WO2013023564A1

WO2013023564A1 - Method and apparatus for flexible raid in ssd

Info

Publication number: WO2013023564A1
Application number: PCT/CN2012/080038
Authority: WO
Inventors: Yiren Huang
Original assignee: Huawei Technologies Co., Ltd.
Priority date: 2011-08-12
Filing date: 2012-08-13
Publication date: 2013-02-21
Also published as: WO2013023564A9

Abstract

A solid state drive (SSD) employing a redundant array of independent disks (RAID) scheme includes a flash memory chip, erasable blocks in the flash memory chip, and a flash controller. The erasable blocks are configured to store flash memory pages. The flash controller is operably coupled to the flash memory chip. The flash controller is also configured to organize certain of the flash memory pages into a RAID line group and to write RAID line group membership information to each of the flash memory pages in the RAID line group.

Description

Method and Apparatus for Flexible RAID in SSD

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This patent application claims priority to U.S. Provisional Application No. 61/523,251, filed on August 12, 2011, entitled "Method and Apparatus for Flexible RAID in SSD," and claims the benefit of U.S. Patent Application Serial No. 13/460686 filed April 30, 2012, entitled "Method and Apparatus for Flexible RAID in SSD", which are incorporated herein by reference as if reproduced in their entirety.

FIELD OF INVENTION

[0002] The present disclosure relates to memories, and, in particular embodiments, to a method and apparatus for flexible RAID in SSD.

BACKGROUND

[0003] Non- volatile memories have been used as memories in computers and portable information apparatuses. Recently, solid state drives (SSD) using NAND flash memory have gained in popularity for use in computers as an alternative to using simply a hard disk drive (HDD). SSD is considered to have great potential for making inroads in areas of use traditionally considered as limited to HDDs.

[0004] Solid State Drive (SSD) has lately been more and more adopted to be used in computer systems either as cache of Hard Disk Drive (HDD) or as a direct replacement of HDD. In such new architectures, the SSD are increasingly used to a) increase access speed to stored (or cached) data, b) reduce power consumption of HDD, c) reduce the size, weight, and power consumption profile of the system, and d) reduce the access latency to the stored (or cached) data. [0005] NAND flash memory based SSD (solid state drive) has been widely adopted in various applications where data access speed is needed. The traditional RAID (redundant array of independent disks) scheme has been adopted in many SSD design. Such traditional RAID schemes, however, can only work if all elements in the RAID group are good.

SUMMARY

[0006] Technical advantages are generally achieved by embodiments of the present disclosure which provide a system and method for flexible RAID in SSD.

[0007] In an embodiment, a solid state drive (SSD) employing a redundant array of independent disks (RAID) scheme includes a flash memory chip, erasable blocks in the flash memory chip, and a flash controller. The erasable blocks are configured to store flash memory pages. The flash controller is operably coupled to the flash memory chip. The flash controller is also configured to organize certain of the flash memory pages into a RAID line group and to write RAID line group membership information to each of the flash memory pages in the RAID line group.

[0008] In an embodiment, a solid state drive (SSD) employing a redundant array of independent disks (RAID) scheme includes a flash memory chip, erasable blocks, and a flash controller. The erasable blocks are in the flash memory chip and configured to store one or more flash memory pages. The flash controller is operably coupled to the flash memory chip. The flash controller is configured to organize certain of the flash memory pages into a RAID line group, to exclude the flash memory pages that are bad from the RAID line group, and to write RAID line group membership information in the form of a binary/hexadecimal number to a RAID line (RAIDL) field within each of the flash memory pages in the RAID line group. [0009] In an embodiment, a method of using a redundant array of independent disks (RAID) scheme for a solid state drive (SSD) includes organizing flash memory pages of an array of flash memory chips into a RAID line group and writing RAID line group membership information to each of the flash memory pages in the RAID line group. The RAID line group information identifies flash memory pages to be excluded during at least one of parity calculations and reconstruction of corrupt flash memory pages.

