High-reliability high-capacity hard disk implementation method and system based on fine-grained ZNS partition

Abstract

The invention belongs to the technical field of SSDs, and particularly relates to a high-reliability large-capacity hard disk implementation method and system based on fine-grained ZNS partition, and the system comprises a PCIe physical link module, a PCIe controller module, a ZNS controller module, a battery management unit module, an NAND controller module, an NAND physical link module and a linear error correction code module. The battery management unit module is provided with a small-capacity SSD cache module; the ZNS controller module is used for realizing fine-grained ZNS partition layout; the battery management unit module manages the specified group SSD cache and is used for receiving write data of the host and storing read data of the host; the linear error correction code module is used for coding and correcting low-probability errors generated in the data reading and writing process; and the NAND controller module and the NAND physical link module are used for processing read-write requests of the flash memory particles. The size of the SSD hard disk partition is reduced through the fine-grained ZNS partition, fine fault isolation is achieved, the SSD hard disk fault domain is reduced, and the hard disk reliability is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of SSD solid state drives, and particularly relates to a method and system for implementing a highly reliable and large-capacity hard disk based on fine-grained ZNS partitioning. Background Art

[0002] With the development of flash memory technology, due to its good storage performance and extremely low power consumption characteristics, it has become a trend for SSD hard disks (Solid State Disk) to replace traditional HDD hard disks (Hard Disk Drive). However, as the capacity of existing SSD hard disks increases and the storage density improves, the failure domain becomes larger, resulting in a problem of reduced reliability. The failure rate of SSD hard disks is high. The main reasons for failures obtained from hard disk manufacturers are as follows: SSD hard disk control software failure; SSD hard disk memory failure; SSD hard disk capacitor failure; NAND (NOT AND, circuit design of logic gates) Flash, flash memory particle failure (the higher the storage density of the particle, the lower the durability).

[0003] With the release of the ZNS (Zoned NameSpace) command set specification in NVMe (Non-Volatile Memory Express) 2.0, ZNS SSD hard disks (hard disks supporting the ZNS command set) provide the NAND physical address or concurrent unit to the host, giving full play to the concurrent characteristics of NAND and improving performance; ZNS partitioning allows the host to write data of different applications into different partitions, and the partition attributes can be aligned with the characteristics of the data written by the host, optimizing the data placement on the SSD hard disk medium. However, ZNS partitioning has two defects: One is that ZNS usually has relatively large partitions with a capacity in the GB level, so redundant algorithms such as RAID (Redundant Array of Independent Disks) are still required to ensure the reliability of ZNS partitions. Although ZNS partitions can give play to the performance advantages of multiple Dies (wafers in flash memory chips, which are the smallest units that can execute commands and return statuses) to perform read and write operations in parallel, since ZNS partitions can only write sequentially and use WP (Write Point) to indicate the next write position, concurrent writing does not play a big role in the case of writing.

[0004] The other is that ZNS contains all the Blocks (flash memory physical blocks, which are the smallest units for data erasure) on all Dies. If there are more than the redundant number of Dies offline in a RAID stripe, all ZNS partitions will fail, and fault isolation between partitions cannot be achieved.

[0005] Therefore, how to improve the existing implementation method of SSD solid-state drives, achieve more refined fault isolation, narrow the SSD hard disk fault domain, and improve the hard disk reliability is a technical problem that urgently needs to be solved at present. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and system for implementing a highly reliable and large-capacity hard disk based on fine-grained ZNS partitioning, so as to improve the existing implementation method of SSD solid-state drives, achieve more refined fault isolation, narrow the SSD hard disk fault domain, and improve the hard disk reliability.

[0007] To solve the above technical problems, the technical solutions adopted by the present invention are as follows: In the first aspect, a system for implementing a highly reliable and large-capacity hard disk based on fine-grained ZNS partitioning is provided, including a PCIe physical link module, a PCIe controller module, a ZNS controller module, a battery management unit module, a NAND controller module, a NAND physical link module, and a linear error correction code module. The battery management unit module is provided with a small-capacity SSD cache module; The PCIe physical link module and the PCIe controller module are used to process the PCIe protocol; The ZNS controller module is used to implement a fine-grained ZNS partition layout; The battery management unit module manages the specified group of SSD caches, and is used to receive the write data of the host and save the read data of the host; The linear error correction code module is used to encode and correct low-probability errors generated during the data reading and writing process; The NAND controller module and the NAND physical link module are used to process the read and write requests of flash memory particles.

