SSD Advantages in Data Centers
Compared with traditional HDDs, SSDs are increasingly used in storage fields, including data centers, due to their many advantages:
- Low Power Consumption – When operating a large number of storage drives, power usage increases. Any opportunity to save power is a win.
- Speed – Faster data access is especially useful for caching databases, applications, or other data that affects system performance.
- Minimal Vibration – Less vibration improves reliability, which leads to fewer issues and less maintenance.
- Low Noise – As more SSDs are deployed, data centers will become quieter.
- Low Heat Generation – The less heat produced, the less power is needed for cooling the data center.
- Faster Boot Time – The quicker a storage chassis comes online after maintenance or troubleshooting, the better. Similarly, the faster a server restarts after these processes, the better it is.
- Higher Data Density – Data centers can store more data in smaller spaces, improving the efficiency of the area used.
Trends in SSD Development
- The speed of both front-end protocol interfaces and back-end flash memory interfaces will continue to increase. This will help boost SSD performance. Although PCIe Gen4 is still being widely adopted, many in the industry believe Gen5 will arrive faster than expected. One reason is that Intel plans to officially support Gen5 next year. This means some Gen5 products in the high-end storage market will be available from early to mid-next year. By 2023, Gen5 products will be more common. PCIe Gen5 offers double the bandwidth of Gen4. It increases from 16 Gb/s to 32 Gb/s per lane. It reaches 256 Gb/s with x8 lanes. On the flash memory side, interface rates are also improving—from 800 MT/s to 1.2 GT/s, 1.6 GT/s, and even higher. Presently, 100 Gbps and 200 Gbps Ethernet backbones already exist. Using PCIe Gen5 x8 to connect servers will allow for much higher throughput. In terms of system architecture, more hardware engines are on the SSD read/write path to increase speed. Meanwhile, firmware is increasingly reserved for handling non-critical or error-handling paths.
- Higher performance leads to greater power consumption and heat generation, which increases the demand for system power improvement and cooling. Typically, the optimal operating temperature for SSDs ranges from room temperature to around 50–55°C. Still, extremely fast performance can push temperatures close to or even beyond 70°C. At such levels, SSDs often suffer from degraded performance or, in worse cases, data loss or hardware damage. Besides reducing the power consumption of flash memory, the controller’s power consumption is also a critical factor. For example, mainstream controllers used 28nm nodes for PCIe Gen3, while PCIe Gen4 has increasingly adopted 12nm nodes. Beyond halving the die size, 12nm consumes only about 40% of the power compared to 28nm. Furthermore, SSDs often use system-level strategies to enhance the balance between power consumption, heat dissipation, and performance. One such approach is dynamic firmware-based temperature monitoring, which adjusts performance in real time to keep optimal balance.
- The cost per GB continues to decrease, thanks to increased 3D NAND stacking layers and higher bits stored per cell. Moving from MLC to TLC has increased the requirements for system error correction. Transitioning from QLC (4 bits) to PLC (5 bits) makes these requirements even more demanding. The device now requires more robust error correction capabilities. Technologies like RAID and algorithms tailored to adjust NAND settings are being widely applied. This also increases the demands on LDPC (Low-Density Parity-Check) capabilities of SSD controllers. In 2020, Samsung introduced 1XX-layer NAND. SK Hynix reached 128 layers. Intel launched 144-layer NAND. These advancements paved the way for next-gen SSDs. By the end of 2020, Micron and SK Hynix had announced breakthroughs with 176-layer 3D NAND. Samsung was preparing to mass-produce its 7th-gen V-NAND in 2021. In December 2020, Intel introduced its 144-layer QLC NAND, offering roughly 50% higher capacity density than the earlier 96-layer QLC. The new U.2 form factor PCIe SSD, D5-P5316, based on this technology, offers a massive 30.72 TB of storage. Besides Intel, Samsung’s 870 QVO series and Micron’s 2300/2210 SSDs also use QLC NAND. Additionally, leading controller manufacturers like Marvell have introduced SSD controllers. Silicon Motion and Phison have also provided controllers. These controllers fully support 3D TLC/QLC. As a result, QLC is expected to be adopted in more and more SSD products.
- Emerging memory as cache: Technologies like Intel’s 3D XPoint, RRAM, and STT-MRAM are already being used in SSDs. An example of this is Intel’s Optane product line.
- In-house SSD design is becoming mainstream. Major SSD consumers are increasingly designing their own SSDs, including both controllers and firmware. By integrating hardware and software, they achieve highly customized solutions with superior performance and reduced costs.