BRIEF DESCRIPTION OF THE DRAWINGS

[0010] For a more complete understanding of the present disclosure, and the advantages thereof, reference is now made to the following descriptions taken in conjunction with the accompanying drawings, in which:

[0011] Figure 1 provides an embodiment solid state drive (SSD);

[0012] Figure 2 illustrates examples of flash memory pages organized in a redundant array of independent disks (RAID) line group and capable of being stored in erasable blocks of flash memory chips in the SDD of Figure 1;

[0013] Figure 3 provides an illustration of a data format for the flash memory pages of Figure 2, the data field including a RAID line (RAIDL) field;

[0014] Figure 4 provides an example of the RAID line group where one of the flash memory pages is corrupt and two of the flash memory pages are excluded from the RAID line group for reconstruction purposes;

[0015] Figure 5 provides an example of the RAIDL field from the data format of Figure 3;

[0016] Figure 6 provides an example where the RAID line group of Figure 2 has been divided into first and second RAID line groups; [0017] Figure 7 provides an example of first and second RAIDL fields corresponding to the first and second RAID line groups of Figure 6;

[0018] Figure 8 provides a block diagram of an embodiment system using the SSD of Figure 1 and that can implement embodiments of the present disclosure; and

[0019] Figure 9 provides an embodiment of a method of using a RAID scheme for the SSD of Figure 1.

DETAILED DESCRIPTION

[0020] The making and using of the present embodiments are discussed in detail below. It should be appreciated, however, that the present disclosure provides many applicable inventive concepts that may be embodied in a wide variety of specific contexts. The specific embodiments discussed are merely illustrative and do not limit the scope of the disclosure.

[0021] Lately, the solid state drive (SSD), which may also be referred to as a solid-state disk or electronic disk, has been incorporated in computer systems either as the cache of a Hard Disk Drive (HDD) or as a direct replacement of the HDD. In such new architectures, the SSD is increasingly used to improve access speed to stored (or cached) data, to reduce power consumption of the HDD, to reduce the size, weight, and power consumption profile of the system, and to reduce the access latency to the stored (or cached) data.

[0022] However, the SSD, which is a NAND flash or other non-volatile memory type, has been error-prone. Therefore, a bit stored therein can easily become corrupted. Also, the error rate increases as the program erase cycles of the SSD increase. As such, even when advanced error correction coding (ECC) algorithms (e.g., the Bose Ray-Chaudhuri (BCH) ECC algorithm) have been used, there is still risk that a unit of data will not be correctable. [0023] As will be more fully explained below, embodiments of this disclosure enable the SSD to flexibly organize flash memory pages, erasable blocks, and/or flash memory chips into a redundant array of independent disks (RAID) group. The RAID line group members may be chosen or selected based on, for example, the state of each silicon die, the write arrival pattern, or some other factor or factors that influence the membership of the RAID line group when data is actually written into the storage media. In other words, membership in the RAID line group is flexible. The members of the RAID line group may be identified by RAID group membership information, which may be written along with data to be stored. When a group member is read and determined to be corrupt by the ECC algorithm, the next neighbor where the RAID line group members reside may be read back and used to reconstruct the corrupt group member using, for example, an exclusive or (XOR) process Analysis shows that such a method may greatly enhance the reliability of the SSD and extend the drive life.

[0024] Referring now to Figure 1, an embodiment SSD 10 utilizing a redundant data storage mechanism or scheme similar to hard drive based traditional RAID (redundant array of inexpensive (or independent) disks) is illustrated. As shown, the SSD 10 generally includes a flash controller 12 and flash memory chips 14, each of which may include several memory dies. The flash controller 12 (a.k.a., NAND flash controller or SSD controller) functions to, among other things, bridge the flash memory chips 14 to the computer hosting or using the SSD 10. In an embodiment, the flash controller 12 is an embedded processor that executes firm-ware level code. The flash controller 12 may, among other things, perform ECC, wear leveling, bad block mapping, read scrubbing and read disturb management, read and write caching, garbage collection, and encryption. [0025] Still referring to Figure 1, in an embodiment the flash memory chips 14 may be nonvolatile NAND flash memory chips. Even so, other types of flash memory chips such as dynamic random-access memory (DRAM) may also be used. In addition, the flash memory chips 14 may be multi-level cell (MLC) flash memory or single-level cell (SLC) flash memory. While two of the flash memory chips 14 are illustrated in Figure 1, more or fewer of the flash memory chips 14 may be included in the SSD 10.

