System and method for improving wear-leveling performance in solid-state memory

[0001] Wear leveling is commonly used in Solid State Drives (SSDs) to prolong their lifecycle. A typical wear leveling operation involves moving data content in a first block of memory to a second block of memory that has just reached a certain erase level/count because of a recent erase operation. The first block of memory is then erased and made available for a future write operation. The overall process of wear leveling ensures that erase operations are evenly spread across blocks of memory in a SSD. Since each erase operation increases the wear of a block by incrementally reducing the block's ability to properly retain data content, wear leveling helps prevent certain blocks of memory from receiving an excessive amount of erase operations relative to other blocks and thus experiencing data failures much earlier than other blocks.

[0009] While certain embodiments of the inventions have been described, these embodiments have been presented by way of example only, and are not intended to limit the scope of the inventions. Indeed, the novel methods and systems described herein may be embodied in a variety of other forms. Furthermore, various omissions, substitutions and changes in the form of the methods and systems described herein may be made without departing from the spirit of the inventions. The accompanying claims and their equivalents are intended to cover such forms or modifications as would fall within the scope and spirit of the inventions.

Overview

[0010] Embodiments of the invention are directed to systems and methods for improving wear leveling performance in solid-state memory. The embodiments described herein make more consistent the number of wear leveling operations that need to be performed, so that sudden spikes in the number wear leveling operations may be reduced in solid-state memory. In one embodiment, a rule-based wear leveling approach is used to spread out the execution of wear leveling operations that otherwise would have been triggered in clusters. Under the rule-based approach, wear leveling is periodically triggered by a specified interval of erase counts associated with a unit of solid-state memory such as a group of blocks, rather than by a threshold based on erase counts.

System Overview

[0011] FIG. 1 is a block diagram illustrating a storage subsystem embodiment with a wear leveling module. As shown, a storage subsystem 140 includes a controller 150, which in turn includes a wear leveling module 156 and a garbage collection module 158. In one embodiment, the wear leveling module 156 is responsible for issuing wear leveling commands and the garbage collection module is responsible for issuing garbage collection commands, with the controller 150 relaying those commands to non-volatile solid-state memory arrays 160 for execution. In one embodiment, the wear leveling module 156 is configured to select blocks for wear leveling and issue wear leveling commands in accordance with the methods described in this disclosure. In one embodiment, the controller 150 is configured to receive memory commands from a driver 132 residing on a host system 130 and execute those host-issued memory commands in the non-volatile solid-state memory arrays 160.

Wear Leveling Operations

[0012] FIGS. 2A and 2B are block diagrams illustrating changes to a simplified example erase counter table in response to the triggering of a threshold-based wear leveling operation. As the name implies, threshold-based wear leveling is triggered when a certain unit of memory (e.g., a block or superblock) reaches a particular erase count threshold. FIG. 2A shows an erase counter table 210 that tracks the erase counts recorded for various superblocks in the non-volatile memory. A superblock is a group of memory blocks, and in one embodiment, each block in a superblock is on one die of a group of dies. Other configurations are possible in other embodiments. For example, in one embodiment, a superblock is comprised of blocks from various planes in a group of dies. In the embodiment shown, the erase counter for each superblock is assumed to begin at zero and erase operations are performed by the garbage collection module 158 largely on a superblock basis.

[0013] In the simplified example of FIGS. 2A and 2B, the threshold for triggering wear leveling is 8 erase counts. In FIG. 2A, Superblock No. 2 has just been erased and reaches the threshold of 8. This means that all blocks within Superblock No. 2 are now in need of wear leveling since they all have been erased eight times. The counter table 210 in FIG. 2B shows the erase counter table after the performance of wear leveling on blocks within Superblock No. 2. Superblock No. 1 is chosen to participate in the wear leveling operation because its blocks, along with those in Superblock No. 7, have been erased the fewest number of times (i.e., they have the least wear). The wear leveling operation in this simplified example involves (1) moving the data contents of blocks in Superblock No. 1 into blocks in Superblock No. 2, (2) erasing those blocks in Superblock No. 1, and (3) swapping the assignment of blocks in Superblock Nos. 1 and 2. The wear leveling operation of blocks in Superblock Nos. 1 and 2 as shown and described is for illustration only in this simplified example, as in various embodiments the particular blocks on which wear leveling are performed are selected in accordance with a number of rules and configurations, as will be further described below. For example, a block that is due to be wear leveled may be wear leveled with another block on the same die. As illustrated in FIGS. 2A and 2B, with threshold-based wear leveling, the number of wear leveling operations that is triggered by an erase operation may be quite large. In the example, if each superblock includes 32 blocks, then 32 wear leveling operations must be performed at once when a superblock reaches a wear leveling threshold. The sudden requirement of having to perform 32 wear leveling operations may negatively impact the throughput of host-issued memory commands that the controller can handle.

