TWI644207B - Method for garbage collection of data storage device - Google Patents

Abstract

A data storage device uses a garbage collection method. The data storage device includes a NAND flash memory, and the NAND flash memory includes a plurality of data blocks, each of which includes a plurality of data pages. In the garbage collection method, a destination data block is selected from the data blocks. Before the user data is written into the destination data block, a number of mapping tables are created for the destination data to record the relationship between the logical address of the user data and the physical address, and an associated bitmap is created. Each bit in the associated bit map corresponds to a mapping table. Then, select a block of data to be recycled. Then, the relevant mapping table is read according to the associated bit map of the to-be-recovered data block. Then, it is judged one by one whether all the data pages of the to-be-recovered data block are in the relevant mapping table. If a data page of the to-be-recovered data block is in the related mapping table, the data page is set as a valid data page. Then, the data of the valid data pages is written into the destination data block.

Description

Method for garbage collection of data storage device

The invention relates to a data storage device, in particular to a garbage collection method for a data storage device.

Flash Translation Layer (FTL) refers to the firmware algorithm implemented in flash memory. Flash memory is divided into NOR and NAND flash memory. The following mainly discusses the FTL (or NFTL) of NAND flash memory. The flash memory of NAND flash memory has three physical characteristics regardless of SLC (Single-Level Chip) or MLC (Multi-Level Chip). First, the written address must be erased before it can be written. Second, the unit that reads or writes once is small and the unit that is erased at one time is large. For example, the unit of one read or write is a page and the unit erased at one time is a block. Typically, a data block includes a number of data pages. Third, the number of erasures per data block is limited. For example, the limit of the number of erasures of the SLC is about 100,000 times, and the limit of the number of erasures of the MLC is about 10,000 times.

As shown in Figure 1, an FTL manages a NAND flash memory and connects to a file system. The NAND flash memory is divided into a number of data blocks.

As shown in Figure 2, each data block has several data pages and a number of spare parts (Spare). As mentioned above, for FTL, the smallest unit of read and write is a data page, and the smallest unit of erasure is a data block. Data page storage User information. The spare part, also known as Metadata, stores information about the corresponding data page, such as the logical location and ECC information corresponding to the data page.

The main job of FTL is to handle the mapping between Logic Address and Physical Address. Therefore, FTL will use some forms to record these relationships for easy table lookup and writing. After the FTL is written, the table is updated. The FTL configures the tablespace in the memory and, after filling the form, flushes the data from the table to the NAND flash memory. The data blocks of NAND flash memory are simply divided into data blocks and table blocks. The data block stores user data and table block storage form.

Figures 3 through 5 show a simple example of writing and recycling. The table in the figure is the correspondence between the physical page number (Physical Page Number: PPN) and the data.

As shown in Figure 3, data block 1000 has three pieces of data x, y, and z. Data block 2000 is a free data block with no data in it.

As shown in Fig. 4, the data of x is updated, that is, a new data x' is written to the data page whose PPN is 3 for the same logical address (not shown). At this time, the data of the data page whose PPN is 0 becomes invalid data. When the data block 1000 is filled, it contains valid data and invalid data.

As shown in FIG. 5, before the data block 1000 is recycled, the valid data page of the data block 1000 is written to the data block 2000. The data block 1000 is then erased and changed back to the free data block.

When the available space of the NAND flash memory drops to a water level, the FTL performs Garbage Collection (GC). There are several steps in garbage collection. First, select a victim block to be recovered from a Data Block Area (DBA), such as shown in Figure 4. Data block 1000. Next, a destination data block is selected, such as data block 2000 shown in FIG. Then, the valid data page in the to-be-recovered data block is moved to the destination data block. Finally, erase and release the data block to be recycled. This process is executed until the number of free data blocks returns above another water level.

To determine whether a data page in the data block to be recycled is a valid data page or an invalid data page, the method is to check whether the physical address of the data page is in a mapping table. If yes, it means that the data page is a valid data page.

When the controller uses a small size of RAM, FTL cannot store all the mapping tables in RAM, so that it can only temporarily cache (CACHE) a certain number of mapping tables in RAM. There are two ways to compare entity addresses with mapping tables.

The first approach is that FTL writes all mapping tables one by one from the NAND flash memory to RAM and compares its contents to the physical address. For example, if there are M mapping tables in the NAND flash memory, the FTL must take M times the loading time of the mapping table. This is a long time.