[0026] In an embodiment, the flash memory chips 14 are divided into erasable blocks 16 (a.k.a., flash memory blocks or RAID blocks). A flash erase block is a minimum unit in NAND flash memory that may be erased at one time. While two of the erasable blocks 16 are illustrated in each of the flash memory chips 14 of Figure 1, more or fewer of the erasable blocks 16 may be included in each of the flash memory chips 14. The erasable blocks 16 are generally configured to store flash memory pages 18 of data, which are illustrated in Figure 2.

[0027] For the purpose of illustration, seventeen (17) of the flash memory pages 18 of data are depicted in Figure 2. As shown, the flash memory pages 18 have been organized, written, or otherwise arranged by the flash controller 12 into a RAID line group 20. The seventeen flash memory pages 18 included in the RAID line group 20 have been designated Page0-Page9, PageA-PageF, and PageP. PageP represents a parity page. Each RAID line group 20 will have at least one parity page to provide redundancy to the RAID line group. Indeed, in case any member is corrupted, the parity page may be used to recover the failed member page using other member pages in the RAID line group and, for example, an XOR operation.

[0028] Because the RAID line group 20 includes seventeen flash memory pages that are written at one time and include one parity page, the RAID line group 20 may be referred to as a 16+1 RAID group. Notably, membership in the RAID line group 20 is flexible. In other words, the RAID line group 20 may include more or fewer flash memory pages 18 available within the erasable block 16 or on the flash memory chip 14.

[0029] Still referring to Figure 2, some of the flash memory pages 18 from the RAID line group 20 are known to be bad (e.g., corrupt, contain an error, have a program failed status, read unsuccessfully due to bit errors or a page read time out, etc.). There may be several cases when a flash memory page 18 is determined to be bad. For example, the flash memory page 18 from the erasable block 16 (a.k.a., NAND flash memory block) may have gone bad due to a bus malfunction or a flash memory chip 14 malfunction.

[0030] In the embodiment of Figure 2, Page9 has been detected as bad during a read process. Because Page9 has been detected as bad, Page9 will not participate in the parity calculation used to derive PageP. Page9 will also not be used to reconstruct corrupt flash memory pages 18 discovered or identified during a subsequent read operation. In other words, Page9 will remain in the physical RAID line group 20 but will be effectively excluded from the RAID line group 20 when the parity calculations are made and when recovering the data of later discovered failed flash memory pages. In an embodiment, the memory bus for Page9 is passed over or skipped to exclude Page9 as described herein.

[0031] In an embodiment, where one or more of the flash memory pages 18 encounters a program fail, such as Page9 in the above example, one option is to remove the failed flash memory page from the RAID line group 20 and then rewrite the RAID line group 20 to the next RAID line address if still using the same erasable block 16. Such an option presumes that the failed flash memory page 18 (e.g., Page9) is likely to fail again in the same erasable block 16. The treatment for parity page program failure will be the same as the regular page program failure. Although one RAID line may not be used to full potential in this scenario, because failure of flash memory pages 18 due to a program failure is infrequent, the impact on the SSD 10 will be very small. In this case, the write formatter may use the copy back command for all other pages except the parity page.

[0032] Referring now to Figure 3, one of the flash memory pages 18 from Figure 2 has been illustrated in further detail. As shown in Figure 3, the flash memory page 18 has been divided into a logical block addressing (LBA) sector 22 and a RAID line (RAIDL) field 24. The LBA sector 22 stores the original data or the meta data corresponding to the original data. The meta data stored in the LBA sector 22 may be associated with the original data, used to protect the original data, or special to the original data (e.g., indicate the LBA address of the original data).

[0033] The RAIDL field 24 is employed to store the RAID line group membership information 26, which identifies the flash memory pages 18 in the physical RAID line group 20 that will be used in parity calculations and/or used in the reconstruction of flash memory pages that later become corrupt. The RAID line group membership information 26 stored in the RAIDL field 24 may also be used to identify which of the flash memory pages 18 will be excluded from the RAID line group 20 for the purpose of parity calculations and recovery or reconstruction of corrupt flash memory pages. In an embodiment, the RAID line group membership information 26 is stored in the RAIDL field 24 as a binary/hexadecimal number. In other embodiments, the RAID line group membership information 26 may be stored in other formats.

