CN115033180A - Cache replacement method and system for SMR-SSD hybrid storage system based on locality awareness - Google Patents

Abstract

The invention discloses a cache replacement method and a cache replacement system of an SMR-SSD hybrid storage system based on locality perception, wherein data of a cache block is divided according to a physical position, access recency and clean/dirty data; according to the physical position of the cache block, a Band-based management mode is adopted, the elimination sequence of bands is established according to the Band weight, and the corresponding bands are selected according to the weight for elimination; according to the access recency of the cache blocks, adopting a dual locality comparison strategy to sort the Band weights of the cache blocks, and placing k bands with best spatial locality and worst temporal locality at the tail end of a queue as candidate obsolete bands; according to clean/dirty data of the cache block, adopting a clean Band priority strategy to recalculate and sort the weights of the candidate obsolete bands, and selecting the Band with the lowest weight for obsolete; and writing the eliminated cache blocks back to the disk, and taking the released SSD space as the address space of the newly written block to complete the cache replacement process. The invention reduces the quantity of RMW to the maximum extent and improves the system performance.

Description

Locality-aware-based cache replacement method for SMR-SSD hybrid storage system system

technical field

The invention belongs to the technical field of computer system architecture, and in particular relates to a method and system for cache replacement of an SMR-SSD hybrid storage system based on locality awareness.

Background technique

Shingled Magnetic Recording (SMR) is a magnetic storage data recording technology for magnetic disks, which increases track density by overlapping tracks, compared with conventional magnetic recording (Conventional Magnetic Recording, CMR) technology. Disk, which can provide higher track density and thus increase storage capacity. Due to its large capacity and low cost, SMR disks are widely used in data center cold archive storage, low frequency storage and other fields. However, with the expansion of the demand for low-cost storage, the challenge of how to apply SMR disks to the standard storage field becomes more and more obvious.

Currently, SMR management can be implemented at three levels: the internal disk device, the host side, the host side, and disk collaboration, corresponding to three types of disk devices: DM-SMR, HM-SMR, and HA-SMR. The driver-managed SMR disk (DM-SMR) is compatible with the traditional disk IO stack, and the software does not need to be modified, and it is plug-and-play; however, there is a persistent buffer inside this type of device that serves non-sequential write requests, and in the face of non-sequential write intensive scenarios At this time, the STL algorithm will periodically clean up the buffer, which will block the upper-layer IO request and cause great jitter in application performance. The optimization scheme of this patent is aimed at DM-SMR disks, and the default SMR disks in subsequent texts refer to DM-SMR disks.

Since the internal tracks of the SMR disk overlap, the SMR disk cannot directly perform in-situ update and write when writing data, but performs the RMW (Read-Modify-Write) operation. While solving the random write problem of SMR disks, RMW operation also brings serious write amplification and performance fluctuation problems. By analyzing Pearson Correlation Coefficients (PCC), SUN et al. concluded that RMW is the key to affecting the performance of SMR disks.

The hybrid storage system composed of SSD and SMR disk can alleviate the above problems to a certain extent, but because traditional cache replacement algorithms, such as LRU, etc., do not take into account the internal organizational structure characteristics of SMR disk, the trigger for reducing RMW cannot be generated. Any help. At present, the cache replacement algorithms for SSD-SMR disks mainly include two categories, one is LRU-like and the other is Band-based.

LRU-like: The LRU-like algorithm still uses the LRU queue as the core. The PORE algorithm uses the LRU to organize the cache queue. By limiting the elimination range of cache blocks, the possibility of write amplification is indirectly reduced, but because the algorithm does not distinguish between Clean and dirty blocks. The SAC algorithm goes a step further on the PORE algorithm, adding dynamic management of clean and dirty blocks, dividing clean blocks and dirty blocks into two LRU queues for management, but because its core algorithm is still LRU, it does not break away from the constraints of traditional algorithms .

Band-based: MOST tends to slow down the write amplification strategy, and proposes the band-based management method for the first time, but it does not consider the impact of the cache hit rate, and only eliminates the cache based on the number of cache blocks in the band, ignoring the core of the cache elements. However, the band-based management method can maximize the aggregation degree of eliminated cache blocks, that is, spatial locality.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to provide a locality-aware-based SMR-SSD hybrid storage system cache replacement method and system to solve the problem of low hit rate of traditional band-based algorithms; To a certain extent, the amount of RMW generated is reduced and the system performance is improved.

