This is achieved by constructing a conductive mold stack of 136 layers and then perpendicularly engaging the cylindrical openings from top to bottom. In this way, uniform 3D cells with a cargo trap (CTF.
Storage 101: Understanding the NAND Flash Solid State Drive
Today, most organizations use SSDs for everything from laptops to enterprise database storage and virtual machines. In this article, Robert Sheldon explains how NAND flash SSDs work.
Solid state drives (SSDs) have made a significant step into enterprise data centers in recent years, handling workloads that were once the exclusive domain of hard drives (HDDs). SSDs are faster, smaller, use less power, and have no moving parts. They also fell in price while supporting higher densities, making them suitable for a wide range of applications.
Despite their growing presence in the enterprise, there is still a lot of confusion about how SSDs work and the features that distinguish one drive from another. Concepts like NAND chips, multi-level cells, and floating gate technologies can be a bit daunting if you still think in terms of spinning disks and movable actuator arms, components that don’t have room for an SSD.
The more you understand how SSDs work, the more efficiently you can select, deploy, and manage them in your organization. To assist with this process, this article introduces some important concepts of SSDs so that you get a clearer picture of the components that make up an SSD and how they work together to provide reliable persistent memory.
However, keep in mind that an SSD is a complex technology and can easily justify a much deeper coverage than what a single article can offer. You should think of this as an introduction, not a complete treatise, a starting point for building the foundations for understanding the inner workings of your SSDs.
Introducing the NAND Flash SSD
Like a hard disk, an SSD is a non-volatile storage device that stores data whether or not it is connected to power. However, a hard drive uses magnetic media to store data, while an SSD uses integrated electronic circuits to preserve certain states of charge which, in turn, map to data bit patterns.
SSDs are based on flash memory technologies that allow data to be written, read, and erased multiple times. Flash memory comes in two varieties: NOR and NAND. While each has its own advantages and disadvantages (a discussion that is beyond the scope of this article), NAND has become a favorite technology as it provides faster erasing and writing. Most modern SSDs are based on NAND flash memory, which is why this is what this article focuses on.
An enterprise SSD contains multiple NAND flash memory chips for data storage. Each token contains one or more cubes, and each cube contains one or more planes. The plane is divided into blocks and the block is divided into pages.
Of these, the biggest problem is blocks and pages, not because you configure them directly or manipulate them directly, but because of the way data is written, read, and deleted on the NAND chip. Data is read and written at the page level but deleted at the block level as shown in Figure 1.
Figure 1. Saving and erasing data on a NAND flash SSD drive (photo by Dmitry Nosachev, under the Creative Commons Attribution-Share Alike 4.0 International license)
In this case, each page is 4 KibiBytes (KiB) and each block is 256 KiB, which is equivalent to 64 pages per block. (Kibibyte is 1024 bytes. Kibibytes are sometimes used instead of kilobytes because they are more precise. A kilobyte can be 1000 bytes or 1024 bytes depending on its use.) Every time an SSD reads or writes data, it does so in Pieces 4 KB, but each time the disk erases data it performs a 256 KB operation. This difference between save and erase has serious ramifications when updating data, as you’ll see later in this article.
Most client systems are no longer limited by CPU performance. They are almost always limited by storage. Hard drives have access latency in milliseconds while SSDs run in hundreds of microseconds.
A New Way To Increase Capacity – 3D V-NAND
3D V-NAND is a cell layering technology where multiple layers are created on a single NAND chip. Samsung implements up to 32 layers for chips that go to the latest SSDs. Note that cells are stacked, not the chips themselves. Of course, more layers means potentially more storage capacity. V-NAND is great because it works on top of existing manufacturing processes such as the 10nm plus class, helping to alleviate the problems of reduced endurance and inferior performance when NAND is further scaled.
This is an important point. As memory cells shrink as a result of moving to smaller manufacturing processes, lower operating voltages and more cells can be expected to fit in a given area, thus increasing the capacity per chip, which is crucial in the quest for SSD capacity over 1 TB. While this is reasonably true, ever smaller processes – where cells are very, very close to each other – have a huge side effect: disruption. The proximity of the cells causes disturbances that inhibit the speed of the entire chip. So, while smaller manufacturing processes are generally a good thing, they also have undesirable problems. 3D V-NAND takes advantage of many of the advantages of low-volume chips, but not the disadvantages.
In addition, switching to the 3D vertical cell stacking model means that compared to regular “flat” flat NANDs, Samsung’s 3D V-NAND requires less physical space for the same capacity – remember that multiple layers of cells are stacked on top of each other – thus enables mature production processes to be used instead.
Reinventing The NAND Chip
3D V-NAND is about more than just stacking more layers on a chip. Flash memory uses an electrically insulated gate floating beneath the transistor’s control gate – CPU transistors only have one gate – and its purpose is to increase the voltage threshold needed to make the transistor conductive. When a load enters the float gate through the control gate, the memory is said to be programmed. Removal of the load causes the content to be deleted.
Instead of using a traditional floating gate to add or remove charge coming from a control gate, Samsung says, which is extremely difficult when building a 3D device, it is replaced by a charge trap, which is a non-conductive layer that effectively wraps around the transistor. This layer is critical in minimizing inter-cell interference, has less leakage properties, requires less voltage to operate, and therefore uses less energy. A byproduct of the lower tension is less stress and increased strength than would be possible with a conventional floating gate.
Samsung took a look at the technology behind the chips that make up SSD capacity and devised a method to increase capacity, reduce power consumption, and increase overall endurance without having to follow an expensive, problematic pathway to a smaller manufacturing process. The transition to 10nm and smaller will take place in due course, but Samsung thinks it makes more sense to use mature processes to minimize risk and keep the NAND chips flowing from the factories. Taken together, these enviable features enable Samsung to offer class-leading performance and power consumption on the latest SSD array for consumers and data centers.
One of the benefits of NAND flash memory is non-volatile data storage. Unlike DRAM, which must be powered continuously to retain data, NAND memory retains data even when powered off – making it ideal for storing portable devices.
256GB V6 V-NAND SATA SSD
Samsung SATA 256GB SSD using 6th generation V-NAND technology.
Details of 250GB V6 SATA SSD:
- 136 layers of flash cells with charging trap
- Single stack design
- 512Gbit matrix
- TLC (3 bits / cell)
- reading delay 45 μs