[0034] As example of how the RAID line group membership information 26 stored in the RAIDL field 24 may be used in an embodiment to identify the flash memory pages used in parity calculations and flash memory page reconstruction efforts is illustrated in Figures 4-5. As shown in Figure 4, the flash controller 12 has organized or written certain flash memory pages 18 into a RAID line group 20. As shown, a total of seventeen flash memory pages 18 from the erasable block 16 on the flash memory chip 14 have been selected. Of the flash memory pages 18 in the RAID line group 20 of Figure 4, two of the flash memory pages 18 are known to be bad, namely Page9 and PageE. Therefore, these pages will be excluded from parity calculations and flash memory page data reconstruction efforts.

[0035] In order to indicate that the flash memory pages Page9 and PageE are known to be bad, RAID line group membership information 26 is written by the flash controller 12 into the RAIDL field 24 as shown in Figure 5. In the illustrative example of Figure 5, because the flash memory pages Page9 and PageE are known to be bad, the hexadecimal number represented as 17'hlbdff is stored in the RAIDL field 24. As shown in Figure 5, when the hexadecimal number is converted to binary, the bits associated with Page9 and with PageE are not set. Therefore, these bits are each a "0." In contrast, the bits associated with the flash memory pages 18 that will be used in parity calculations using the XOR process and flash memory page reconstruction are set. Therefore, these bits are each a "1."

[0036] Still referring to Figures 4-5, if during a subsequent read operation Page5 is discovered to be bad or corrupt due to, for example, bit errors, the flash memory pages 18 included in the RAID line group 20, with the exception of Page5, are read out. In other words, the flash memory pages 18 designated PageO, Page2, Page2, Page3, Page4, Page6, Page7, Page8, PageA, PageB, PageC, PageD, and PageF are read. Once read, these flash memory pages are used along with the parity page, PageP, in an XOR process to reconstruct the flash memory page Page5. Notably, the known bad flash memory pages 18 designated Page9 and PageE and the newly discovered bad flash memory page designated Page5 are not used to reconstruct the flash memory page designated Page5. In addition, if Page5 cannot, for some reason, be recovered or reconstructed using the above process due to, for example, an uncorrectable error, then the RAIDL field 24 may be updated such that the bit associated with Page5 is set to "0."

[0037] In an embodiment as shown in Figures 6-7, the RAID line group 20 may be divided into smaller RAID line groups, for example, a first smaller RAID line group 28 and a second smaller RAID line group 30. In other words, flexible RAID techniques may be used to break down the number of pages generally required by a fixed RAID group. In such embodiments, the RAIDL field 24 for the first smaller RAID line group 28 may be populated with the hexadecimal number is represented as 17h'000ff and the RAIDL field 24 for the second smaller RAID line group 30 may be populated with the hexadecimal number represented as 17'hlbd00. As shown in Figure 7, the RAID line group membership information 26 stored in the RAIDL fields is complementary and collectively identifies the flash memory pages 18 included in the RAID line group 20. Indeed, the first eight flash memory pages 18 (Page0-Page7) are included in the first smaller RAID line group 28 (where Page5 may be discovered to be bad and later excluded) and the next eight flash memory pages 18 are included in the second smaller RAID line group 30 (where Page9 and PageE will be excluded from parity/reconstruction calculations). In this embodiment, one of the flash memory pages in the first smaller RAID line group 28 (e.g., Page7) may have to operate as a parity page similar to PageP.

[0038] Referring now to Figure 8, the SSD may be incorporated into a computer system 32. The computer system 32 may be utilized to implement methods of the present disclosure. In this case, the main processing may be performed in a processor 34, which may be a microprocessor or any other appropriate processing device. Program code (e.g., the code implementing the algorithms or methods disclosed above) and data may be stored in a random access memory 36. The memory 36 may be local memory such as DRAM or mass storage such as a hard drive, optical drive, or other storage (which may be local or remote). While the memory 36 is illustrated functionally with a single block, it is understood that one or more hardware blocks may be used to implement this function.

