interesting read - SSD reliability
- Thread starter cesmith9999
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
And here is the paper rather than ZDnet...
My summery ...
Without listing what drives were used, what process nand and who made it this data is utterly useless.
They have literally grouped together all the SLC drives all the MLC drives and all the MLCe drives.
The biggest risk with consumer drives is data loss on hard power loss due to the lack of plp.
The data retention tests done by techreport have better use.
My summery ...
Without listing what drives were used, what process nand and who made it this data is utterly useless.
They have literally grouped together all the SLC drives all the MLC drives and all the MLCe drives.
The biggest risk with consumer drives is data loss on hard power loss due to the lack of plp.
The data retention tests done by techreport have better use.
Last edited:
Why is that surprising?
No more reliable is not the same as No more endurance
SLC Chip = MLC Chip I believe -- right? One just stores more, and has less endurance.
So it would mkae sense drive for drive that they are the same
well then i interpreted that wrong, i read: "Is just as reliable, similiar if not the same endurance, more or less the same in practice as"
You can use MLC die in 'SLC'mode', that is each cell can store 4 levels (2 bits) but you can choose to program only 2 levels (1 bit). I believe SLC flash is still different in regards to the other hardware on the die (the read/program/erase control circuitry). In the end, SLC should be quicker to program and erase (though I suspect intimate knowledge of the underlying structures for a MLC flash used in 'SLC mode' can get them to program/erase quicker than SLC flash).
As I mentioned, I didn't read the entire thing. Is your quote what you read or what you 'said to yourself' in your head? Not sure I'm interpreting right myself
Thanks!The statement by itself is of no use. Let say SSDs of both types fail because their power supplies give out. Sure, you could say they have equivalent reliability, but that doesn't mean SLC is really no better than MLC flash. It just means some other component present on both on the MLC and SLC SSD fails before the flash can. So in a very limited case you could say that the type of flash didn't impact reliability, but that doesn't mean you can apply that statement to a broader scope.
Since no manufacturers or models are referenced it basically means nothing.
what i posted came from the article - the author, i guess, came to that conclusion as the actual paper that Google did wasn't linked as far as i can tell.As I mentioned, I didn't read the entire thing. Is your quote what you read or what you 'said to yourself' in your head? Not sure I'm interpreting right myself
Gotta read the paper closely...
Regarding the bit about SLC being no more reliable than MLC, they also said "within typical drive lifetimes" - an SLC drive has more endurance and a higher expected lifetime under a high write workload, and so the statement isn't really that surprising. It's really just saying that regardless of NAND type, SSDs are lasting as long as they are supposed to.
And yes - the NAND chips used for SLC/MLC/TLC are the same, every cell stores a single analog value. The differences are in how the controller evaluates that analog value, and whether it can differentiate between 2 values (SLC), 4 values (2-bit MLC), or 8 values (TLC). Taking the same analog range and splitting it more ways makes things harder, which is why MLC/TLC is slower and has lower endurance. Well - thats probably not the most technically correct way to put it, but is generally true - go google it yourself if you want really technical descriptions.
As to the way they grouped up the different kinds of drives - keep this statement from the paper in mind:
"The drives in our study are custom designed high performance solid state drives, which are based on commodity flash chips, but use a custom PCIe interface, firmware and driver." So the only thing these have in common with the drives we see are the NAND chips. They do have the details on the process used for the NAND, but the manufacturer data has been cleansed down to 'vendor 1', 'vendor 2', etc.
Regarding the bit about SLC being no more reliable than MLC, they also said "within typical drive lifetimes" - an SLC drive has more endurance and a higher expected lifetime under a high write workload, and so the statement isn't really that surprising. It's really just saying that regardless of NAND type, SSDs are lasting as long as they are supposed to.
And yes - the NAND chips used for SLC/MLC/TLC are the same, every cell stores a single analog value. The differences are in how the controller evaluates that analog value, and whether it can differentiate between 2 values (SLC), 4 values (2-bit MLC), or 8 values (TLC). Taking the same analog range and splitting it more ways makes things harder, which is why MLC/TLC is slower and has lower endurance. Well - thats probably not the most technically correct way to put it, but is generally true - go google it yourself if you want really technical descriptions.
As to the way they grouped up the different kinds of drives - keep this statement from the paper in mind:
"The drives in our study are custom designed high performance solid state drives, which are based on commodity flash chips, but use a custom PCIe interface, firmware and driver." So the only thing these have in common with the drives we see are the NAND chips. They do have the details on the process used for the NAND, but the manufacturer data has been cleansed down to 'vendor 1', 'vendor 2', etc.