A heat-aware based SSD mapping region adaptive management method

Abstract

The application discloses a kind of based on heat perception's SSD mapping area adaptive management method in the technical field of solid state disk.The method includes: obtaining the heat data of each mapping area, and it is calculated based on write frequency, access frequency and recent access time weighting;According to heat data and the comparison table of preset heat interval, determine target area granularity;When heat data crosses split threshold or merging threshold, adjust the granularity of region to target granularity, wherein split threshold is higher than merging threshold, and when heat is located between them, keep granularity unchanged;Adjustment mode includes splitting mapping area into smaller sub-region, or merging multiple continuous mapping areas into larger region.The application realizes the adaptive dynamic adjustment of mapping area granularity with access heat, and hot data is managed with small granularity to reduce write amplification, cold data is managed with large granularity to reduce resource occupation, and the hysteresis threshold mechanism avoids frequent splitting and merging, which improves the storage management efficiency and life of solid state disk.

Description

Technical Field

[0001] This invention relates to the field of flash translation layer (FTL) technology for solid-state drives (SSDs), and more particularly to an adaptive management method for SSD mapping regions based on heat perception. Background Technology

[0002] In solid-state drives (SSDs) employing page-level mapping (L2P), the logical address to physical address mapping table is typically large, accounting for approximately 0.1% of the total SSD capacity. For example, in a 1TB SSD with a page granularity of 4KB, the mapping table size is approximately 1GB. For ease of management, the industry typically divides the L2P table into multiple fixed-size mapping regions based on logical address ranges, using these regions as the basic unit for managing, caching, and persisting the mapping table.

[0003] Existing mapping area management schemes typically employ the following methods: First, they use fixed-granularity area division, with all mapping areas being the same size; second, they adopt a static area management strategy, where the management method of each area does not change with changes in data access patterns; and third, they lack a dynamic adjustment mechanism, with the number and granularity of areas remaining constant throughout the entire lifecycle of the solid-state drive.

[0004] The above fixed-granularity scheme has the following drawbacks: First, due to its overly large granularity, the mapping regions with high access frequency require persistence of all mapping relationships when only a few logical addresses within a region are modified, leading to severe write amplification. Second, cold data regions consume the same management resources as hot data regions, resulting in wasted management resources. Third, the number of regions is fixed and cannot be adjusted dynamically to accommodate changes in data access patterns. Fourth, it cannot adapt to dynamic changes in user input load, causing storage performance to fluctuate significantly with load characteristics. Fifth, overall management efficiency is low, failing to balance fine-grained management of hot data with efficient storage of cold data. Summary of the Invention

[0005] This application provides a heat-aware SSD mapping area adaptive management method, which solves the problem that existing solid-state drive mapping area management schemes, which use fixed granularity division and static management strategies, cannot adapt to dynamic load changes and have low overall management efficiency. It realizes the adaptive dynamic adjustment of mapping area granularity according to data access heat.

[0006] This application provides an adaptive management method for SSD mapping regions based on heat perception, comprising the following steps: The heat data of each mapped region in the solid-state storage device is obtained. The heat data is calculated by weighting the number of writes, access frequency and recent access time of the mapped region. Based on the heat data and the preset heat interval comparison table, the target region granularity corresponding to each of the mapping regions is determined; wherein, the heat interval comparison table establishes a hierarchical correspondence between multiple heat intervals from the lowest heat to the highest heat and multiple standard region granularities from the largest granularity to the smallest granularity, and the higher the heat, the smaller the corresponding standard region granularity; For each mapped region, if its current region granularity is not equal to the target region granularity, and the popularity data exceeds a preset splitting threshold or merging threshold, then the region granularity of the mapped region is adjusted to the target region granularity. The adjustment includes: when the popularity data of the mapped region increases and exceeds the splitting threshold, splitting the mapped region into multiple sub-mapped regions with smaller granularity; when the popularity data of the mapped region and at least one adjacent mapped region decreases and exceeds the merging threshold, merging the mapped region and the corresponding adjacent mapped region into a new mapped region with larger granularity. Wherein, for two adjacent standard region granularity levels, the splitting threshold is higher than the merging threshold, and when the popularity data is between the splitting threshold and the merging threshold, the current region granularity remains unchanged.