[0039] In one embodiment, the processor 34 may be used to implement various ones (or all) of the functions discussed above. For example, the processor 34 may serve as a specific functional unit at different times to implement the subtasks involved in performing the techniques of the present disclosure. Alternatively, different hardware blocks (e.g., the same as or different than the processor 34) may be used to perform different functions. In other embodiments, some subtasks are performed by the processor 34 while others are performed using a separate circuitry.

[0040] Figure 4 also illustrates an input/output (I O) 38, which may be used to provide information to the system 32. This information could include, for example, video to be stored in the SSD 10. The I/O 38 may include ports for connection to a network, including the Internet or local interfaces (e.g., a USB or LAN interface). The I/O 38 can also include local user operated I/O device such as display, keyboard, mouse, and so on. In other words, the single box is provided to represent a potentially large number of hardware items.

[0041] The SSD 10 and other memory 40 (e.g., HDD, optical drive, or other) are shown connected to the same bus as the I/O 38. This is but one example of an architecture that may be suitably used. For purposes of simplicity, control circuitry, such as a memory controller or bridge, has been omitted from the drawing even though a practical system will include additional circuitry as appropriate.

[0042] Referring now to Figure 9, an embodiment of a method 42 of is illustrated. In block 44, selected ones of the flash memory pages 18 of the flash memory chip 14 are organized into the RAID line group 20. In block 46, the RAID line group membership information 26 is written to each of the flash memory pages 18 in the RAID line group 20. As noted above, the RAID line group member information 26 identifies the flash memory pages 18 to be excluded during parity calculations and/or reconstruction of corrupt flash memory pages 18. In an embodiment, the flash controller 12 or the processor 34 performs or coordinates these steps.

[0043] Using the flexible RAID scheme, algorithm, or processes disclosed herein, single or multiple flash memory page 18 failures may be tolerated and permit the SSD 10 to continue working properly. In addition, where the flexible RAID techniques are applied to the erasable blocks 16 or the flash memory chips 14 in the SSD 10, failures may also be tolerated and permit the SSD 10 to continue working properly.

[0044] Embodiments of the present disclosure provide a number of new and advantageous features. For example, one feature of this scheme is to use a RAID membership to indicate the member of a RAID line group and therefore allow flexible grouping of a RAID line. Such scheme allows tolerance of one or multiple members of a RAID group to malfunction without having to abandon the remaining good storage media. This scheme may be applied to the latest NAND flash based SSDs. Such scheme may also be applied to other storage media such as, for example, hard disk drive based RAID systems.

[0045] Embodiments of the present disclosure may be used in a number of products, processes, and services. For example, embodiments could be used in next generation Peripheral Component Interconnect Express (PCIE) SSD to improve the utilization of the SSD and life span of the SSD as a result. The flexible membership is superior to existing fixed RAID algorithms. This flexible RAID technique is advantageous in a number of markets including, but not limited to, enterprise storage systems. [0046] As discussed above, embodiments have a number of benefits. By storing the data using RAID mechanism in SSD, a single page data loss may be recovered. Embodiments of the flexible RAID algorithm as described herein provide a way to flexibly decide the number of members in a RAID group. This enables the SSD to tolerate one or more members in RAID to fail and still able to recover data.

[0047] Benefits are achieved by having a flexible membership of the RAID algorithm to allow the design to tolerate single or multiple pages, blocks, or silicon dies to fail in SSD and still able to recover data lost.

[0048] In a disk drive based storage system, the RAID mechanism may be implemented to prevent one single disk to fail such as RAID 4 or RAID 5 mechanism by using one parity entry in a RAID group. In one embodiment, this disclosure calculates and stores the RAID membership along with data such that in case one device in RAID group fails, it may be skipped from the RAID membership when data is stored to the solid state drive. This way, when one member failed over time, the RAID membership may be read from the other pages in the RAID group and only these members in the RAID group will be used to determine the recovery by RAID algorithm.

[0049] While the disclosure has been made with reference to illustrative embodiments, this description is not intended to be construed in a limiting sense. Various modifications and combinations of the illustrative embodiments, as well as other embodiments, will be apparent to persons skilled in the art upon reference to the description. It is therefore intended that the appended claims encompass any such modifications or embodiments.