[0008] Preferably, the PCIe physical link module is connected to the PCIe controller module, the PCIe controller module is connected to the ZNS control module, the ZNS controller module is connected to the small-capacity SSD cache module, the small-capacity SSD cache module is connected to the NAND controller module, the NAND controller module is connected to the NAND physical link module, and the linear error correction code module is connected to the small-capacity SSD cache module.

[0009] Preferably, the fine-grained ZNS partition layout implemented by the ZNS controller module aggregates the Blocks with the same block ID on all Dies within a Target into a ZNS partition.

[0010] Preferably, the ZNS controller module realizes fault isolation between fine-grained ZNS partitions through the process of fine-grained ZNS partitioning. The ZNS partitions are distributed on the Blocks within the same Target. A single Target failure or Die failure only affects the ZNS partitions distributed on the Target or the Die. A single Block failure only affects the ZNS partition where the Block is located, removing the correlation between the Block failure and the whole-disk failure.

[0011] In a second aspect, a method for implementing a highly reliable and large-capacity hard disk based on fine-grained ZNS partitioning is provided, which is implemented based on the system for implementing a highly reliable and large-capacity hard disk based on fine-grained ZNS partitioning according to any one of the above, and includes the following steps: S1: Deploy an SSD hard disk controller in the server. The SSD hard disk controller deploys a PCIe physical link module, a PCIe controller module, a ZNS controller module, a battery management unit module, a NAND controller module, a NAND physical link module, and a linear error correction code module. The battery management unit module sets a small-capacity SSD cache module; S2: The ZNS controller module aggregates the Blocks with the same block ID on all Dies within a Target into a ZNS partition to implement a fine-grained ZNS partition layout; S3: When a single Target fails, mark the failed Target. Fault isolation is achieved between the failed Target and other Targets based on the ZNS partition; When a single Die fails, mark the failed Die. Fault isolation is achieved between the failed Die and other Dies based on the ZNS partition; When a single Block fails, mark the partition where the failed Block is located. Fault isolation is achieved between the partition where the failed Block is located and other partitions based on the ZNS partition, and the distributed storage software quickly reconstructs the failed partition.

[0012] Preferably, while performing step S2, the setting of the hard disk reserved space OP is removed.

[0013] The beneficial effects of the present invention include: The method and system for implementing a highly reliable and large-capacity hard disk based on fine-grained ZNS partitioning provided by the present invention include a PCIe physical link module, a PCIe controller module, a ZNS controller module, a battery management unit module, a NAND controller module, a NAND physical link module, and a linear error correction code module. The battery management unit module is provided with a small-capacity SSD cache module; the ZNS controller module is used to implement a fine-grained ZNS partition layout; the battery management unit module manages a specified group of SSD caches, is used to receive write data from the host and save read data from the host; the linear error correction code module is used to encode and correct low-probability errors generated during the data reading and writing process; the NAND controller module and the NAND physical link module are used to process read and write requests for flash memory particles. By reducing the SSD hard disk partition size through fine-grained ZNS partitioning, refined fault isolation is achieved, and the SSD hard disk fault domain is reduced to improve the reliability of the hard disk.

[0014] First, through the system for implementing a highly reliable and large-capacity hard disk based on fine-grained ZNS partitioning, set up a PCIe physical link module, a PCIe controller module, a ZNS controller module, a battery management unit module, a NAND controller module, a NAND physical link module, and a linear error correction code module. The battery management unit module is provided with a small-capacity SSD cache module, streamlining the SSD hard disk control software and only retaining the ZNS-related command set; removing the hard disk memory and capacitors, as well as functions such as HWACC, reduces the complexity of the hard disk design.

