Intel unveils itsy-bitsy, teeny-weeny SSDs
Same performance, one-eighth the size.
Intel announced on Wednesday a new line of tiny solid-state drives, the SSD 310 series, that have performance specs comparable to those of their 2.5-inch X-25 and 1.8-inch X-18 SSD brethren – and at least one manufacturer is already on board to put the little fellows into dual-drive notebooks.
"The Intel SSD 310 series will allow us to provide the advantages of a full-performance Intel SSD paired with the storage of a hard disk drive in a small, dual-drive system," Lenovo's ThinkPad honcho Tom Butler said in a prepared statement.
A dual-drive system will allow users to put their system and apps on the SDD for fast boot and launch times, while storing their files and data on relatively pokey, inexpensive 2.5-inch hard drives.
The latest example of how storage systems are shrinking
The SSD 310 series devices – available in 40GB and 80GB capacities, both composed of 34nm NAND chips – are small, indeed. Intel crows that the devices are "up to 8 times smaller than a 2.5-inch hard-disk drive," but that characterization takes all three axes – width, length, and thickness – into account.
To be exact, your mainstream 2.5-inch HDD is typically 70mm wide, 100mm long, and 9.5mm thick. The SDD 310 series devices are 29.85mm wide, 50.80mm long, and a bit less than 4.85mm thick. A typical 2.5-inch HDD weighs between 95 and 105gm, while the SSD 310s weigh under 10gm.
Ten years ago, would you have thought a 2.5-inch drive could ever look so gargantuan?
On paper, the performance of the SSD 310s stacks up quite nicely when compared with that of the Intel X18-M and X18-M Mainstream SSDs, as well as with that of a snappy 7200rpm 2.5-inch HDD such as the Hitachi Travelstar Z7K320, not to mention a garden-variety 5400rpm 2.5-incher such as the Travelstar 5K320.
Although comparing an SSD's sustained read/write rates with Hitachi's advertised media transfer rates is not as exact a comparison as we'd prefer, it's what we have to go on:
If not exactly apples to oranges, perhaps at least Granny Smiths to Gravensteins
The SSD 310 series also matches the X18/25-M series' MTBF of 1.2 million hours and operating shock of 1,500G/0.5ms – far superior to the Travelstar Z7K320's 400G/2ms, 225G/1ms and the 5K320's 400G/2ms, 200G/1ms.
The SSD 310 series is already shipping to customers, and at decent prices – for SSDs – when purchased in 1,000-unit quantities: $99 for 40GB and $179 for 80GB. ®
*pets his /very cheap/ 40TB RAID 6*
G in this respect will be the acceleration due to gravity at the Earth's surface, not the universal gravitational constant. The former being or rather more practical use to the average person, not to mention the units of the universal gravitational constant make no sense at all in this context.
As for 400G/2ms this is not a single unit, but a statement that the device can withstand a rate of deceleration of 400G (or about 3,924 metres per second squared) for 2 milliseconds. That is not the same as 200G for 1 millisecond (which would be half the rate of deceleration for half the period). Do the mathematics on this, and you find that this will allow for this to be installed into a device that can be dropped a metre or two (depending on the degree of bounce) provided that the SSD is decelerated over a few millimetres (which can be achieved through a deformable case, rubber mountings or some combination of such). Most laptops won't survive a 2 metre fall onto a concrete surface, although they might on a heavily carpeted one.
So the 1,500G in 0.5 ms means that the device will stand almost 4 times the rate of deceleration albeit for a quarter of the time. What this means is that for the same collision speed it can be slowed a lot faster and hence the mounting and casing needs to allow for less movement for the same drop height.
As this device is only 10gms, then the 1,500G is equivalent to a force of about 15N or about that experience by a 1.5Kg mass at the Earth's surface. On that basis I suspect that force could well be sustained much longer than 0.5ms, possibly indefinitely. However, it has to be installed into real devices and those will probably have to be designed for real world situations (like being dropped off a 1 metre worktop).
"In a data center people would generally prefer to have big storage capacity as they can access drives in parallel for speed. The greater density allows more storage for the power and cooling required."
It's perfectly true that you can get more IOPs (and throughput) by installing more drives at the cost of power consumption, cost and data centre footprint. However, there is one think hard drives can never do, and that is consistently achieve random read latency times of less than about 6ms. With SSDs you can get down to tens of microseconds (or, realistically, on a SAN hundres of milliseconds). If you have an application which is I/O bound on random reads, then only SSDs will get you out of it once you;'ve exhausted the practicalities of caching. (Random writes are not such an issue - enterprise arrays will cache those in NV store).
As for 4TB drives in the enterprise, then we have had real issues with the reliability of high density drives. They are OK for some semi-archival uses (for which you'd never use SSD anyway - at least not for the forseeable future). However, use them hard, and they fail at a much higher rate than the lower density enterprise drives. Then there is the problem that re-building RAID sets with such high density drives takes a very long time as capacity inevitably outgrows throughput (read/write throughput goes up in proportion to linear bit density, capacity as to square of linear bit density). That pushes you to double parity which removes some of the capacity advantages and can also impact performance.
SSDs will erode the top end of the enterprise drive market as prices drop. Realistically, write endurance is not going to be too much of an issue. Enterprise drives fail too, and once they get to the limit of their practical operating life, failure rates start to increase and swap-outs happen more frequently. Of course it's generally covered by maintenance contracts, and exactly the same thing will happen with enterprise SSDs.
Disks won't go away, at least for the forseeable future, it will just get pushed more and more to the bulk storage area.