Claims

1. A storage device employing a redundant array of independent disks (RAID) scheme, comprising:

a flash memory chip;

erasable blocks in the flash memory chip, the erasable blocks configured to store one or more flash memory pages; and

a flash controller operably coupled to the flash memory chip, the flash controller configured to organize certain of the flash memory pages into a RAID line group and to write RAID line group membership information to each of the flash memory pages in the RAID line group.

2. The drive of claim 1, wherein each of the flash memory pages includes a RAID line (RAIDL) field configured to store the RAID line group membership information.

3. The storage device of claim 2, wherein the RAIDL field is used to indicate the flash memory pages in the RAID line group used for parity calculations.

4. The storage device of claim 2, wherein the RAIDL field is used to indicate the flash memory pages in the RAID line group used for page reconstruction.

5. The storage device of claim 2, wherein the RAIDL field is used to indicate the flash memory pages excluded from the RAID line group during at least one of parity calculations and page reconstruction.

6. The storage device of claim 2, wherein the RAIDL field is used to reconstruct a subsequent bad flash memory page in the RAID line group.

7. The storage device of claim 2, wherein the RAIDL field stores a hexadecimal number, the hexadecimal number used to indicate the flash memory pages included during at least one of parity calculations and page reconstruction.

8. The storage device of claim 2, wherein the RAIDL field stores a hexadecimal number, the hexadecimal number used to indicate the flash memory pages excluded during at least one of parity calculations and page reconstruction.

9. The storage device of claim 2, wherein each of the flash memory pages includes a logical block addressing (LBA) sector, the LBA sector configured to store at least one of data and meta data.

10. The storage device of claim 1, wherein at least one of the flash memory pages in the RAID line group is a parity block.

11. The storage device of claim 1, wherein the flash controller is configured to divide the RAID line group into smaller RAID line groups having complementary RAID line group membership information collectively forming the RAID line group membership information. A storage device employing a redundant array of independent disks (RAID) scheme, comprising: a flash memory chip;

a flash controller operably coupled to the flash memory chip, the flash controller configured to organize certain of the flash memory pages into a RAID line group and to write RAID line group membership information to a RAID line (RAIDL) field within each of the flash memory pages in the RAID line group.

12. The storage device of claim 12, wherein the flash controller is configured to exclude the flash memory pages that are bad from the RAID line group.

13. The storage device of claim 12, wherein the RAID line group membership information comprises a hexadecimal number.

14. The storage device of claim 14, wherein the hexadecimal number of the RAIDL field identifies which of the flash memory pages is included in the RAID line group and which of the flash memory pages is excluded from the RAID line group for parity calculations and for reconstruction of corrupt flash memory pages.

15. The storage device of claim 14, wherein the hexadecimal number of the RAIDL field is employed to reconstruct a failed one of the flash memory pages in the RAID line group discovered during a read operation of the flash memory pages in the RAID line group.

16. The storage device of claim 12, wherein at least one of the flash memory pages in the RAID line group is a parity block.

17. A method of using a redundant array of independent disks (RAID) scheme for a solid state drive (SSD), comprising:

organizing flash memory pages of a flash memory chip into a RAID line group; and writing RAID line group membership information to each of the flash memory pages in the RAID line group, the RAID line group information identifying flash memory pages to be excluded during at least one of parity calculations and reconstruction of corrupt flash memory pages.

18. The method of claim 18, further comprising reconstructing the corrupt flash memory pages in the RAID line group using the RAID line group membership information.

19. The method of claim 18, wherein writing RAID line group membership information to each of the flash memory pages in the RAID line group comprises writing to a RAID line (RAIDL) field in each of the flash memory pages of the RAID line group.

20. The method of claim 18, wherein writing RAID line group membership information to each of the flash memory pages in the RAID line group comprises writing a hexadecimal number to a RAID line (RAIDL) field in each of the flash memory pages of the RAID line group.

21. The method of claim 21, further comprising identifying the flash memory pages in the RAID line group and the flash memory pages excluded from the RAID line group using the hexadecimal number.