The second method is that the FTL reads the logical addresses from the metadata of each data page of the data block to be recovered, and the logical addresses know that the mapping tables should be read from the NAND flash memory. For example, if a data block has a K data page and the L mapping table needs to be read from the NAND flash memory, the FTL must spend K times the data reading time plus the mapping table loading time L times. L is less than or equal to K. This time is not short.

In view of the above problems and deficiencies of the prior art, the object of the present invention is to provide a data storage device with a garbage requiring only a small amount of resources. Recycling method.

To achieve the above object, the data storage device includes a NAND flash memory, and the NAND flash memory includes a plurality of data blocks, each of which includes a plurality of data pages. In the garbage collection method, a destination data block is selected from the data blocks. Before the user data is written into the destination data block, a number of mapping tables are created for the destination data to record the relationship between the logical address of the user data and the physical address, and an associated bitmap is created. Each bit in the associated bit map corresponds to a mapping table. Then, select a block of data to be recycled. Then, the relevant mapping table is read according to the associated bit map of the to-be-recovered data block. Then, it is judged one by one whether all the data pages of the to-be-recovered data block are in the relevant mapping table. If a data page of the to-be-recovered data block is in the related mapping table, the data page is set as a valid data page. Then, the data of the valid data pages is written into another data block.

10‧‧‧Host

12‧‧‧SSD Controller

14‧‧‧SSD memory

16‧‧‧NAND flash memory

18‧‧‧ processor

20‧‧‧ Effective data page bitmap

22‧‧‧ mapping table

24‧‧‧ Relevance bit map

26‧‧‧Data buffer

28‧‧‧Table index one-dimensional array

30‧‧‧Information block

32‧‧‧Form block

34‧‧‧Form Index Block

1 is a block diagram of a typical data storage system; FIG. 2 is a simplified block diagram of a NAND flash memory of the data storage system shown in FIG. 1; FIGS. 3 to 5 are a second diagram A block diagram of two data blocks of the NAND flash memory in different states; FIG. 6 is a block diagram of the data storage system of the preferred embodiment of the present invention in a writing subroutine; The block diagram of the data storage system shown in the figure in the recycling subroutine; Figure 8 is a flow chart of a writing subroutine executed by the data storage system shown in Figure 6; and Figure 9 is a flow chart of the recycling subroutine executed by the data storage system shown in Figure 7; and Figure 10 It is a flowchart of a part of the recycling subroutine shown in Fig. 9.

Hereinafter, embodiments of the garbage collection method of the data storage device of the present invention will be further described with reference to the related drawings, and the same components are denoted by the same reference numerals.

As shown in FIGS. 6 and 7, a data storage system includes a host 10, an SSD controller 12, an SSD memory 14 and a NAND flash memory 16. The data storage system can be used in the garbage collection method of the preferred embodiment of the present invention. The garbage collection method includes a writing subroutine and a recycling subroutine.

Figure 6 shows the state of the data storage system in the writing subroutine. First, the host 10 provides a writing instruction. A write command includes user data to be stored in NAND flash memory 16.

The SSD controller 12 has a processor 18. The processor 18 executes the FTL. The SSD memory 14 stores a mapping table 22, a Relevance Bit Map 24, a Data Buffer 26, and a Table Index Array 28. The NAND flash memory 16 includes a plurality of data blocks 30, a plurality of table blocks 32, and a table index block 34. The data blocks 30 are essentially identical, but are distinguished by adjectives because they play different roles at some stage in the garbage collection method of the present invention. The free data block 30 is a data block 30 that does not contain any user data. The destination data block 30 is a data block 30 to which the user data is to be entered. The data block to be recycled 30 is about to be The data block 30 is erased and recovered.

The mapping table 22 stores the correspondence between the logical address and the physical address. The size and number of the mapping table 22 depends on the size of the SSD memory 14.

The associative bit map 24 stores the association of a corresponding data block 30 with a plurality of mapping tables 22. Each bit of the associated bit map 24 represents a mapping table 22. If the current system has N mapping tables 22, then the associated bit map 24 will have at least N bits. When the associative bit map 24 is initialized, all of its bits are set to "0". If the corresponding data block 30 is associated with the Mth mapping table 22, the Mth bit of the associated bit map 24 is set to "1".

The data buffer 26 temporarily stores user data that will be written to the NAND flash memory 16 from the host 10. Once the address translation is completed and the physical address of the user data is determined, the user data is carried into the data block 30 of the NAND flash memory 16 by the data buffer 26.

