SanDisk Announces 4TB SSD, Plans For 8TB Next Year

Lucas123 (935744) writes "SanDisk has announced what it's calling the world's highest capacity 2.5-in SAS SSD, the 4TB Optimus MAX line. The flash drive uses eMLC (enterprise multi-level cell) NAND built with 19nm process technology. The company said it plans on doubling the capacity of its SAS SSDs every one to two years and expects to release an 8TB model next year, dwarfing anything hard disk drives can ever offer over the same amount of time. he Optimus MAX SAS SSD is capable of up to 400 MBps sequential reads and writes and up to 75,000 random I/Os per second (IOPS) for both reads and writes, the company said."

Comments

Oh goody

[Anonymous Coward]

Now you can pay $4000 for a drive that won't last 2 years! Yeah.. sign me up.

Re: Oh goody

[Anonymous Coward]

My primary OS is running on an SSD going on 4 years old now... Out of 5 that I have only one had had issues, which was actually it's controller catastrophically failing and not a NAND issue - could have just as easily happened to a HDD.

Re:Oh goody

[beelsebob]

Assuming you write an average of 100GB a day to this drive (which is... an enormous overestimate for anything except a video editor's scratch disk), that's 40,000 days before you write over every cell on the disk 1000 times. Aka, 100 years before it reaches its write limit. So no... SSDs are far from the 2 year proposition that people who bought first gen 16/32GB drives make them out to be.

Re:Oh goody

[Anonymous Coward]

Just turn on DoubleSpace ... might buy you a couple of more years.

Re: Oh goody

[Bryan Ischo]

False. Your one anecdotal story does not negate the collective wisdom of the entire computer industry.

As far as anecdotal evidence goes, here's some more worthless info: I've owned 8 SSD drives going all the way back to 2009 and not a single one has ever failed. They're all currently in use and still going strong. I have:

- 32 GB Mtron PATA SLC drive from 2009
- 64 GB Kingston from 2010 (crappy JMicron controller but it was cheap)
- 80 GB Intel G2 from 2010
- 80 GB Intel G3 from 2011
- 2x 80 GB Intel 320 from 2011
- 2x 240 GB Intel 520 in my work computer, it gets pretty heavily used, from 2012
- Whatever is in my Macbook Pro from 2012
- Just purchased a 250GB Samsung 840 Evo

Not a single failure on any of them, even the old 32 GB Mtron and the piece of crap JMicron controller Kingston.

But this evidence doesn't really matter; it's the broad experience of the industry as a whole that matters, and I assure you, SSDs have already been decided as ready for prime time.

For a recent example, linode.com, my data center host for like 10 years now, just switched over to all SSDs in all of their systems.

Re: Oh goody

[shitzu]

We have ~100 SSDs installed in our company, workstations, laptops and servers. Over five years only 3 of them died, all Kingstons. Samsung and Intel have been spotless. All of those that died had the following symptoms - if you accessed a certain sector the drive just dropped off - as if you switched off its power. The drive did not remap them as it always dropped off before it could do so. Otherwise the drive remained functional. Got them replaced under warranty.

Re:Oh goody

[Mr Z]

If you know something about the drive's sector migration policies, in theory you could construct a worst-case amplification attack against a given drive. Leverage that against the drive's wear leveling policies. But, that seems rather unlikely.

Flash pages retain their data until they're erased. You can write at the byte level, but you must erase at the full page level. You can't rewrite a byte until you erase the page that contains it. That's the heart of the attack: Rewriting sectors with new data. You can't rewrite a sector in-place. You mark the old location as "dirty but free", and write the new data to a new location. The SSD can't reclaim the dirty-but-free sectors for writing until they're erased.

Thus, the basic idea goes something like this: Fill the disk to 99.9% full. Then, selectively rewrite individual sectors, forcing the sector to migrate to a new flash page. Wash, rinse, repeat until the drive fails.

If the drive only performs dynamic wear leveling, all subsequent rewrites will erase and reuse only among the free space. (Note: This free space includes all of the space the drive reserves to itself for dynamic wear leveling purposes.) Now all you need to do is reach the erase/rewrite limit among the available dynamic wear leveling pool, which is significantly smaller than the full drive capacity. You can achieve this by rewriting a small subset of sectors until the disk falls over.

Modern drives perform a blend of dynamic and static wear leveling. Dynamic wear leveling only erases/rewrites among the "free" space. Static wear leveling gets otherwise untouched sectors into the fray by wear leveling over all sectors. This blended approach defers static wear leveling until it becomes absolutely necessary. The flash translation layer (FTL) detects when the wear difference between sectors gets too imbalanced, and migrates static sectors into the worn regions and wear-levels over the previously "static" sectors.

A successful attack would take this into account and attempt to keep track of which sectors would be marked "static" vs. "dynamic". It would also predict how the static sectors were grouped together into pages, so it could cherry-pick and inflict the maximum damage: All it needs to do is write to a single sector in each static flash page (creating a bunch of unallocated "dirty-but-free" holes), continuing until the SSD was forced into a garbage collection cycle. That GC cycle then would have to touch all the static pages (or at least a significant fraction) to compact the holes away and make space available for future writes.

If you can keep that up, you can magnify your writes by the ratio between the page size and the sector size. If you have 512 byte sectors and 512K bytes pages, the amplification factor is 1024.

But, as I suggested above, to achieve this directly, you need to have some idea of how the SSD marks things static vs. dynamic. Without such knowledge, you have to approximate.