[0014] FIG. 3 further illustrates the threshold-based wear leveling approach (illustration 310) and compares it to the rule-based wear leveling approach (illustration 320) in accordance with one embodiment. In the example shown in illustration 310, the threshold for wear leveling is 4, and Superblock No. 1 consists of four blocks, each from Dies No. 1-4. Both illustrations 310 and 320 are block diagrams that illustrate operations performed in sequence and arranged chronologically from top to bottom. Illustration 310 shows that four consecutive erase operations are performed on Superblock No. 1. At the end of the fourth erase operation, wear leveling is triggered for all four blocks within Superblock No. 1, and hence four wear leveling operations (one for each block in Dies No. 1-4) must be performed as shown.

Rule-Based Wear Leveling

[0015] Illustration 320 shows a different sequence under the rule-based wear leveling approach, which spreads out the execution of wear leveling operations that otherwise would have been triggered together. As its name suggests, in one embodiment, the rule-based approach ensures that the blocks in a superblock are periodically wear leveled according to a set of rules instead of being triggered by thresholds. In one embodiment blocks in a superblock are periodically wear leveled on a rotational basis even if the superblock erase counter has not yet reached a wear leveling threshold. The period in the rule-based approach is set in accordance with an interval value denoting a number of erase operations between two wear leveling operations. The interval is set to 1 in the example shown in FIG. 3. In other words, according to the example shown in FIG. 3, a wear leveling operation is performed for every one erase operation performed in the superblock, regardless of whether the wear leveling threshold has been reached. If the interval is set at 10, a wear leveling operation is performed for every ten erase operations performed in the superblock. In one embodiment, the interval value is based on the wear leveling threshold (by erase counts) divided by the number of blocks in a superblock. For example, a superblock with a threshold of 400 and 32 blocks may have the interval value set at 12. Executing the wear leveling operation periodically based on the interval and regardless of the wear levels of the individual blocks in the superblock makes the overall wear leveling deterministic. Because blocks are selected from dies or planes on a rotational basis each time wear leveling is triggered, the rule-based approach avoids the sudden need to execute a large number of wear leveling operations while ensuring that blocks are properly selected for wear leveling over a period of time for even wear.

[0016] Returning to FIG. 3, as shown in illustration 320, as the interval is set at 1, a wear leveling operation is performed for each erase operation. In the example shown, blocks from the dies are selected in a rotation fashion, with a block in Die No. 1 selected first, followed by a block in Die No. 2, and so on. In one embodiment, the block that is selected is the most worn block in the die, and the selected block is wear leveled with the least worn block in the die. The rule-based wear leveling approach will be further described in FIG. 4.

[0017] FIG. 4 is a flow diagram showing a process 400 of rule-based wear leveling in accordance with one embodiment. At block 402, an erase operation is performed on a superblock. At block 404, a superblock erase counter is incremented in response to the erase operation. At block 406, the value of the superblock erase counter is divided by a wear level interval, and the process 400 determines whether the remainder is zero. If it is not, the process terminates and waits until the next erase operation is performed on the same superblock, at which time the process will begin anew at block 400. If the remainder is zero, the process 400 moves to block 408, where a wear leveling operation is initiated. In one embodiment, the determination in block 406 is performed through a modulo operation on the superblock erase counter and the wear level interval. In other embodiments, the wear level interval may be a set time period or a set number of erase counts and the determination as to whether a wear leveling operation should be performed is made in accordance with the type of wear level interval used. Also, in some embodiments, the interval may be dynamically adjusted in accordance with the overall wear level of the solid-state memory and/or the amount of static data within the memory.

[0018] At block 410, a block within a selected die is wear leveled with another block within the same die. The die is selected based on a current die index (represented by a value X in FIG. 4). In one embodiment, the wear leveling operation selects the most worn block on the selected die that is currently in the free pool, and performs wear leveling with the least worn block on the die that is not in the free pool. In one embodiment, the free pool stores blocks that are available for writing. Blocks that are not in the free pool contain valid data and are otherwise not available for writing. Returning to FIG. 4, once the wear leveling operation is complete, the current die index is incremented at block 412 so a block from the next die in the rotation is selected in the next wear leveling operation, and the process 400 ends.

[0019] FIGS. 5A-5H are block diagrams that illustrate an example set of superblocks that are undergoing rule-based wear leveling operations. The tables in FIGS. 5A-5H show the erase counters for eight example superblocks (Superblocks Nos. 1-8). Each example superblock has eight blocks, which are tracked across the columns as Blocks Nos. 1-8. In one embodiment, each block of Blocks Nos. 1-8 is located on a different die or plane, though for the purpose of illustration they will be described as being located on a different die of Dies Nos. 1-8. In addition, in the example of FIGS. 5A-5H, the wear level interval is set at 1. FIG. 5A shows the state of the superblocks before an erase operation is performed. FIG. 5B shows that Superblock No. 2 is erased, and the erase counters for Superblock No. 2 are incremented (a process previously shown in block 404 of FIG. 4). Because the wear level interval is 1 in this example, a wear leveling operation is initiated. In one embodiment, the wear leveling operation is performed on a rotational basis, and in this example the wear leveling moves sequentially from Die No. 1 to Die No. 8. Therefore, in FIG. 5C, a wear leveling operation is performed on Die No. 1.

