Solid-state storage (SSS) is a type of computer storage media that stores data electronically and has no moving parts. Solid state storage is made from silicon microchips. Because there are no moving parts, SSDs require less power and produce far less heat than spinning hard disk drives or magnetic tape. In addition to providing faster and more consistent input/output (I/O) times, solid-state storage media offers the same levels of data integrity and endurance as other electronic devices.
Solid-state storage can be found generally in three form factors: solid-state drives (SSD), solid-state cards (SSC) and solid-state modules (SSM). An important advantage of solid-state storage is that it contains no mechanical parts, allowing data transfer to and from storage media to take place at a much higher speed and providing a more predictable lifespan for the storage media.
How solid-state storage can replace multiple hard drives
Although solid-state storage technology is not new, interest in how the technology can be used to improve enterprise storage has been relatively recent. Part of this trend can be attributed to reductions in price, but hardware performance also plays a role. Since the turn of this century, processor speeds have continued to increase dramatically while read and write times for mechanical HDDs have not.
Today's CPUs can process data much faster than HDD storage can supply it. The resulting lag time is known as latency, and one way enterprise administrators have traditionally dealt with high storage latency is by short-stroking the hard disk drives in the storage system. Short-stroking is done by deliberately limiting the disk drive capacity to the outer tracks of the drive so the disk drive actuator only has to move its heads across a shorter distance and fewer tracks which reduces seek time. Environments that implement short-stroking typically have to make up for the reduced capacity used in each disk drive by increasing the number of disk drives in these configurations. While short stroking might deliver the required performance, it's an uneconomical solution.
In contrast, solid-state storage devices have zero seek time—it doesn't matter where the needed data is stored. This considerably reduces their latencies, which makes them faster than HDDs, especially for random read/write operations. SSS devices have less of a performance advantage when it comes to sequential read/write operations.
Enterprise applications for solid-state storage
In the enterprise, solid-state storage technology is used for primary storage and also as cache in front of traditional spinning disks, introducing a new layer between the processor and storage. Inevitably, solid-state storage will eventually replace most hard disk storage as silicon factories ramp up to meet the increasing demand for product and the price for SSS continues to decline. With capacities of 12 terabyte (TB) or 14 TB, hard disks may still be the preferred media for archive and other high-capacity/limited-access applications. However, solid-state may soon rival hard disk systems for those applications, too, as a number of solid-state vendors have introduced even higher capacity products, such as Western Digital's 15 TB Ultrastar SAS SSD; Samsung's 30.7 TB PM1643 SAS SSD; and Toshiba Memory Corporation's CM5-R series (up to 15.3 TB) and its PM5-R series that tops out at 30.7 TB.
Advances in solid-state storage haven't been limited to increasing capacities. A new protocol, NVM Express (NVMe), leverages the PCI Express interface to provide more direct and faster communications between the solid-state storage and the server's CPU. NVMe can dramatically reduce latency and pump up I/O for more demanding applications.
Cost of solid-state storage
Not that long ago, a 2 TB RAM-based solid-state external storage system, such as those marketed by Texas Memory Systems (acquired by IBM in 2012) would cost hundreds of thousands of dollars. Today, a 2 TB RAID 0 PCI Express-based SSD can be purchased for as low as $300. A 1 TB SSD can be purchased for $100 to $150. While SSD prices have decreased dramatically, SSDs are still somewhat more expensive than HDDs. A mechanical HDD with 2 TB of storage can be purchased for less than $100.
Types of solid-state storage systems
NAND flash is generally used in enterprise SSS products because of its higher capacities and faster erase and write times. It is non-volatile, which means the data on the storage media remains in memory after the power is turned off. In contrast, RAM-based solid-state storage is volatile -- the storage media requires constant power to retain the data it holds. RAM-based systems have the advantage of being relatively insensitive to the number of times data is written to them, but because of the high cost of RAM and its volatility it's used almost exclusively as system memory rather than storage.
A new category of solid-state storage called storage-class memory (SCM) has emerged that bridges the gap between system RAM and persistent storage. SCM provides greater performance than typical solid-state storage devices but is still slower than RAM. It does, however, mimic RAM in how it addresses and accesses data, so it can work more closely with system memory to further boost performance.
Under the hood of solid-state storage
Flash-based SSDs store data in cells on the NAND chip. Currently there are four ways data is stored in those cell: single-level (SLC) stores one bit of data per cell; multi-level cell (MLC) puts two bits in a cell; triple-level (TLC) stores three bits per cell; and quad level (QLC) stores four bits of data per cell. As the number of bits per cell increases, capacity goes up, but performance may dip.
Capacity may be further increased by shifting from a planar to 3D architecture. In a planar design, all the cells are laid out next to each other on one level; 3D NAND layers cells on top of each other, thus making it possible to increase capacity without increasing the chip's footprint. Chip makers are marketing 3D NAND flash with up to 96 layers of cells.
Flash-based systems have a finite number of writes and, like magnetic tape, the media can wear out. For solid-state storage devices, writing involves a process called the program/erase cycle (P/E cycle) where data that occupies a cell must be deleted before new data is written to that cell. That's why write operations are slower than reads with solid-state storage. The P/E cycle is the main cause of wear for NAND flash; solid-state storage vendors typically estimate the amount of wear its products can endure before failing in terms of how many device/drive writes per day (DWPD) the device can handle.
This video (at left) with Demartek LLC founder and president Dennis Martin delves into solid-state storage technology best practices and developments surrounding the technology.