The present invention adopts following technical scheme:

A locality-aware SMR-SSD hybrid storage system cache replacement method, and a locality-aware SMR-SSD hybrid storage system cache replacement method, characterized in that it includes the following steps:

Divide cache block data by physical location, access recency, and clean/dirty data;

According to the physical location of the cache block, the Band-based management method is adopted, the band elimination order is established according to the band weight, and the corresponding band is selected for elimination according to the weight;

According to the access recency of the cache block, the double locality comparison strategy DLC is used to sort the band weights of the cache block, and the k Bands with the best spatial locality and the worst temporal locality are placed at the end of the queue as candidate elimination bands;

According to the clean/dirty data of the cache block, the clean band priority strategy is used to recalculate and sort the weights of the candidate bands for elimination, and select the band with the lowest weight for elimination; write the eliminated cache blocks back to the disk, and the freed SSD space is used as a new write. block address space to complete the cache replacement process.

Specifically, the double locality comparison strategy DLC is as follows:

Distinguish the cache blocks in the band according to the access frequency and access recency; calculate the temporal locality weight of the cache blocks in the band according to the calculation formula of time locality; according to the calculation formula of spatial locality, calculate the cache blocks in the band The spatial locality weight of the band is calculated; the total weight performance of the band is calculated and sorted.

Further, the temporal locality weight W _TPL (b) is specifically calculated as follows:

W _TPL (b)=∑F(cur_time-reference_time)

Among them, F(x) is the time locality of a single block, cur_time is the current timestamp, and reference_time is the timestamp when it was accessed.

Further, the spatial locality weight W _SPL (b) is specifically calculated as follows:

W _SPL (b)=SPL _band (b)

where F(x) is the temporal locality of a single block, and SPL _band (b) is the number of cache blocks in a single band.

Further, the total weight of the band W _band (b) is:

W _band (b)=αW _SPL (b)-βW _TPL (b)+γW

Among them, W is a numerical value related to clean blocks and dirty blocks, and α, β, and γ are coefficients.

Specifically, in the relatively clean band-first RCBF strategy, the cost of band elimination, Cost _band , is positively correlated with the number of dirty blocks eliminated, as follows:

Time _band ∝Write _time *N _dirty

Among them, Time _band is the time required to eliminate the band, Write _time is the time required to write back the cache block, and N _dirty is the number of dirty blocks.

Further, the elimination cost of clean blocks, Cost _clean , and the elimination cost of dirty blocks, Cost _dirty , are as follows:

Cost _clean =Cost _{access_miss} *P(access)

Cost _dirty =Cost _{write_back} +Cost _{access_miss} *P(access)

Among them, Cache _clean and Cache _dirty represent the time it takes for a block to be eliminated from the cache, P(access) represents the probability that the block is re-read, Cost _{write_back} represents the time it takes for a dirty block to be written back to disk, and Cost _{access_miss} represents an access miss cost to be spent.

Further, the cost of eliminating the band Cost _band is:

Cost _band =Cost _{write_back} *N _dirty +Cost _{access_miss} *N*P(access)

Among them, N _clean and N _dirty represent the number of clean and dirty blocks in the band, respectively, and N represents the number of all blocks in the band.

Specifically, the Band-based management method is as follows:

When inserting a cache block, calculate the band to which the cache block belongs based on the offset of the cache block; build a hash table according to the offset of the cache block and the band to which the cache block belongs; calculate the band weight, and establish the band elimination order according to the calculated band weight; when the SSDcache space is insufficient , select a suitable band in the band elimination order for elimination according to the weight.

In a second aspect, an embodiment of the present invention provides a locality awareness-based SMR-SSD hybrid storage system cache replacement system, including:

The division module divides the data of cache blocks according to physical location, access recency and clean/dirty data;

The elimination module, according to the physical location of the cache block, adopts the Band-based management method, establishes the elimination order of the bands according to the band weight, and selects the corresponding band for elimination according to the weight;

The sorting module uses the double locality comparison strategy DLC to sort the band weights of the cache blocks according to the access recency of the cache blocks, and places the k bands with the best spatial locality and the worst time locality at the end of the queue as candidates for elimination band;

The replacement module, according to the clean/dirty data of the cache block, adopts the clean band priority strategy to recalculate and sort the weights of the candidate bands for elimination, and select the band with the lowest weight for elimination; write the eliminated cache blocks back to the disk, and the freed SSD space is used as The address space of the newly written block completes the cache replacement process.

