“At the beginning of 2021, the price of SSD solid state disks began to rise due to the shortage of capacity supply. Currently, the quotation of 240GB SSD has risen from US$24.2 to US$24.6; the quotation of 480GB SSD has risen from US$44 to US$44.5.
Author: Inspur Storage Product Manager Zong Ganggang, Inspur Storage SSD Senior Architect Zhao Baolin
At the beginning of 2021, the price of SSD solid state disks began to rise due to the shortage of capacity supply. Currently, the quotation of 240GB SSD has risen from US$24.2 to US$24.6; the quotation of 480GB SSD has risen from US$44 to US$44.5.
Faced with the rapid increase in the price of SSD solid state drives, Inspur has continued to innovate in solid state drives and launched a new generation of ZNS (Zoned Namespace) SSDs, achieving a leap in performance, capacity, life, cost, and ease of use. promote. This is a comprehensive upgrade of Inspur Storage in terms of solid-state disk technology after Inspur Storage released NVMe SSD solid state disks at the IPF Conference in 2019.
What is ZNS SSD?
ZNS SSD is a partitioned namespace solid state drive. ZNS is developed on the basis of OC (Open Channel) SSD, which realizes the migration of FTL (Flash Translation Layer) from the SSD to the upper Host side. , Open the SSD internal to the Host side, so that users can have their own specific FTL more flexibly according to their needs. The cost is that the software architecture must be redesigned, the cost is very high, and the requirements for users’ technical strength are also very high.
The ZNS protocol standardizes part of NVMe 2.0 to meet industry needs. In contrast, ZNS SSD specifications are more standard, the ecological environment is more friendly, and the software architecture is simplified. Users can more easily develop specific software according to their own scenario requirements.
ZNS SSD moves FTL to the Host side for data management and I/O scheduling. NVMe ZNS provides a partitioned storage device interface, which allows the SSD and the host to cooperate in data placement, and select separate storage locations according to different data types. In this way, the data can be aligned with the physical medium of the SSD, which improves the overall performance and increases the capacity that the host can directly read and write, and realizes I/O isolation.
In the ZNS specification, the entire LBA (logical block address) range of the SSD namespace is divided into several equal-length intervals. The equal-length LBA intervals become a zone, and each zone must be written in a sequential manner. In, each zone has its own independent state machine for management.
ZNS SSD can map different zones and different physical locations according to different application scenarios; configA is a configuration that is most conducive to sequential read and write. This method can ensure the best concurrency of the read and write process. For standard SSDs, they will be written in this way. configB is more conducive to Latency-sensitive scenes, each user controls their own Die, thereby reducing mutual influence.
Because the zone requires sequential writes, when multiple threads write to the same zone, the write speed will be limited or the write location will be incorrect due to the competition of the write pointer between the threads, so the same zone The write queue depth cannot exceed 1. In order to solve this problem, ZNS provides append additional write and ZRWA (Zone Random Write Area) two write modes:
First, Append write mode. The Append command will determine the LBA written by the SSD and return it to the user, which eliminates the problem of zone write location competition between host software.
Second, ZRWA allows random writing and in-situ overwriting of data in the SSD cache. Compared with the Zone Append command, this method requires more resources on the SSD.
Ordinary SSD VS ZNS SSD
For ordinary SSD products, files cannot be written separately by the drive, which ultimately results in files with different lifetimes being stored in the same erase unit, while ZNS SSD supports the coordinated placement of data between the host and the storage device, which can realize the data in the application control area Placed, that is, partitioned storage.
ZNS SSD is mainly to be able to make full use of the storage capacity of the underlying media, while taking advantage of the characteristics of the media itself. For example, for QLC NAND flash SSDs, ZNS can adopt different data partitioning methods according to the type of data and access frequency to reduce overall writing Amplify, thereby prolonging the life of the hard disk, at the same time, it can also improve the I/O access delay, and support the host and the storage device to coordinate data placement. The following figure shows the difference between ordinary SSD and ZNS.
ZNS SSD requires that all zones can only be written sequentially (or append, etc.), and no replication operations can occur. The zone reset command needs to be executed on the zone before reuse, so that applications and drivers need to be more compatible with ZNS SSD, but the current major The storage system and open source tools have already completed or planned to complete the adaptation and support of ZNS SSD, and the standardized interface has realized a powerful software and hardware ecosystem.
Compared with ordinary SSD products, ZNS SSD has five major advantages: eliminates the WAF (Write Amplification) of the GC (Garbage Collection) process of the SSD, and eliminates the traditional OP (Over provisioning, reserved space) space , Due to the smaller mapping table, less DRAM, higher throughput and lower latency, greatly improving the life of SSD disks.
Introduction of Inspur ZNS SSD
Inspur is the first manufacturer in China to launch a new ZNS SSD architecture solution. ZNS storage technology is aimed at cloud scene applications and is mainly suitable for large-capacity storage data, such as high-definition videos and images.
Inspur Storage SSD R&D team worked closely with cloud service providers to optimize its software stack from top to bottom, completed business adaptation and customized SSD development, realized partition storage, and its technical strength was highly recognized by users. At the same time as business development needs, the total cost of ownership of TCO has been greatly reduced, which has brought considerable economic benefits to customers.
Inspur’s self-developed ZNS brand-new SSD architecture has achieved leapfrog improvements in terms of capacity, life, cost, ease of use, performance, etc. The following quantitative figures are presented with 8TB disks against the standard Inspur NS8600G1 6.4TB capacity point:
First of all, cost advantages. Compared with ordinary SSDs, Inspur ZNS SSDs with the same capacity have less demand for DRAM, and DDR costs are reduced by 97%; the internal OP space of SSDs is greatly reduced, and the capacity is increased by 34%, which releases more capacity for the storage system; Improved SSD write amplification, increased DWPD (Drive Writes Per Day, drive writes per day), and increased SSD lifespan by 70%;
Secondly, the performance advantage. Inspur ZNS SSD can significantly reduce the write delay. Through the physical isolation of the zone, the write delay can be greatly reduced; the throughput can be improved, and the write amplification can be reduced to close to 1, providing more bandwidth to users. , All SSD resources are used for application reading and writing, and the overall performance is improved by 210%.
In the future, Inspur Storage will adhere to the storage concept of “cloud storage and intelligent use of new data”. In addition to the two major storage platforms of distributed storage and all-flash storage, it will continue to introduce higher-performance and larger-capacity SSDs to achieve Comprehensive innovation from storage components to storage systems accelerates the digital transformation of enterprises.