[0015] Second, through the ZNS controller module, the Blocks with the same block ID on all Dies within a Target are aggregated into a ZNS partition to implement a fine-grained ZNS partition layout. Implementing fine-grained ZNS partitioning into the Die, the hard disk fault domain changes from a whole-disk fault to a partition fault, and the fault range is reduced by a hundred times; removing the SSD hard disk OP, a single ZNS partition fault only affects that partition, eliminating the probability of a whole-disk fault. As long as there are available ZNS partitions, the hard disk can still be used, improving the reliability of the SSD solid-state drive. Brief Description of the Drawings

[0016] Figure 1 It is a composition structure diagram of the system for implementing a highly reliable and large-capacity hard disk based on fine-grained ZNS partitioning of the present invention.

[0017] Figure 2 It is a schematic diagram of the ZNS partition layout of the present invention.

[0018] Figure 3 It is a composition structure diagram of the controller implementation system of the SSD hard disk in the prior art.

[0019] Figure 4 It is a schematic diagram of organizing NAND chips with a complex hierarchical architecture in the prior art. Detailed implementation manners

[0020] The following further elaborates on the present invention in conjunction with the Figures 1 to 4 accompanying drawings: Embodiment 1 Refer to the Figure 1 accompanying drawings. The high-reliability and large-capacity hard disk implementation system based on fine-grained ZNS partitioning includes a PCIe physical link module, a PCIe controller module, a ZNS controller module, a battery management unit module, a NAND controller module, a NAND physical link module, and a linear error correction code module. The battery management unit module is provided with a small-capacity SSD cache module. The PCIe physical link module and the PCIe controller module are used to process the PCIe protocol. The ZNS controller module is used to implement the fine-grained ZNS partitioning layout. The battery management unit module manages the specified group of SSD caches, and is used to receive the write data of the host and store the read data of the host. The linear error correction code module is used to encode and correct the low-probability errors generated during the data reading and writing process. The NAND controller module and the NAND physical link module are used to process the read and write requests of the flash memory particles.

[0021] In the controller design of the SSD hard disk in the prior art, refer to Figure 3 , which includes a front-end interface part, including a PCIe (Peripheral Component Interconnect Express, high-speed serial computer expansion bus standard) physical link, a PCIe controller, and an NVMe controller. This part processes the PCIe and NVMe protocols, receives the host data into the SSD hard disk cache, or sends the data in the cache to the host's memory. The battery management unit module (BMU, Buffer Management Unit) is responsible for the management of the SSD cache and includes a RAID engine. The NAND chip control module includes a NAND controller (a controller for processing NAND read and write) and a NAND physical link. This part realizes the control, management, and read and write of the NAND flash memory chip. The performance acceleration module (HWACC (Hardware Acceleration Compiler, hardware acceleration compiler), this part realizes hardware acceleration and function acceleration such as FTL look-up table and garbage collection), and this module mainly realizes acceleration functions such as look-up table and scanning. The CPU and memory module runs the SSD hard disk control software code.

[0022] Refer to Figure 4Schematic diagram of a complex hierarchical architecture organization NAND chip in the prior art, which increases the concurrency of the hard disk through multiple Channels (data transmission channels) and different hierarchical structures, and improves the capacity and performance of the SSD hard disk. The flash memory chip organization architecture of the solid-state drive of the present invention is as shown in the appendix Figure 2 As shown, an SSD hard disk usually contains multiple Packages, and one Package is the encapsulation of one NAND chip. One Package contains multiple Targets, each Target has an independent storage unit, data bus, and chip select signal. The Target is connected to the Channel of the SSD hard disk through the bus, and the Targets connected to the same Channel are selected through chip enable. Inside one Target, there are multiple Dies. A Die is the smallest unit that can execute commands and return status independently. One Die contains multiple Planes, each Plane contains multiple Blocks. A Block is the smallest unit for data erasure, that is, the smallest unit for garbage collection. Data can only be written in sequence within a Block. Inside one Block, there are multiple Pages. One Page is the smallest unit for single read and write. Then ZNS contains all the Blocks (flash physical blocks, which are the smallest units for data erasure) on all Dies. If there are more than redundant Dies offline in a RAID stripe, all ZNS partitions will fail completely, and fault isolation between partitions cannot be achieved.

