The 14th Five-Year Plan calls for “accelerating digital development and building a digital China”. IDC predicts that by 2022, 65% of global GDP will be driven by digitalization. In recent years, the rapid development of new infrastructures, the digital economy, and the platform economy have presented data centers with new challenges. The all-flash data center has the advantages of high speed, green energy saving, etc., which will be the development trend of the future data center. SSD (Solid State Disk) will also be more common.
※ Why focus on SSD lifespan prediction?
Enterprises pay attention to SSD lifespan prediction technology because the application prospects of SSD are very broad and the market is more and more utilized. Second, the data loss caused by SSD damage is huge. Three, because flash memory has erased time-limit functions
In comparison to traditional mechanical drives, SSDs have obvious advantages. For example, SSDs have faster data access speeds, 100x higher throughput, and 1,000x higher IOPS than mechanical hard drives. Rapid technological developments such as NVMe and PCIe are also unleashing additional SSD performance. In terms of reliability performance, SSD is more reliable due to its lightweight, small size, shock resistance, and better drop resistance. For example, SSDs save 70% of power compared to mechanical hard drives. This year’s government work report proposes “peak CO2” by 2030 and “carbon neutrality” by 2060. The use of SSDs can significantly reduce the energy consumption of data centers. In the past, the use of SSDs was mainly due to the high price. According to IDC, SSD prices have fallen an average of 25% per year between 2015 and 2020, and this trend is set to continue for the next five years. In 2020, global enterprise SSD spending exceeded traditional hard drive spending.
Due to the high speed but relatively high price of SSDs, SSDs are generally used to store metadata or core data, and losing this part of the data will bring more serious losses to users. In addition, SSD flash media has a limited erase time, so SSD lifespan deserves more attention and attention.
※ How to predict the SSD lifespan?
The SSD consists of a control chip and a memory chip. The control chip is the brain of the SSD and is used for implementation and data transfer. The memory chip is used to store data. The current main memory chip is a NAND flash memory chip. NAND uses a floating gate transistor to store data that must be erased before writing. The essence of the write operation is to inject charge into the floating gate, while the erase operation is to remove the charge from the floating gate. Ultimately, there is no guarantee that there will be enough charge on the floating gate. Therefore, the number of NAND erases is limited. A full erase of flash memory is referred to as P/E, and flash memory endurance is based on P/E. For example, the MLC SSD erase count is 10,000. The main purpose of predicting SSD lifespan is to predict how long the NAND chip P/E cycles will be available.
Disk vendors generally follow the S. standard for data protection. The standard is an automatic detection and warning system that drive health specification. SSDs include some lifespan indicators that vary depending on the interface and vendor. Inspur NVMe SSD is NVMe 1 compatible.3 and provides S.Standard output. The following table lists the endurance specifications for some vendors’ SATA ports and PCIe ports:
| Interface | Vendor | Name of Index | Index Meaning |
|---|---|---|---|
| SATA | Intel | Smart_233 Media Wearout Indicator | Medium wear indicator |
| SATA | Micron | Smart_202 Percent lifetime remaining | Percentage of remaining life |
| SATA | Samsung | Smart_177 Wear Leveling Count | Erase count |
| SATA | Seagate | Smart_231 SSD Life Left | Remaining SSD Lifespan |
| PCIe | DiskMFR etc. | percentage_used | Percentage used |
SSD Lifespan Prediction predicts the number of days an SSD can be used based on the level of flash wear and tear. At the same time, the lifetime prediction model can support DiskMFR‘s self-developed PCIe SSD.
SSD lifespan prediction uses time-series prediction technology, based on HDDs. Prediction requires a standard collection of historical time-series data sets and then uses the AI’s prediction algorithm developed by DiskMFR to predict. If the indicator changes in the future, the remaining days of SSD lifeSPAN will be retrieved.
※ Six-fold protection business is always online, and reliable!
At the management software level, the InView platform regularly collects data and forecasts on a daily basis, and DiskMFR automatically develops the product SSDs Function to display the percentage of usage and available reserve and the lifespan of the managed SSD. When the forecast result is less than two weeks, an alarm will be generated to inform users about backup and disk replacement schedules to avoid service degradation or even maintenance downtime due to unexpected disk replacement.
In addition to management software, DiskMFR Storage also integrates end-to-end business through core software, devices, components, systems, and solutions, and strives to provide customers with integrated solution services to ensure failures are detected in time. they have no effect and do not fail permanently.
With the development of the digital economy, data elements will play an increasingly important role in the digital transformation of companies in the future. DiskMFR Storage will continue to increase its investment in enterprise-level SSD R&D, promote continuous innovation of centralized full-flash and distributed full-flash technology, and work with green partners such as industry, university, research, and application to achieve provide solutions. of the data lifecycle, help organizations accelerate digital transformation, and unlock the value of data.
