Interview I Samsung Semiconductor
The history of flash innovation - from D3 V-NAND to NVMe
Over the past decade, flash storage has undergone a huge transformation - and is almost unrecognisable to what it once was. Driven by the need for instant access to information, high quality storage for data, and more efficient ways to cool data centre infrastructure, flash has moved on from its adolescence into full adulthood with two landmark innovations. Thomas Arenz, director Marcom/SBD at Samsung Semiconductor Europe, talks to CIE magazine
O
f course, flash has undergone many ‘innovative’ changes over the last year, but I want to focus on two, without which, flash storage would simply be too impractical, expensive and unappealing for the data centre.
3D V-NAND - doing more with less The growing volume, velocity and variety of big data has highlighted the need to be able to fit more data into servers more quickly than ever before. Priorities for data centre managers were (and still are), data storage reliability, performance and capacity. However, the older style data centre was in danger of physically filling up with more and more servers to accommodate more data. There’s only so much space within a data centre to play with.
12 December 2015/January 2016
SSD manufacturers began to shrink the individual components to save space without compromising on storage capacity, which to a point worked well but then the technology hit a wall. To reduce the size of components, manufacturers had to push memory cells closer together. As the space between cells became smaller, SSDs were more likely to suffer from ‘coupling’, leading to a higher level of interference and even data corruption. This scaling limit would have made storage products less reliable for inclusion within servers, and ultimately within data centres themselves. Since Samsung did not want to
compromise between physical size and reliability, Samsung developed a new form of ‘vertical’ memory, which incorporates
Components in Electronics
more cells without pushing them closer together. Rather than packing in more cells into the same horizontal space, this new approach accommodates more memory by building on top of existing structures in a new vertical structure called ‘3D V-NAND’. This new technology provides a ten-fold improvement in longevity for read intensive and an average of five-fold for mixed use. Besides improved write performance, the vertical structure also makes way for much higher densities on the same footprint. Starting with 24 layers, Samsung’s V-NAND base components came at 128GB a piece whereas not even two years later, the current generation has 48 layers and allows for multi- terabyte densities already in small 2.5” form factors. This not only helps to solve the issue of big data, but it also helps to reduce energy consumption and long term maintenance costs.
NVMe and flash - a match made in heaven Non-Volatile Memory Express (NVMe) is firmware technology that gives SSDs the ability to reach their full speed potential. However, before we get onto how NVMe improves flash speed, it’s important to remember how data centres once used SSDs.
A few years ago when flash was still
relatively young, manufacturers built in SSDs with SATA and SAS buses to interface with the rest of the server. This approach worked well, because at the time, solid state operated roughly at the same speed as HDDs. However, problems
arose when flash speeds began to increase because the technology became more and more held back by these dated buses. SATA’s 600Gbps limit forced most state of the art SSDs to operate at roughly 500Mbps no matter what speed capabilities those SSDs have.
Fixing flash with NVMe NVMe works as part of the server’s existing high bandwidth peripheral component interconnect express (PCIe) bus. PCIe is the underlying data transport layer and defines a command and feature set for PCIe-based SSDs.
The NVMe-enabled PCIe works specifically with the latest SSDs and is capable of extracting the most value from an SSD. PCIe 2.0 offers 500Mbps per lane and PCIe 3.0 offers 985Mbps. However, by installing a card in a four lane PCIe 2.0 slot, bandwidth will increase to 2Gbps, and by installing a card in a PCIe 3.0 slot, bandwidth will soar to 4Gbps - much better for today’s fastest SSDs.
Using NVMe, PCIe offers much faster speeds, lower latency and lower energy consumption than SATA. This enables an entirely new breed of storage architecture, making data available to the CPU in large chunks, ultimately offering massive bandwidth and IOPS (input/output operations per second) improvements over both SATA and SAS. Whereas SATA requires two messages for 4KB transfers and can only process a single queue, NVMe requires just one message and can process up to 65,536
www.cieonline.co.uk
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44