[0007] The beneficial effects of the above embodiments are as follows: This management method establishes a hierarchical correspondence between heat and granularity and combines it with a hysteresis threshold mechanism to achieve adaptive dynamic adjustment of the mapping region granularity according to the access heat. Hot data is managed with small granularity to reduce write amplification, while cold data is managed with large granularity to reduce resource consumption. At the same time, it avoids frequent splitting and merging caused by heat boundary fluctuations.

[0008] Based on the above embodiments, this application can be further improved as follows: In one embodiment of this application, obtaining the heat data of each mapped region includes: triggering heat calculation at a preset period using a timer, storing the heat data obtained each time into a sliding window sequence, and using the statistical average of the heat data in the sliding window sequence as the final heat data for the current period; wherein, when the solid-state storage device is detected to be in an idle state, the timer counting is stopped.

[0009] Technical benefits: The sliding window smooths out heat data, eliminating the interference of instantaneous fluctuations on granularity adjustment; the timer count is stopped when idle, avoiding unnecessary heat updates and resource consumption in idle state.

[0010] In one embodiment of this application, the splitting includes: When the heat data increases and exceeds the splitting threshold corresponding to the current region granularity, the mapping region is evenly divided into two sub-mapping regions with the same granularity as the target region, starting from the logical starting address of the mapping region. Each of the sub-mapping regions is assigned a new region identifier, and a corresponding management entry is created based on its logical starting address and length. The address mapping data of each sub-mapping region is persisted to the physical storage medium, and the persistent data and corresponding management entries of the original mapping region are deleted.

[0011] Technical effects: It clarifies the complete process of the splitting operation, ensures that the sub-mapped region inherits the continuity of the original region in the logical address space, and guarantees power-loss safety through persistence.

[0012] In one embodiment of this application, the merging includes: Two or more mapping regions to be merged are identified and arranged consecutively along the logical address order. The heat data of each mapping region to be merged is lower than the merging threshold corresponding to the granularity of the new mapping region formed after merging, and the logical address intervals of each mapping region to be merged are consecutive. Determine whether the total length of each of the mapping regions to be merged is one of the preset standard region granularities; if so, merge them into a new mapping region. The minimum logical starting address among the various mapping regions to be merged is used as the logical starting address of the new mapping region, the sum of the lengths of the various mapping regions to be merged is used as the length of the new mapping region, and a new region identifier is assigned to the new mapping region. Delete the management information for each of the mapping regions to be merged.

[0013] Technical benefits: By restricting the merged result to a standard granularity, fragmented mapping regions with non-standard granularity are avoided, reducing management complexity; the new mapping region only occupies one set of management resources, and the overall merge reduces the management resource consumption.

[0014] In one embodiment of this application, when there are multiple feasible merging combinations within the same logical address range, the judgment is made in ascending order of logical starting address, and the combination with the largest granularity after merging is selected first; when there are multiple combinations with the same granularity after merging, the combination with the smallest logical starting address is merged first; if the total length of each of the mapping regions to be merged cannot form a preset standard region granularity, then merging is not performed in this round.

[0015] Technical effects: Clear merge priority rules were established, resolving decision conflicts when multiple combinations coexist, maximizing the benefits of a single merge, and avoiding ineffective merge operations.

[0016] In one embodiment of this application, the initial heat data of the new mapping region is taken as the highest heat of each of the mapping regions to be merged, and then recalculated using a timer and a sliding window.

[0017] Technical effect: After merging, the highest heat value is used as the initial value, which avoids the heat value of the new mapping region being underestimated after merging, reduces the probability of hot data being incorrectly included in the large-granularity mapping region and the new mapping region continuing to be over-merged, and further enhances the stability of granularity adjustment.

[0018] In one embodiment of this application, the splitting or merging operation is based on the logical starting address and length of the mapping region to determine the operation object. The region identifier is only used as a unique identifier of the mapping region and does not indicate the arrangement position of the mapping region in the logical address space. After splitting or merging, an unused region identifier is assigned to the new mapping region, and the management information corresponding to the original mapping region identifier is deleted.

[0019] Technical effect: Decoupling the region identifier from the logical address location makes the splitting and merging operations not restricted by the order of the region identifier numbers, thus improving operational flexibility.

[0020] In one embodiment of this application, the heat data is calculated using the following formula: in, This represents the total number of writes to the mapped region. The access frequency of the mapped region. The recent access frequency of the mapped region is scored; , , These are the preset weighting coefficients.