Compared with the prior art, the present invention at least has the following beneficial effects:

The cache replacement method of SMR-SSD hybrid storage system based on locality awareness divides the data of cache blocks into three types according to physical location, access recency and clean/dirty data; The weight establishes the band elimination order, and selects the corresponding band for elimination according to the weight, which improves the hit rate of the algorithm; the k bands with the best spatial locality and the worst time locality are placed at the end of the queue as candidate elimination bands; clean The band priority strategy recalculates and sorts the weights of the candidate bands for elimination, and selects the band with the lowest weight for elimination; writes the eliminated cache blocks back to the disk, and the released SSD space is used as the address space for the newly written blocks to complete the cache replacement process. The goal of minimizing the number of disk read, modify, and write operations is achieved, improving overall system performance.

Further, the double locality comparison strategy comprehensively considers the temporal locality and spatial locality of the Band cache block, avoiding the repeated reading and writing of hot blocks caused by the traditional Band-based algorithm only relying on spatial locality, and the traditional LRU. The algorithm only relies on the SMR disk write amplification problem caused by time locality, thereby improving the hit rate of the cache system.

Further, the temporal locality weight is an indicator to measure the hotness of the cache block, which is calculated by using the timestamp of the cache block, and fully considers the coldness and hotness of the cache block in the Band.

Further, the spatial locality weight is an index to measure the aggregation degree of cache blocks. This index uses the physical location information of cache blocks to count the number of cache blocks in a single band, and obtains its spatial locality according to the number of cache blocks.

Further, the total weight of the band comprehensively considers the temporal locality and the spatial locality. The choice of this formula is that the total weight of the band W _band (b) is inversely proportional to the temporal locality and proportional to the spatial locality, and the locality value for linear accumulation.

Further, the relatively clean band priority strategy selects the relatively clean cache block sequence. This is because the elimination cost is positively correlated with the number of dirty blocks, that is to say, the more clean blocks eliminated, the lower the cost. Therefore, the elimination cost can be minimized by minimizing the number of eliminated blocks. , to achieve the goal of minimizing the number of disk read, modify, and write operations.

Further, the cost _dean of clean block elimination is calculated by assuming the cost of reading clean data blocks after an access miss. The elimination cost of dirty blocks, Cost _dirty , is also calculated based on the cost of re-reading dirty blocks after an access miss. From these two formulas, it is obvious that the cost of eliminating clean blocks is much lower than dirty blocks.

Further, the cost of the band, Cost _band , combines the clean block and dirty block costs in the previous step, which more intuitively shows the relationship between the entire band elimination cost and the number of clean blocks.

Further, the Band-based algorithm obtains the band position by calculating the physical position of the cache block, and the spatial locality of the cache block can be improved to the greatest extent by unified management of the band.

It can be understood that, for the beneficial effects of the foregoing second aspect, reference may be made to the relevant descriptions in the foregoing first aspect, which will not be repeated here.

To sum up, the present invention improves the problem of the low hit rate of the traditional band-based algorithm, and can achieve a hit rate similar to that of the LRU in the actual test. At the same time, the invention significantly reduces the number of triggers of RMW in the SMR-SSD hybrid storage system, and improves the system performance. In practical application, the present invention can alleviate the phenomenon of abnormal performance fluctuation of SMR disks in standard storage scenarios, expand the application field of SMR disks, and reduce the cost of data centers.

The technical solutions of the present invention will be further described in detail below through the accompanying drawings and embodiments.

Description of drawings

FIG. 1 is a schematic diagram of the method of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

In the description of the present invention, it is to be understood that the terms "comprising" and "comprising" indicate the presence of the described features, integers, steps, operations, elements and/or components, but do not exclude one or more other features, The existence or addition of a whole, step, operation, element, component, and/or a collection thereof.

It should also be understood that the terminology used in the present specification is for the purpose of describing particular embodiments only and is not intended to limit the present invention. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural unless the context clearly dictates otherwise.

It should further be understood that, as used in this specification and the appended claims, the term "and/or" refers to and including any and all possible combinations of one or more of the associated listed items , for example, A and/or B, can mean that A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" in this document generally indicates that the related objects are an "or" relationship.

It should be understood that although the terms first, second, third, etc. may be used in the embodiments of the present invention to describe the preset range and the like, these preset ranges should not be limited to these terms. These terms are only used to distinguish preset ranges from one another. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

Depending on the context, the word "if" as used herein can be interpreted as "at" or "when" or "in response to determining" or "in response to detecting." Similarly, the phrases "if determined" or "if detected (the stated condition or event)" can be interpreted as "when determined" or "in response to determining" or "when detected (the stated condition or event)," depending on the context )" or "in response to detection (a stated condition or event)".