The table index one-dimensional array 28 is a table and records a one-dimensional array of the physical addresses of the mapping table 22 and the associated bit map 24. If the current system has N mapping tables 22 and K association bitmaps 24, the size of the one-dimensional array is N+K columns. The table index one-dimensional array 28 stores the physical addresses of the individual tables in the NAND flash memory 16.

When a mapping table 22 is written, its contents are swiped to the table block 32 of the NAND flash memory 16, and this mapping table 22 is updated in the table index one-dimensional array 28 in the NAND flash memory 16 The physical address in .

When a data block 30 is written, the content of its associated bit map 24 is swiped to the table block 32 of the NAND flash memory 16, and the association is updated in the table index one-dimensional array 28. The physical address of the bit map 24 in the NAND flash memory 16.

At regular intervals, the table index one-dimensional array 28 is swiped to a specific address of the NAND flash memory 16.

The data block 30 stores user data from the host computer 10. Table block 32 stores various tables, primarily mapping table 22 and associated bit map 24. Table index block 34 stores the physical addresses of the various tables, located at a particular address of NAND flash memory 16.

Figure 7 shows the state of the data storage system in the recycling subroutine. The state shown in Figure 7 is similar to the state shown in Figure 6, but with a few differences. The first difference is that the data flow is two-way. The second difference is that the data buffer 26 stores the data recovered from the data block to be recovered 30 and writes the data to the target data block 30. The third difference is the addition of a valid data page bitmap.

In the SSD memory 14, a new associated bit map 24 is initialized, which will indicate the valid data page of the data block 30 to be recovered. For example, if a data block 30 has an M data page, its associated bit map 24 has at least M bits. If it is confirmed that the Kth data page of the to-be-recovered data block 30 is valid, the Kth bit of its associated bit map 24 is set to "1". At the initial stage, all the bits of the associated bit map 24 are set to "0". At the end of garbage collection, the space occupied by the associated bit map 24 will be released.

Figure 8 is a flow chart of the writing subroutine. At S10, host 10 provides a writing command. The writing instruction includes a piece of user data.

Before writing a piece of user data into the NAND flash memory 16, at S12, it is determined whether the user data is written into an idle data block 30, that is, whether the destination data block 30 is a free data block 30. . To write the book user data to a free data block 30, go to S14, otherwise go to S16. To write a user data into the free data block 30, there are two originals. because. First, this NAND flash memory 16 has never been written. This user data is the first data. Second, the previous destination data block 30 is full. For that reason, in S14, a new mapping table 22 and an associated bit map 24 must be initialized in the SSD memory 14. If the target data block is not an idle data block 30, it is not necessary to initialize a new table.

At S16, the user data is written into the data buffer 26, and after the FTL completes the conversion and determines the physical address to be written, the user data is swiped from the data buffer 26 to the destination data of the NAND flash memory 16. Block 30.

At S18, the mapping table 22 and the associated bit map 24 in the SSD memory 14 are updated.

At S20, it is judged whether or not the destination material block 30 is full. If the destination data block 30 is full, then go to S22, otherwise go to S24.

At S22, the associated bit map 24 is swiped down to the NAND flash memory 16.

At S24, it is judged whether or not this mapping table 22 is full. If the mapping table 22 is full, then go to S26, otherwise go to S28.

At S26, the mapping table 22 is swiped down to the NAND flash memory 16, and the FTL writes the physical address of this mapping table 22 to the table index one-dimensional array 28.

At S28, the writing subroutine ends.

Referring to Fig. 9, at S30, when the number of free data blocks 30 is detected to fall to a certain water level, the recycling subroutine is executed.

At S32, the data block 30 containing the least valid data page is selected as the data block 30 to be recycled.

At S34, the query table indexes the one-dimensional array 28 and finds the entity address of the associated bit map 24 corresponding to the data block to be recovered 30, and associates the association. The data of the bit map 24 is loaded from the corresponding table block 32 into the SSD memory 14.

At S36, scanning the associated bit map 24, it can be known that the to-be-recovered data block 30 is associated with which mapping table 22, and the physical addresses of the mapping tables 22 are searched one by one, and they are loaded from the NAND flash memory 16 Into the SSD memory 14. From the data of the mapping table 22, it can be known that those data pages of the data block 30 to be recovered are still valid.

At S38, a valid data page bit map 20 is created. The bit of the valid data page bit map 20 corresponding to the valid material page is set to "1". A method of how to create a valid material page bit map 20 will be described in detail later with reference to FIG.