[0021] Technical effect: The popularity is comprehensively evaluated from three dimensions: number of writes, access frequency, and recent access time, avoiding the bias caused by a single indicator and making the granular adjustment more accurate and reasonable.

[0022] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. This invention splits regions by using a heat-based classification system. Hot data is managed with fine granularity to reduce write amplification, while cold data is managed with large granularity to reduce resource consumption.

[0023] 2. This invention sets splitting and merging thresholds for two adjacent Region granularity levels, with the splitting threshold being higher than the merging threshold. When the Region's popularity is between these two thresholds, the current granularity remains unchanged. Splitting or merging operations are only performed when the popularity exceeds the splitting or merging threshold. This reduces redundant splitting, merging, and duplicate persistence of management entries caused by fluctuations in popularity near granularity boundaries.

[0024] 3. This invention calculates popularity information by weighting it from three perspectives: number of writes, number of accesses, and frequency of recent accesses, thus avoiding the distortion of popularity data caused by a single indicator. Attached Figure Description

[0025] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. The elements or parts in the drawings are not necessarily drawn to scale.

[0026] Figure 1 This is a flowchart illustrating the steps of an adaptive management method for SSD mapping regions based on heat perception, as described in an embodiment of this application. Figure 2 This is a schematic diagram showing the distribution of mapping regions and logical addresses in the logical space in an embodiment of the present invention; Figure 3 This is a diagram showing the correspondence between the granularity of the mapping region and the heat data in an embodiment of the present invention; Figure 4 This is a flowchart of the mapping region splitting in an embodiment of the present invention; Figure 5 This is a flowchart of the mapping region merging process in an embodiment of the present invention. Detailed Implementation

[0027] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading the present invention, any modifications of the present invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0028] Example 1: like Figure 1 As shown, an adaptive management method for SSD mapping regions based on heat awareness includes the following steps: S1: Obtain the heat data of each mapped region in the solid-state storage device. The heat data is calculated by weighting the number of writes, access frequency and the most recent access time of the mapped region.

[0029] like Figure 2 As shown, in the initial state of a solid-state storage device, the logical address space is divided into multiple contiguous mapping regions, each containing a fixed number of logical addresses (LCAs). Initially, all mapping regions have the same granularity.

[0030] The Flash Translation Layer (FTL) of a solid-state storage device maintains heat data for each mapped region. This heat data includes write counts, access frequency, and recent access score. The heat data is calculated using the following formula: ; in, , , These are preset weighting coefficients that can be configured based on the application scenario and load characteristics of the SSD. For example, they can be increased in write-intensive scenarios. In mixed read / write scenarios, the values ​​of these three factors can be appropriately balanced.

[0031] Popularity data is updated every 24 hours via a timer. Each update stores the calculated popularity data into a sliding window sequence of length 6. When the sliding window is full, new data replaces the oldest data, and the arithmetic mean of all popularity data within the sliding window is taken as the final popularity data for the current period. The timer stops counting when the solid-state storage device is detected to be idle, avoiding unnecessary popularity updates when there is no access operation.

[0032] S2: Determine the target region granularity corresponding to each mapping region based on the heat data and the preset heat interval comparison table.

[0033] To more intuitively represent the correspondence, this embodiment transforms the heat range lookup table into the following: Figure 3 The diagram shows the correspondence between the granularity of the mapping region and the heat data.

[0034] A popularity interval comparison table is pre-established, dividing the popularity data into multiple popularity intervals from the lowest to the highest value (e.g., Figure 3 The system is divided into eight intervals: below T1, T0-T3, T2-T5, T4-T7, T6-T9, T8-T11, T10-T13, and above T12. T0-T13 represent the heat values ​​arranged from smallest to largest, with some overlap between adjacent heat intervals. The region granularity is further divided into multiple standard region granularities (e.g., 128B, 256B, 512B, 1KB, 2KB, 4KB, 8KB, and 16KB). A heat interval mapping table establishes a hierarchical correspondence between multiple heat intervals from lowest to highest heat and multiple standard region granularities from largest to smallest granularity; higher heat corresponds to smaller standard region granularities.

[0035] Specifically, when the heat of the mapped region is T0, the corresponding standard region granularity is 16KB; when the heat increases to T1 or above, the standard region granularity is 8KB; when the heat increases to T3 or above, the standard region granularity is 4KB; and so on. The higher the heat, the smaller the region granularity and the higher the management precision.

