Burying its head in the NAND: Samsung boosts 64-layer 3D flash chip production Samsung says it is boosting its 64-layer V-NAND flash chip production after Toshiba and WDC have introduced 64-layer NAND drives. Micron and flash foundry partner Intel also have 64-layer product coming. The Samsung V-NAND 3bits/cell (TLC) chips have 256Gbit capacity and are being punted at mobile, PC and server applications …
The idea that write cycles is a major limitation for flash, or that disk is somehow better in that regard is a badly outdated if not outright ignorant argument at this point and needs to die.
Disk has unlimited writes? What a joke! Take a 10TB spinning disk. What's it's expected life? About 5 years? Now hammer it with 100mb/sec of writes continuously. You'll fill it up once a day. What's the effective write cycles over 5 years? About 1900! That's it! Not "unlimited". Plenty of flash has more endurance than that!
"Take a 10TB spinning disk. What's it's expected life? About 5 years? Now hammer it with 100mb/sec of writes continuously"
Or hammer it with random access and it'll die quickly - mechanical drives effectively shake themselves to pieces under such loads.
The answer to how many write cycles an HDD is good for is clearly not infinite for the very good reason that they are mechanical devices and will, sooner or later, wear out. Hammer them with random I/Os with a very high duty cycle rate will accelerate that and can only be mitigated if you buy very expensive enterprise HDDs with reduced capacity. People should also not mistake MTBFs as expected lifetimes for HDDs - they are statistical failure rates during a devices expected lifetime, which is typically quite a lot shorter than the MTBF figure, especially if used intensively.
As with tapes, HDDs will gradually move to support the type of workloads where their performance characteristics and cost profiles make sense. The high latency and related limited access rates will gradually make them less viable for many workloads where those characteristics are important.
It's also worth noting that the controllers in modern SSDs are much more capable than earlier ones in the manner in which they deal with bad cells and write levelling.
No storage device, whether HDD, SSD or tape is forever. Personally, not wasting hours a week for the cumulative effect of poor HDD performance (on personal devices and back end systems) is a considerable cost in itself and, increasingly, large businesses are appreciating that too due to productivity issues even before customer service considerations are taken into account.
Flash storage has reached the point where it is "good enough" and "cheap enough" that it makes sense for a lot of individuals and companies, at least for I/O intensive parts of workloads if not backup or archival.
Performance of SATA 3 SSDs peaked about 5 years ago and all we've seen is a capacity increase and (seemingly slowing down) price drop since then. As far I'm concerned, this makes SATA 3 SSD "old, boring and slow" tech and it's the PCIe SSDs where you'd hope Samsung and others would ramp up production on.
Yes, I know older motherboards don't have M.2 slots, but that's what an adapter is for - we need PCIe SSDs to get close to SATA 3 SSD prices (they've still got a premium that puts people off, despite being up to 6 times faster than SATA 3 SSDs giving them a massively better "performance per buck") and then people's head lightbulbs will illuminate and they'll finally realise that PCIe SSDs are the way to go.
I think IT sites should stop doing SATA 3 SSD reviews unless a) the price per GB is significantly better than other SATA 3 SSDs or b) the capacity is huge (e.g. 2TB or more).
2TB SATA still cost an arm and a leg. I shudder to think what a PCI-E version would cost.
Then OTOH, I can get some spinning rust 4TB for a little over £130.00. I have 4 in my Drobo. Great for backups.
As much as I'd like to get rid of all HDD's the cost for SSD's make it impossible.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
