Slashdot Mirror
NAND Flash Density Surpasses HDDs', But Price Is Still a Sticking Point (computerworld.com)
Lucas123 writes: With the introduction of 3D or stacked NAND flash memory, non-volatile memory has for the first time surpassed that of hard disk drives in density. This year, Micron revealed it had demonstrated areal densities in its laboratories of up to 2.77 terabits per square inch (Tbpsi) for its 3D NAND. That compares with the densest HDDs of about 1.3Tbpsi. While NAND flash may have surpassed hard drives in density, it doesn't mean the medium has reached price parity with HDDs — nor will it anytime soon. One roadblock to price parity is the cost of revamping existing or building new 3D NAND fabrication plant, which far exceeds that of hard drive manufacturing facilities, according to market research firm Coughlin Associates. HDD makers are also preparing to launch even denser products using technologies such as heat assisted magnetic recording.
That isn't how wear leveling algorithm work. Yes, once you hit 99%, every write does involve a rewrite somewhere, but those writes are not concentrated in the 1% free area. Instead, the drive controller is reading sections of already written disk and moving them around.
That's 4.294 Gb/mm^2 and 2.02 Gb/mm^2, respectively, for us SI folks.
Most newer SSDs are designed to fail gracefully. When they die, they become a read-only device. All your data is still accessible. Many USB flash drives are designed to fail the same way - if you've ever had a USB flash drive mysteriously become "write-protected", it probably died and set itself to read-only mode. Unfortunately, Samsung seems to be one of the SSD manufacturers which hasn't yet adopted this philosophy for failure. But I can understand their reasoning because...
That problem was solved in the 2000s with wear-leveling algorithms. Basically, the "sectors" the SSD presents to the computer aren't actual physical locations. They're virtual locations stored in a table. If the SSD senses certain blocks being used too much or other blocks sitting unused, it moves the data around behind the scenes so that writes hit all flash memory cells about equally. It updates the virtual table every time it does this, to fool the computer into thinking the drive is physically the same as it has always been.
The rated endurance on most consumer SSDs is around 2000-3000 cycles. For a 250 GB SSD, that means you can write 625 terabytes to it before expecting a failure. If you write 100 GB of data to the drive every day, you can expect it to last nearly 20 years. In torture tests, most SSDs have lasted about 2-3x longer than their rating. And no, the first cell failure is not catastrophic. Pretty much all SSDs have a number of reserve cells sitting on the sidelines to take over for cells which fail early.
If your duty cycle is higher than 100 GB/day, they make special enterprise SSDs rated for 10k-100k writes per cell. The price is correspondingly higher of course, primarily due to using SLC (one bit stored per cell) instead of MLC (2 bits) or TLC (3 bits).
Limited number of writes were more of a problem in the early days of SSDs when they were like 32 GB in size. In that case, the exact same characteristics as the above 250 GB SSD would yield only 2.2 years of longevity. But the problem has pretty much become a non-factor as capacities have increased.
OTA channels = 19.38 Mbps (max)
2 channels = 38.76 Mbps = 4.845 MB/sec
1 Terabyte SSD = 1,000,000 MB
1,000,000 / 4.845 = 206,398 seconds, or 2.3 days
Nand flash write cycle life : 10,000
Total life 10,000 * 2.3 days = 23,000 days or 65 years
If you don't like the assumptions, feel free to make your own, but I think it's clear that write cycle life isn't going to be the limiting factor.
The largest recording I've ever seen off of cable TV is about 8GB/hr. I know OTA broadcasts can be slightly bigger, so lets say 10GB/hr. To record that 24/7 requires about 87 TB/year.
There was a long term test of SSDs done here:
http://techreport.com/review/2...
Many of the drives ended up getting close to 1 PB of writes, and the best even got over 2PB. Thats enough for you to run 2 tuners 24/7 for a decade. And note, their tests were with 250GB drives. As you increase SSD capacity, longevity increases almost linearly. If you were building a DVR, you'd probably want something like a 1TB drive.
As far as the original question of whether the SSD can outlive HDD in the most extreme application....probably at the most extreme, no. But for the vast majority of cases, including a DVR, most likely yes.
Let's work it out. A few years ago, TechReport ran an SSD endurance experiment to figure out how much punishment current-gen SSDs could take before failing. Their test setup essentially involved writing random data at maximum speed for 18 months straight. The results indicated that the worst SSD in their bunch, a Intel's 335 Series, wrote about 700 TB before dying, and the best SSD, a Samsung 840 Pro SSD, went on to 2.4 PB.
Various estimates say you can put between 60-75 hours of HD content on a 500GB drive, so, assuming the largest possible size, that works out to about 8.3 GB/hour. Since you're writing two streams, that's 16.6 GB/hour, or 145 TB per year. For the worst drive in the bunch, that's about 4.8 years of service (right at the upper end of your HDD's service life); for the best drive, it's over 16 years.
Keep in mind that these tests were all run on 250GB drives. Smaller drives have less flash to work with, and have to write over the same flash cells more often. Therefore, if you bought a 1TB drive, you can expect the lifetime to be easily 4x better (more if you're using a more recent drive, such as the Samsung 850 Pro) - 64 years of DVR recording should be more than sufficient.
Hint: every SSD has *at least* 6% extra space for wear leveling - 1TB drives are internally 1024TiB.
P.S. If you wanna counter my first argument, fill your SSD up to 99% and then try to work with it continuously for quite some time. That 1% will get overwritten multiple times and your whole SSD will be prone to a failure.
P.S. Bullshit
SMART Attributes Data Structure revision number: 1
Vendor Specific SMART Attributes with Thresholds:
9 Power_On_Hours 0x0032 094 094 000 Old_age Always - 28138
177 Wear_Leveling_Count 0x0013 097 097 000 Pre-fail Always - 98
241 Total_LBAs_Written 0x0032 099 099 000 Old_age Always - 9528109928
That's a 100% full 128GB Samsung 830 - there's a headerless dm-crypt volume on it, so from the point of view of the drive every single user visible block contains data. ... 100 years or so.
4.87TB written and it's at 98/3000 erase cycles, a WA factor of about 2.76.
Considering that's after 3 years of continuous operation, at this rate it should hit the rated erase count in
You didn't read the report correctly. And I wouldn't plan on pushing anything past its rated limits for writes. That being said, "writes" are the limiting factor, just before general failure. Longevity is more than just writes, it is component failures as well.
It is important to note, that spinning drives have significant drop in reliability at about 42 months (See BackBlaze stats). And while drives can last WELL into 8 years, if you're dealing with critical data, you really don't want to push it much past 42 months. I personally had drives last 10 years. I didn't have any trust in them for the last half of that time.
Agent K: A *person* is smart. People are dumb, stupid, panicky animals, and you know it.