[0036] For two adjacent standard region granularity levels, splitting thresholds (T1, T3, T5, T7, T9, T11, T13) and merging thresholds (T0, T2, T4, T6, T8, T10, T12) are set respectively. The splitting threshold is located near the lower limit of the higher popularity range, and the merging threshold is located near the upper limit of the lower popularity range, with the splitting threshold being higher than the merging threshold. When the popularity data of the mapped region is between the splitting threshold and the merging threshold, the target region granularity is the current region granularity, thus maintaining the current region granularity unchanged subsequently. This hysteresis mechanism effectively prevents repeated switching of region granularity (i.e., the ping-pong effect) caused by small fluctuations in popularity near the boundary.

[0037] S3: For each mapped region, if its current region granularity is not equal to the target region granularity, and the heat data exceeds the preset splitting threshold or merging threshold, then the region granularity of the mapped region will be adjusted to the target region granularity.

[0038] The adjustments include: when the popularity data of a mapped region increases beyond the splitting threshold, the mapped region is split into multiple sub-mapped regions with smaller granularity; when the popularity data of a mapped region and at least one adjacent mapped region decreases beyond the merging threshold, the mapped region and the corresponding adjacent mapped region are merged into a new mapped region with larger granularity.

[0039] like Figure 4 As shown, the hotspot region splitting process is as follows: Assume the original mapped region is 16KB in size and the logical address (LCA) range is Region0[0, 16K-1]. Due to the influence of front-end instructions, the hot data in this region increases. When the hot data in this region increases and exceeds the splitting threshold (T1) corresponding to the interval, the target region granularity is determined to be 8KB (standard region granularity), which meets the splitting condition, and it is split.

[0040] The specific process for performing the splitting operation is as follows: First, starting from the logical start address 0 of the original mapped region, the 16KB Region0 is evenly split into two 8KB sub-mapped regions, namely Region1[0, 8K-1] and Region2[8K, 16K-1]. A new region identifier (e.g., RegionID=1) is assigned to Region1, with a logical start address (LCA_START) of 0 and a length of 8K; a new region identifier (e.g., RegionID=2) is assigned to Region2, with a logical start address of 8K and a length of 8K. Region1 and Region2 inherit the heat data of the original Region0 as their respective initial heat values.

[0041] Subsequently, the address mapping data of Region1 and Region2 are persisted to the physical storage medium, and the corresponding physical addresses are recorded in the management table. At the same time, the persistent address of the original Region0 is removed from the management table. Finally, the management information corresponding to the original Region0 is deleted.

[0042] By splitting the data, the original situation where only a few logical addresses in the 16KB granular mapping region needed to be persisted when they were modified was optimized to only needing to persist the mapping relationship of the more frequently accessed 8KB sub-regions, effectively reducing write amplification.

[0043] like Figure 5 As shown, the cold zone merging process is as follows: The timer is triggered periodically to update the heat data information of each mapping region. The heat data is judged from small to large according to the logical starting address. When the heat data of two or more mapping regions arranged consecutively along the logical address sequence are all lower than the corresponding merging threshold, and the logical address intervals of each mapping region are consecutive, the merging judgment is triggered. If the total length of each candidate merging mapping region is the standard region granularity, then each candidate merging mapping region is merged.

[0044] Assume there are three consecutive mapped regions in logical address order: Region0 (granularity 2KB, LCA range [0, 2K-1]), Region1 (granularity 4KB, LCA range [2K, 6K-1]), and Region2 (granularity 2KB, LCA range [6K, 8K-1]). The heat data of all three regions is below the merging threshold corresponding to the 8KB granularity. The sum of the lengths of the three regions is 2KB + 4KB + 2KB = 8KB, which is one of the preset standard region granularities. Therefore, the three regions are merged into a new 8KB mapped region.

[0045] The specific process of the merge operation is as follows: Assign a new region identifier (e.g., RegionID=3) to the new mapped region. The initial heat of the new mapped region is the highest heat among the three original mapped regions, and it is recalculated subsequently using a timer and sliding window. Use the logical starting address 0 of the original Region0 as the logical starting address of the new mapped region, and set the Length to 8K. Persist the data in the new mapped region Region3, and update the persistent Flash address to the management table entry.