At S52, it is judged whether or not all associated mapping tables 22 have been carried and scanned. If all of the associated mapping tables 22 have been carried and scanned, then go to S54, otherwise go back to S36.

At S54, a valid data page bit map 20 has been created. From this valid data page bit map 20, it is known that the physical address of those data pages is valid, and the user data is from the NAND flash memory 16 to be recovered. Block 30 loads the data buffer 26 of the SSD memory 14. Later, these user profiles are swiped to the destination data block 30.

At S56, all valid data pages of the data block to be recovered 30 have been recovered, and the garbage collection operation is completed.

Referring to Figure 10, a detailed description of how to create a valid data page bitmap 20 is described. At S40, the mapping table 22 is scanned.

At S42, the value of each entry in the mapping table 22 is read, and this value is the physical address.

At S44, it is determined whether the physical address is within the to-be-recovered data block 30. If the physical address is within the to-be-recovered data block 30, then go to S46, otherwise go to S48.

At S46, in the valid material page bit map 20, the bit corresponding to the physical address is set to "1".

At S48, it is judged whether the current entry is the last entry. If the current entry is the last entry, go to S50, otherwise go back to S42.

At S50, the scan ends. At this point, all entries of the current mapping table 22 have been scanned and the valid data page bitmap 20 has been completed.

The garbage collection method of the present invention is particularly suitable for a data storage device having limited SSD memory. The associated bit map 24 is established to reduce the number of scanned maps 22 and thus quickly find valid pages. In addition, a valid data page bit map 20 is created to indicate the valid data page, and thus the valid data page is quickly written from the to-be-recovered data block 30 to a new data block 30.

The above description is only for the preferred embodiment of the present invention, and is intended to clarify the features of the present invention, and is not intended to limit the scope of the embodiments of the present invention. Changes that are well known to those skilled in the art are still within the scope of the invention.

Claims (6)

A garbage collection method for a data storage device, comprising the steps of: (a) providing a data storage device, the data storage device comprising a NAND flash memory (16), the NAND flash memory (16) comprising a plurality of materials Block (30), each data block (30) includes a number of data pages; (b) selects a destination data block (30) from the data blocks (30); (c) writes user data Before the destination data block (30), a plurality of mapping tables (22) are created for the purpose data to record the relationship between the logical address of the user data and the physical address, and an associated bit map (24) is established, the association Each bit in the bit map (24) corresponds to a mapping table (22); (d) a data block to be recovered (30) is selected from the data blocks (30); (e) The correlation bit map (24) of the data block (30) reads the related mapping table (22); (f) determines one by one whether all the data pages of the to-be-recovered data block (30) are in the related mapping table (22) (g) if a data page of the data block to be recycled (30) is in the relevant mapping table (22), the data page is set as a valid data page; and (h) The data of the valid data page is written into another data block (30). The garbage collection method of the data storage device according to claim 1, wherein the step (g) comprises the steps of: (g1) establishing a valid data page bitmap (20), and the valid data page bitmap (20) Each bit in the pair corresponds to a data page of the data block (30) to be recycled. A method of garbage collection of a data storage device according to claim 2, wherein Step (g1) includes the steps of: (g1A) setting the bit of the corresponding valid data page in the valid data page bit map (20) to "1"; and (g1B) setting the valid data page bit map (20) The other bits in the ) are set to "0". The method for garbage collection of a data storage device according to claim 1, wherein the step (c) comprises the following steps: (c1) determining whether the destination data block (30) is a new data block (30); (c2) If the destination data block (30) is a new data block (30), a new associated bit map (24) and a new map (22) are initialized; (c3) the user is The data is written into the destination data block (30); and (c4) the associated bit map (24) and the map (22) are updated. The method for garbage collection of a data storage device according to claim 4, wherein the step (c) comprises the steps of: (c5) determining whether the destination data block (30) is full; (c6) if the destination data When the block (30) is full, the associated bit map (24) is written into the NAND flash memory (16); (c7) determines whether the mapping table (22) is full; (c8) if When the mapping table (22) is full, the mapping table (22) is written to the NAND flash memory (16); and (c9) is written. The garbage collection method of the data storage device according to claim 1, wherein the step (c) comprises the step of: (c10) locating the bit corresponding to the mapping table (22) in the associated bit map (24). Set to "1"; and (c11) set the other bits in the associated bit map (24) to "0".