmare" because of the slow response. Due to location they usually had a 48 hour wait when something went wrong. Do you have any suggestions? If this isn't allowed on the open forum, feel free to pm me please.

No.
The new methods employed by Samsung have no competition.

I was afraid of this.

 

[JoeComp]

They were previously burned by data loss. Almost a month of data was lost and had to be reconstructed/rebuilt from books/accounting records.

WTF?

Where were the backups? If data is important, there should be at least two backups, at least one of which is off site.

 

[Ocellaris]

Any drive can die at any time regardless of the technology or warranty period. You need a properly structured backup plan to mitigate lost data when drives die.

 

[Aesma]

Don't even look at performance for the tasks you mention, any SSD will be fine. Now talking about 32GB being enough with Windows 7, we must not be talking about the same OS !

So like the other I say 128/256 GB drives are the way to go.

And yes RAID1 is your best bet for availability.

Then a hot network backup of some sort, hourly.

Then a cloud backup and/or a cold backup, preferably off-site.

 

[drescherjm]

E.g. the same 32GB won't be written and read in the same place on the 256 GB drive, even though it's a very small quantity compared to the size of the drive?

The drive will spread the wear over the entire drive (+ the hidden extra cells that each drive has). It does not read and write in the same place. It maps each block using a mapping table so that to the OS it looks like the data is sequential but as far as the NAND goes the wear is distributed over the cells. So with a larger SSD that has 8 times the cells you get many more cells to wear level.

 

[drescherjm]

Any drive can die at any time regardless of the technology or warranty period. You need a properly structured backup plan to mitigate lost data when drives die.

Fully agreed. Regardless of the technology any drive can die at any time at all. Always have a backup (or better yet have several possibly offsite) if you have data that can not be lost.

 

[Vesa]

WTF?

Where were the backups? If data is important, there should be at least two backups, at least one of which is off site.

I wholly agree. They got burned by their vendor. Their backup stopped running and they got tired of calling the vendor to fix things. In their version of the story it wasn't uncommon to have the vendor on-site at least once every month or two.

Any drive can die at any time regardless of the technology or warranty period. You need a properly structured backup plan to mitigate lost data when drives die.

I understand this. My goal is to ensure the hardware is stable and that the data does not become lost again. I should post another thread or modify my post to make it more clear that I am open for suggestions regarding redundancy and backup systems. Also, I need to ask what setups people have had success with for SSD redundancy.

Don't even look at performance for the tasks you mention, any SSD will be fine. Now talking about 32GB being enough with Windows 7, we must not be talking about the same OS !

So like the other I say 128/256 GB drives are the way to go.

And yes RAID1 is your best bet for availability.

Then a hot network backup of some sort, hourly.

Then a cloud backup and/or a cold backup, preferably off-site.

Your thoughts are valued. I initially assessed 32 GB as being enough space with a little wiggle room as the setup is Windows 7 with one program installed and a very small DB to host. In retrospect, I do see your point.

I think I am going to order one more 128GB Samsung 850 Pro for one of the entry machines/workstations. As for the final drive, I'm stuck between a rock and a hard place with the 128GB Intel 530 at $80 and the 256 GB Crucial MX100 at $110. Any insight as to how this difference in capacity would play out in reality with regard to wear-level?

It appears that the effectiveness of RAID 1 for SSDs is a hot topic on many forums. Are you referring to hardware or software RAID? Windows 7 Professional, Enterprise, and Ultimate are all capable of software RAID. I don't have any experience here with SSDs and am unable to find very much material that addresses software RAID 1 and reliability and/or availability. Any experience or input here?

Do you have any input or any experience with any specific hot network backup solution?

 

[Vesa]

The drive will spread the wear over the entire drive (+ the hidden extra cells that each drive has). It does not read and write in the same place. It maps each block using a mapping table so that to the OS it looks like the data is sequential but as far as the NAND goes the wear is distributed over the cells. So with a larger SSD that has 8 times the cells you get many more cells to wear level.

This is what I have been looking for. I suspected, but wasn't positive. Thank you for providing clarity.

Fully agreed. Regardless of the technology any drive can die at any time at all. Always have a backup (or better yet have several possibly offsite) if you have data that can not be lost.

This is true unless it is a drive in an Apple computer. Parts on Macs never fail, nor do Macs get malware. Only stolen. Naturally, I'm trolling. This piece of advice is definitely the one I'm most concerned with trying to implement and drive home.

 

[Concentric]

... I would be more than happy to get them in contact with a vendor, however they want someone they know they can trust as they felt the last deal was, "an absolute nightmare" because of the slow response. Due to location they usually had a 48 hour wait when something went wrong. Do you have any suggestions? If this isn't allowed on the open forum, feel free to pm me please. ..

I'm sure you already know this, but the point I was making was that you can buy a new server from someone like Dell/HP/etc and get a support contract to fit your requirements - 4-hour call out, 24-hour, whatever, including replacing failed drives. They will provide your client with that level of service and you no longer need to worry about it.

It's very interesting from a computer enthusiast's point-of-view to look at the different components and talk about which is faster or more reliable or what-not. But for a business critical system you want to avoid all this hassle by just letting Dell/HP/whoever provide the parts with guaranteed compatibility.

I'm saying this because it sounds like your client has a "buy off-the-shelf components to fix it as cheaply as possible" approach, so I wanted to advise otherwise.

 

[Vesa]

I'm sure you already know this, but the point I was making was that you can buy a new server from someone like Dell/HP/etc and get a support contract to fit your requirements - 4-hour call out, 24-hour, whatever, including replacing failed drives. They will provide your client with that level of service and you no longer need to worry about it.

I do know that is possible, but I wasn't sure of what exactly you were suggesting in the statement. I'm apprehensive to depend on any specific vendor and leave a bad taste in their mouths. I have considered the option. Perhaps I could persuade them to purchase a new server.

It's very interesting from a computer enthusiast's point-of-view to look at the different components and talk about which is faster or more reliable or what-not. But for a business critical system you want to avoid all this hassle by just letting Dell/HP/whoever provide the parts with guaranteed compatibility.

Regarding the discussion itself, I do consider it fascinating seeing the many ways that we apply and create technology to achieve results and meet needs. From all the hassle comes the reward--billable hours! Jokes aside, it is a very solid approach to let someone else handle the h/w and compatibility support that can be an absolute proprietary pain.

I'm saying this because it sounds like your client has a "buy off-the-shelf components to fix it as cheaply as possible" approach, so I wanted to advise otherwise.

My client does have that mindset. Everyone wants to feel like they're getting a good deal, and they are. The Samsung 850 Pro may not be the greatest drive ever manufactured, but it is very capable of holding its own even if it's over-priced IMO. They have approached me requesting that I help them find a solution that works and provides them the least headache going forward.

At the end of the day, they want someone and something they can trust. In the past, they were taught to not trust people in our industry, and I want to help them restore that faith.


Please, do feel free to PM me.