[0046] After the merge, the management information for Region0, Region1, and Region2 is deleted, and the original persistent Flash addresses are removed from the management table entries. The new mapped region after the merge occupies only one set of management resources, a reduction of two-thirds compared to the three sets of management resources before the merge.

[0047] The following points require further clarification regarding the merger rules: 1. The merging operation is limited to cases where the total length can form a standard region granularity. If the total length of three adjacent regions is 10KB (e.g., 4KB+2KB+4KB), and cannot form a standard granularity such as 128B, 256B, 512B, 1KB, 2KB, 4KB, 8KB, or 16KB, then merging will not be performed in this round, and will be reassessed after subsequent changes in popularity or region granularity.

[0048] 2. The merging rule requires that the merged objects be mapping regions arranged consecutively along the logical address sequence, that is, the logical address intervals of each mapping region are connected end to end. For example, the LCA intervals [0, 8K] and [8K+1, 16K] are consecutive and can be merged; however, there are address gaps between the LCA intervals [0, 8K] and [16K, 24K], so they cannot be merged.

[0049] 3. When there are multiple feasible merging combinations within the same logical address range, the following priority rules shall be followed for decision-making: First, judge according to the logical starting address from smallest to largest; second, prioritize the combination with the largest granularity after merging; third, when there are multiple combinations with the same granularity after merging, prioritize merging the combination with the smallest logical starting address.

[0050] For example, suppose there are consecutive mapped regions A[0, 2K-1] (granularity 2KB), B[2K, 4K-1] (granularity 2KB), and C[4K, 6K-1] (granularity 2KB) in logical address order, with a total length of 6KB. The heat of A, B, and C is all below the 4KB merging threshold. There are three possible merging options: merge A, B, and C into 6KB (non-standard granularity, not feasible), merge A and B into 4KB (standard granularity, feasible), or merge B and C into 4KB (standard granularity, feasible). According to the rules, the feasible combination with the largest granularity after merging is preferred, i.e., merging A and B and merging B and C both result in 4KB; when combinations with the same granularity exist, the combination with the smallest logical starting address is prioritized, i.e., merging A and B.

[0051] Furthermore, in this invention, the RegionID serves only as a unique identifier for the mapped region and does not indicate its position in the logical address space. The position and range of the mapped region in the logical address space are determined by the logical start address (LCA_START) and the length. When performing a split or merge operation, an unused RegionID is assigned to the new mapped region, and the management information corresponding to the original mapped region ID is deleted. The objects of splitting and merging are determined based on the logical start address and length, not on the numerical value or continuity of the RegionID. This design makes the allocation of RegionIDs unrestricted by the logical address order, making splitting and merging operations more flexible and unconstrained by the RegionID numbering.

[0052] Example 2: A solid-state drive employs a management method as shown in Example 1 to adaptively manage the mapped region.

[0053] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: 1. Dynamically adjust the granularity of the mapping region based on changes in user input patterns. Hot data mapping regions have finer granularity; when only a few logical addresses within a region are modified, only the mapping relationship of the small-granularity mapping region needs to be persisted, significantly reducing write amplification. Cold data mapping regions have larger granularity, reducing the consumption of management resources.

[0054] 2. For two adjacent region granularity levels, set a split threshold and a merge threshold respectively, with the split threshold being higher than the merge threshold. When the region's popularity is between the two, the current granularity remains unchanged. The corresponding split or merge operation is only performed when the popularity exceeds the split threshold or merge threshold, effectively reducing duplicate splits, merges, and duplicate persistence of management entries caused by fluctuations in popularity near the granularity boundary.

[0055] 3. The popularity information is calculated by weighting it from three dimensions: number of writes, number of accesses, and frequency of recent accesses. This avoids the distortion of popularity data caused by a single indicator and makes the granular adjustment more reasonable and accurate.

[0056] 4. Merge multiple cold regions with different granularities but consecutive addresses. Merge is only performed when the total length can form a preset standard region granularity, avoiding the generation of non-standard granularity management fragments, improving the merging rate of cold data regions, and further reducing the consumption of region management resources.

[0057] 5. When an enterprise-grade solid-state drive (SSD) experiences an abnormal power outage, temporary data persistence is achieved through a brief period provided by supercapacitors. Because the granularity of hot data regions is sufficiently small, the corresponding data persistence time is significantly reduced. Similarly, during the abnormal power outage recovery phase, when recovering hot data, the small granularity of the hot data regions reduces the amount of data recovered, drastically shortening the data recovery time.