[0020] In one embodiment, two different types of wear leveling operations may be performed: static and dynamic. In static wear leveling, the most worn block on a die is wear leveled with the least worn block on the die. In dynamic wear leveling, the most worn block on a die in the free pool is wear leveled with the least worn block on the die that contains valid data (i.e., not in the free pool). FIG. 5C illustrates dynamic wear leveling, so the block assigned to Superblock No. 2 on Die No. 1 is selected for wear leveling because it has the highest erase count and Superblock No. 2 has just been erased. The wear leveling operation moves data from the least worn block with valid data on Die No. 1 (the block assigned to Superblock No. 1) to the block assigned to Superblock No. 2, and then erases the block assigned to Superblock No. 1. Then the superblock assignments of the two blocks are switched so that the block assigned to Superblock No. 1 is re-assigned to Superblock No. 2 and vice versa. The updated erase counts are shown in FIG. 5C. After the wear leveling is complete, the current die index is incremented (as previously shown in block 412 of FIG. 4).

[0021] FIGS. 5D-5F illustrate the next erase cycle, with FIG. 5D showing the state of the superblocks before the next erase operation. FIG. 5E shows that Superblock No. 2 is erased again. Because the wear level interval is one, the erase operation necessitates a wear leveling operation. In FIG. 5F, wear leveling is now performed on Die No. 2 (as a result of the current die index having been previously incremented). As discussed above, dynamic wear leveling in one embodiment is performed on the most worn of all blocks on a die that are in the free pool. In FIG. 5F, of all the blocks in Die No. 2, the block assigned to Superblock No. 2 is selected because, prior to the performance of the wear leveling operation, it has the highest erase count (most worn) and Superblock No. 2 has just been erased. The wear leveling operation moves data from the least worn block on Die No. 2 (the block assigned to Superblock No. 1) to the block assigned to Superblock No. 2, and then erases the block assigned to Superblock No. 1. Then the superblock assignments of the two blocks are switched so that the block assigned to Superblock No. 1 is re-assigned to Superblock No. 2 and vice versa. The updated erase counts are shown in FIG. 5F. After the wear leveling is complete, the current die index is again incremented.

[0022] FIGS. 5G-5I illustrate the next erase cycle, with FIG. 5G showing the state of the superblocks before the next erase operation. FIG. 5H shows that Superblock No. 1 is erased this time. Because the wear level interval is one, the erase operation necessitates another wear leveling operation. In FIG. 5I, wear leveling is now performed on Die No. 3 (as a result of the current die index having been previously incremented). In FIG. 5I, the wear leveling operation moves data from the least worn block with valid data on Die No. 3 (the block assigned to Superblock No. 7) to the block assigned to Superblock No. 2 (most worn block on Die No. 3 in the free pool), and then erases the block assigned to Superblock No. 7. Then the superblock assignments of the two blocks are switched so that the block assigned to Superblock No. 7 is re-assigned to Superblock No. 2 and vice versa. The updated erase counts are shown in FIG. 5I.

[0023] FIG. 6 is a flow diagram showing a process 600 of switching between threshold-based wear leveling and rule-based wear leveling in accordance with one embodiment. In one embodiment, the switching between the two methods of wear leveling is intended to maximize performance of the storage subsystem, since each wear leveling method has an associated percentage range of static data in the overall memory in which the method performs optimally, i.e., the wear leveling operations triggered by the method are evenly distributed with respect to time. In one embodiment, as shown in block 602, the controller 150 and/or the wear leveling module 156 is configured to perform process 600, which continuously and/or periodically monitors the level of static data that is present in the non-volatile memory arrays 160. The level of static data may be determined in a variety of ways, including using an invalid page counter based on information from an invalid page table to determine the amount of invalid data stored. For example, a superblock that has a zero invalid page counter may be considered to have static data, and the percentage of static data may be calculated by summing the number of superblocks that have such zero counters. At block 604, the process 600 determines whether the level of static data is within a range that is suitable for rule-based wear leveling. If so, it initiates rule-based wear leveling (or maintains it if it is already using it) at block 606. Otherwise, it initiates threshold-based wear leveling (or maintains it if it is already using it) at block 608. In either case, the process returns to block 600. In one embodiment, the process shown in FIG. 6 is performed periodically, for example, in accordance with a time interval or an interval based on a total erase count of the storage subsystem. In one embodiment, the range of static data in which rule-based wear leveling is used is 10%-25%.

[0024] The features and attributes of the specific embodiments disclosed above may be combined in different ways to form additional embodiments, all of which fall within the scope of the present disclosure. Although the present disclosure provides certain embodiments and applications, other embodiments that are apparent to those of ordinary skill in the art, including embodiments which do not provide all of the features and advantages set forth herein, are also within the scope of this disclosure. Accordingly, the scope of the present disclosure is intended to be defined only by reference to the appended claims.