[0023] Therefore, the high-reliability large-capacity hard disk implementation system based on fine-grained ZNS partitioning of the present invention aggregates the Blocks with the same block ID on all Dies within one Target into one ZNS partition, and such partition granularity is finer. It realizes fault isolation between partitions. The partitions are distributed on the Blocks within the same Target. A single Target failure or Die failure will only affect the partitions distributed on that Target or Die. A single Block failure will only affect the partition where the Block is located, and good fault isolation between partitions can be achieved. At the same time, the OP (Over-Provision, hard disk reserved space) design is abandoned. A single partition failure will only affect that partition, rather than causing the entire disk to fail, eliminating the correlation between Block failures and entire disk failures, and improving the reliability of the SSD hard disk.

[0024] As the partition size decreases, the impact of a single partition failure also decreases accordingly. Therefore, the system does not need to ensure the reliability of the partition, and there is no need to implement the RAID function within the partition. Instead, if a Block fails within the partition, the entire partition is marked as faulty, and the distributed storage software quickly reconstructs the faulty partition. This improves the write concurrency and, to a certain extent, enhances the write performance. The traditional ZNS partition uses a larger partition design and can only write sequentially within a partition. By using the WP to mark the position of the next IO write, although the partition spans multiple components that can work concurrently, the sequential write results in a single ZNS partition being able to handle fewer concurrent IOs and does not allow multiple components to work in parallel. Since the partition size is reduced, compared with the traditional ZNS partition design, the distributed storage software can operate more partitions simultaneously and enable more Dies to write concurrently.

[0025] Embodiment 2 Based on Embodiment 1, the PCIe physical link module is connected to the PCIe controller module, the PCIe controller module is connected to the ZNS control module, the ZNS controller module is connected to the small-capacity SSD cache module, the small-capacity SSD cache module is connected to the NAND controller module, the NAND controller module is connected to the NAND physical link module, and the linear error correction code module is connected to the small-capacity SSD cache module.

[0026] In this embodiment, the fine-grained ZNS partition layout implemented by the ZNS controller module aggregates Blocks with the same block ID on all Dies within a Target into a ZNS partition. Through the process of fine-grained ZNS partitioning, the ZNS controller module realizes fault isolation between fine-grained ZNS partitions. The ZNS partitions are distributed on the Blocks within the same Target. A single Target failure or Die failure only affects the ZNS partitions distributed on the Target or the Die, and a single Block failure only affects the ZNS partition where the Block is located, removing the correlation between Block failures and whole-disk failures.

[0027] Embodiment 3 Based on Embodiment 1 or Embodiment 2, a method for implementing a highly reliable large-capacity hard disk based on fine-grained ZNS partitioning is implemented based on the system for implementing a highly reliable large-capacity hard disk based on fine-grained ZNS partitioning according to any one of the above, and includes the following steps: S1: Deploy an SSD hard disk controller in the server. The SSD hard disk controller deploys a PCIe physical link module, a PCIe controller module, a ZNS controller module, a battery management unit module, a NAND controller module, a NAND physical link module, and a linear error correction code module. The battery management unit module is provided with a small-capacity SSD cache module; S2: The ZNS controller module aggregates Blocks with the same block ID on all Dies within a Target into a ZNS partition to implement a fine-grained ZNS partition layout. At the same time, the setting of the hard disk reserved space OP is removed; S3: When a single Target fails, mark the failed Target. Fault isolation is achieved between the failed Target and other Targets based on the ZNS partition; When a single Die fails, mark the failed Die. Fault isolation is achieved between the failed Die and other Dies based on the ZNS partition; When a single Block fails, mark the partition where the failed Block is located. Fault isolation is achieved between the partition where the failed Block is located and other partitions based on the ZNS partition, and the distributed storage software quickly reconstructs the failed partition.

[0028] In this embodiment, by streamlining the SSD hard disk control software, only the ZNS-related command set is retained; functions such as the hard disk memory, capacitor, and HWACC are removed to reduce the complexity of the hard disk design. Implement a fine-grained ZNS partition into the Die. The hard disk failure domain changes from a whole-disk failure to a partition failure, and the failure range is reduced by a hundred times; the SSD hard disk OP is removed, and the failure of a single ZNS partition only affects that partition, eliminating the probability of a whole-disk failure. As long as there are available ZNS partitions, the hard disk can still be used.