[0058] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A heat-aware SSD mapping region adaptive management method, characterized in that, Includes the following steps: The heat data of each mapped region in the solid-state storage device is obtained. The heat data is calculated by weighting the number of writes, access frequency and recent access time of the mapped region. Based on the heat data and the preset heat interval comparison table, the target region granularity corresponding to each of the mapping regions is determined; wherein, the heat interval comparison table establishes a hierarchical correspondence between multiple heat intervals from the lowest heat to the highest heat and multiple standard region granularities from the largest granularity to the smallest granularity, and the higher the heat, the smaller the corresponding standard region granularity; For each mapped region, if its current region granularity is not equal to the target region granularity, and the popularity data exceeds a preset splitting threshold or merging threshold, then the region granularity of the mapped region is adjusted to the target region granularity. The adjustment includes: when the popularity data of the mapped region increases and exceeds the splitting threshold, splitting the mapped region into multiple sub-mapped regions with smaller granularity; when the popularity data of the mapped region and at least one adjacent mapped region decreases and exceeds the merging threshold, merging the mapped region and the corresponding adjacent mapped region into a new mapped region with larger granularity. Wherein, for two adjacent standard region granularity levels, the splitting threshold is higher than the merging threshold, and when the popularity data is between the splitting threshold and the merging threshold, the current region granularity remains unchanged.

2. The management method according to claim 1, characterized in that: Acquiring heat data for each mapped region includes: triggering heat calculation at a preset period using a timer, storing the heat data obtained each time into a sliding window sequence, and using the statistical average of the heat data in the sliding window sequence as the final heat data for the current period; wherein, when the solid-state storage device is detected to be in an idle state, the timer counting is stopped.

3. The management method according to claim 1, characterized in that, The splitting includes: When the heat data increases and exceeds the splitting threshold corresponding to the current region granularity, the mapping region is evenly divided into two sub-mapping regions with the same granularity as the target region, starting from the logical starting address of the mapping region. Each of the sub-mapping regions is assigned a new region identifier, and a corresponding management entry is created based on its logical starting address and length. The address mapping data of each sub-mapping region is persisted to the physical storage medium, and the persistent data and corresponding management entries of the original mapping region are deleted.

4. The management method according to claim 1, characterized in that, The merger includes: Two or more mapping regions to be merged are identified and arranged consecutively along the logical address order. The heat data of each mapping region to be merged is lower than the merging threshold corresponding to the granularity of the new mapping region formed after merging, and the logical address intervals of each mapping region to be merged are consecutive. Determine whether the total length of each of the mapping regions to be merged is one of the preset standard region granularities; if so, merge them into a new mapping region. The minimum logical starting address among the various mapping regions to be merged is used as the logical starting address of the new mapping region, the sum of the lengths of the various mapping regions to be merged is used as the length of the new mapping region, and a new region identifier is assigned to the new mapping region. Delete the management information for each of the mapping regions to be merged.

5. The management method according to claim 4, characterized in that: When there are multiple feasible merging combinations within the same logical address range, the judgment is made in ascending order of logical starting address, and the combination with the largest granularity after merging is selected first; when there are multiple combinations with the same granularity after merging, the combination with the smallest logical starting address is merged first; if the total length of each of the mapping regions to be merged cannot form the preset standard region granularity, then they are not merged.

6. The management method according to claim 4, characterized in that: The initial heat data of the new mapping region is taken as the highest heat of each of the mapping regions to be merged, and then recalculated using a timer and a sliding window.

7. The management method according to claim 1, characterized in that: The splitting or merging operations are all based on the logical starting address and length of the mapping region to determine the operation object. The region identifier is only used as a unique identifier of the mapping region and does not indicate the arrangement position of the mapping region in the logical address space. After splitting or merging, an unused region identifier is assigned to the new mapping region, and the management information corresponding to the original mapping region identifier is deleted.

8. The management method according to claim 1, characterized in that: The heat data is calculated using the following formula: in, This represents the total number of writes to the mapped region. The access frequency of the mapped region. The recent access frequency of the mapped region is scored; , , These are the preset weighting coefficients.

9. A solid-state drive, characterized in that: The management method described in any one of claims 1-8 shall be adopted.

Images