I imagine if you really wanted to kill an SSD without any knowledge of its algorithms, you could do something simple like rewrite every allocated sector in an arbitrary order, shuffling the order each time. SSD algorithms assume a distribution of "hotness" (ie. some sectors are "hot" and will be rewritten regularly, and most are "cold" and will be rewritten rarely if ever), and so rewriting all sectors in a random order will cause rather persistent fragmentation, recurring GC cycles, and pretty noticeable amplification.

You wouldn't get to the 40 day mark, but if you started with a mostly full SSD, you might get to a few months.

That's my back-of-the-napkin, "I wrote an FTL once and had to reason through all this" estimate.

Re:Oh goody

[Tapewolf]

If you only write infrequently (use for image editing) and then backup storage - how many years would the SSD maintain values?

If the drive is powered down, I wouldn't bet on it lasting the year. Intel only seem to guarantee up to 3 months without power for their drives: http://www.intel.co.uk/content...

Note also that the retention is said to go downwards as P/E cycles are used up. For me, I think they make great system drives, but I don't use them for anything precious.

Re: Oh goody

[Anonymous Coward]

It's highly unlikely that his parents were fucking yours.

Re:Oh goody

[jon3k]

You do not want to use SSDs for long term storage: http://www.intel.co.uk/content...

"In JESD218, SSD endurance for data center applications is specified as the total amount of host data that can be written to an SSD , guaranteeing no greater than a specified error rate (1E - 16) and data retention of no less than three months at 40 C when the SSD is powered off."

Re: Oh goody

[Hadlock]

While I'm happy for you and your luck so far, the hard numbers don't lie, one in 20 OCZ drives were returned over a two year period. Closer to 7 percent for particular models. Compare to half a percent for Samsung or Intel.

Finally the disk drive can die

[Billly Gates]

It is so archaic in this day and age of microization to have something mechanic bottlenecking the whole computer. It just doesn't mix in the 21st century.

For those who have used them will agree with me. It is like light and day and there is no way in hell you could pay me to do things like run several domain VM's on a mid 20th century spinning mechanical disk. No more 15 minute waits to start up and shutdown all 7 vms at the same time.

Not even a 100 disk array can match the IOPS (interrupts and operations per second) that a single ssd can provide. If the price goes down in 5 years from now only walmart specials will have any mechanical disk.

Like tape drive and paper punch cards I am sure it will live someone in a storage oriented server IDF closet or something. But for real work it is SSD all the way.

Not in my experience.

[aussersterne]

Anecdotal and small sample size caveats aside, I've had 4 (of 15) mechanical drives fail in my small business over the last two years and 0 (of 8) SSDs over the same time period fail on me.

The oldest mechanical drive that failed was around 2 years old. The oldest SSD currently in service is over 4 years old.

More to the point, the SSDs are all in laptops, getting jostled, bumped around, used at odd angles, and subject to routine temperature fluctuations. The mechanical drives were all case-mounted, stationary, and with adequate cooling.

This isn't enough to base an industry report on, but certainly my experience doesn't bear out the common idea that SSDs are catastrophically unreliable in comparison to mechanical drives.

dwarfing? Not quite yet!

[Nagilum23]

Seagate already announced 8-10TB disks for next year: http://www.bit-tech.net/news/h... .
Now if SanDisk can deliver 16TB SSDs in 2016 then they might be indeed ahead of the hard-disks but not in 2015.

Where are the 3.5" SSDs?

[swb]

Why do SSD makers only make 2.5" SSDs? It seems like a lot of the capacity limitation is self-enforced by constraining themselves to laptop-sized drives.

Why can't they sell "yesterday's" flash density at larger storage capacities in the 3.5" disk form factor? For a a lot of the use cases, the 3.5" form factor isn't an issue. More, cheaper flash would enable greater capacities at lower prices.

The same thing is true for hybrid drives -- the 2.5" ones I've used have barely enough flash to make acceleration happen, a 3.5" case with a 2.5" platter and 120GB flash would be able to keep a lot more blocks in flash and reserve meaningful amounts for write caching to flash.

Re:Where are the 3.5" SSDs?

[jon3k]

It's not constrained by size. It's the cost of NAND flash that's the limiting factor. And no one is going to manufacture last generation's NAND, it doesn't make any business sense. Ask Intel why they don't sell last years CPUs at cut rate prices. Same reason.

Re:Where are the 3.5" SSDs?

[Amouth]

there are a few reasons they don't make 3.5's

1: physical size isn't an issue, for the sizes they release that people are willing to pay for it all fits nicely in 2.5
2: 2.5's work in more devices, including in desktops where 3.5's live. if noting is forcing the 3.5 usage then it would be bad for them to artificially handicap them selves.

now for your commend on larger physical drives being cheaper. Flash does not work the way that normal dries to.

Normal platter drives the areal density directly impacts pricing as it drives the platter surface to be smoother, the film to be more evenly distributed, the head to be more sensitive, the accurater to be more precise, all things that cause higher precision that drive up costs as it increases failure rates and manufacturing defects causing product failure.

Now in the flash world. they use the same silicon lithography that they use for making all other chips. there are two costs involved here.

1: the one time sunk cost of the lithography tech (22nm, 19nm, 14nm...) This cost is spread across everything that goes though it. And in reality evens out to no cost increase for the final product because the more you spend the smaller the feature the more end product you can get out per raw product put in.
2: the cost of the raw material in. It does not matter what level of lithography you are using the raw material is nearly exactly the same (some require doping but costs are on par with each other). So in fact your larger lithographic methods become more expensive to produce product once there is newer tech on the market.

No please note that in the CPU world where you have complex logic sets and designs there is an added cost for the newer lithography as it adds to the design costs. but for flash sets there is nearly zero impact form this as it is such a simple circuit design.