[0029] In summary, the high-reliability large-capacity hard disk implementation method based on fine-grained ZNS partitioning provided by the present invention includes a PCIe physical link module, a PCIe controller module, a ZNS controller module, a battery management unit module, a NAND controller module, a NAND physical link module, and a linear error correction code module. The battery management unit module is provided with a small-capacity SSD cache module; the ZNS controller module is used to implement a fine-grained ZNS partition layout; the battery management unit module manages a specified group of SSD caches, is used to receive write data from the host and save read data from the host; the linear error correction code module is used to encode and correct low-probability errors generated during the data reading and writing process; the NAND controller module and the NAND physical link module are used to process read and write requests for flash memory particles. By reducing the SSD hard disk partition size through fine-grained ZNS partitioning, refined fault isolation is achieved, and the SSD hard disk fault domain is reduced to improve the hard disk reliability.

Claims

1. A high-reliability, large-capacity hard disk implementation system based on fine-grained ZNS partitioning, characterized in that: It includes a PCIe physical link module, a PCIe controller module, a ZNS controller module, a battery management unit module, a NAND controller module, a NAND physical link module and a linear error correction code module, wherein the battery management unit module is provided with a small-capacity SSD cache module; The PCIe physical link module and the PCIe controller module are used to process the PCIe protocol; The ZNS controller module is used to implement fine-grained ZNS partition layout; The battery management unit module manages a specified group of SSD caches for receiving write data from the host and storing read data from the host; The linear error correction code module is used to encode and correct low-probability errors generated during data reading and writing; The NAND controller module and the NAND physical link module are used to process read and write requests of flash memory particles.

2. According to claim 1, the high-reliability large-capacity hard disk implementation system based on fine-grained ZNS partitioning is characterized in that: The PCIe physical link module is connected to the PCIe controller module, the PCIe controller module is connected to the ZNS control module, the ZNS controller module is connected to the small-capacity SSD cache module, the small-capacity SSD cache module is connected to the NAND controller module, the NAND controller module is connected to the NAND physical link module, and the linear error correction code module is connected to the small-capacity SSD cache module.

3. The high-reliability large-capacity hard disk implementation system based on fine-grained ZNS partitioning according to claim 1 is characterized in that: The fine-grained ZNS partition layout implemented by the ZNS controller module aggregates the blocks with the same block ID on all Dies in a Target into a ZNS partition.

4. The high-reliability large-capacity hard disk implementation system based on fine-grained ZNS partitioning according to claim 3 is characterized in that: The ZNS controller module implements fault isolation between fine-grained ZNS partitions through the process of fine-grained ZNS partitioning. ZNS partitions are distributed on blocks within the same Target. A single Target failure or Die failure only affects the ZNS partitions distributed on the Target or the Die. A single Block failure only affects the ZNS partition where the Block is located, thereby removing the correlation between Block failure and entire disk failure.

5. A method for implementing a high-reliability large-capacity hard disk based on fine-grained ZNS partitioning, which is implemented based on the high-reliability large-capacity hard disk implementation system based on fine-grained ZNS partitioning according to any one of claims 1 to 4, characterized in that: The following steps are involved: S1: Deploy an SSD hard disk controller in the server, wherein the SSD hard disk controller deploys a PCIe physical link module, a PCIe controller module, a ZNS controller module, a battery management unit module, a NAND controller module, a NAND physical link module and a linear error correction code module, and the battery management unit module is provided with a small-capacity SSD cache module; S2: The ZNS controller module aggregates the blocks with the same block ID on all Dies in a Target into a fine-grained ZNS partition layout implemented by a ZNS partition; S3: When a single Target fails, the failed Target is marked as a fault, and the fault isolation is achieved between the failed Target and other Targets based on ZNS partitions; When a single Die fails, the failed Die is marked as a fault, and the fault isolation is achieved between the failed Die and other Dies based on ZNS partitioning; When a single Block fails, the partition where the failed Block is located is marked as a fault, and the partition where the failed Block is located is isolated from other partitions based on the ZNS partition, and the distributed storage software quickly reconstructs the faulty partition.

6. The method for realizing a high-reliability and large-capacity hard disk based on fine-grained ZNS partitioning according to claim 5 is characterized in that: At the same time as step S2, the setting of the hard disk reserved space OP is removed.

Images