Various structural schematic diagrams according to the disclosed embodiments of the present invention are shown in the accompanying drawings. The figures are not to scale, some details have been exaggerated for clarity, and some details may have been omitted. The shapes of various regions and layers shown in the figures and their relative sizes and positional relationships are only exemplary, and in practice, there may be deviations due to manufacturing tolerances or technical limitations, and those skilled in the art should Regions/layers with different shapes, sizes, relative positions can be additionally designed as desired.

Referring to FIG. 1, the present invention provides a locality-aware-based SMR-SSD hybrid storage system cache replacement method. The SLA (Structural Learning Algorithm) algorithm architecture includes a Band-based management method (BM), a double locality comparison strategy ( DLC) and Relatively Clean Band Priority (RCBF);

The Band-based management method and the relatively clean band priority strategy are responsible for reducing the number of RMWs generated. The double locality comparison strategy is responsible for solving the problem of low hit rate caused by the Band-based management method. There are three types of clean/dirty data. Data blocks with the same color in Figure 1 represent the same band, F represents the number of data blocks in the band, R represents recently accessed data, and C represents clean data.

The present invention is a locality-aware-based SMR-SSD hybrid storage system cache replacement method, comprising the following steps:

S1, Band-based management method;

S101. When inserting a cache block, calculate the band to which the cache block belongs by using the offset of the cache block;

S102, build a hash table according to the offset of the cache block and the band to which the cache block obtained in step S101 belongs;

S103. Calculate the weight of the band, and establish the elimination order of the band according to the calculated weight of the band;

S104 , when the cache space is insufficient, select a suitable band for elimination from the band elimination order established in step S103 according to the weight.

The weight calculation of the traditional band-based management method is only based on the number of cache blocks in the band. Elimination is to give priority to the band with the largest number of blocks, regardless of whether the data is hot or cold, which will cause a serious hit rate drop. Therefore, the present invention will use the double locality comparison strategy DLC and the relatively clean band priority RCBF strategy to recalculate and sort the weights of the bands. Specific steps are as follows:

S2, double locality comparison strategy;

S201. Distinguish the cache blocks in the band according to access frequency and access recency;

S202, according to the calculation formula of time locality, calculate the overall time locality of the cache block in the band;

By analyzing the spatial locality and time locality of the band, it is found that the problem with the traditional band-based management method is that when the band with the largest number of blocks is eliminated first, more hot data blocks will be eliminated, which is also a band-based algorithm. The reason for the low hit rate.

The temporal locality weight and the spatial locality weight W _TPL (b) are calculated as follows:

W _TPL (b)=∑F(cur_time-reference_time) (1)

Among them, F(x) is the time locality of a single block, cur_time is the current timestamp, and reference_time is the timestamp when it was accessed.

S203, according to the calculation formula of spatial locality, calculate the overall spatial locality of the cache block in the band;

The spatial locality weight W _SPL (b) is calculated as follows:

W _SPL (b) = SPL _band (b) (2)

where F(x) is the temporal locality of a single block, and SPL _band (b) is the number of cache blocks in the band.

The temporal locality F(x) of a single block is calculated as:

Among them, F(x) is the difference between the current time and the last access time of the block, and the SPL _band (b) of the entire band represents the total number of blocks.

S204. Calculate and sort the overall locality performance of the band.

Due to the complexity of counting the access time and calculation weight of each block, it will increase the complexity of the algorithm. The idea of calculus is used here, and multiple blocks accessed within a similar time period are regarded as the same calculation, or the number of recent accesses is used to calculate replace.

W _band (b)=αW _SPL (b)-βW _TPL (b)+γW (3)

Among them, W is a value related to clean blocks and dirty blocks, α, β, γ are coefficients, and the default is 1.

It can be seen from the above discussion that the total weight of the band W _band (b) is inversely proportional to the temporal locality and proportional to the spatial locality, and since the locality value is linearly accumulated, the formula (3) is used to calculate W _band (b). ).

S3, relatively clean band priority strategy.

In traditional HDDs, the cost of clean blocks and dirty blocks is not much different. However, in SMR disks, due to the write amplification problem, the elimination cost of dirty blocks is much higher than that of clean blocks; but pure CF (clean first) The strategy doesn't work very well because it can seriously affect the cache hit rate.

Considering the problem of miss rereading, the elimination cost of clean block Cost _clean and the elimination cost of dirty block Cost _dirty are expressed as follows:

Cost _clean =Cost _{access_miss} *P(access) (4)

Cost _dirty =Cost _{write_back} +Cost _{access_miss} *P(access)

Among them, Cache _clean and Cache _dirty represent the time it takes for a block to be eliminated from the cache, P(access) represents the probability that the block is re-read, Cost _{write_back} represents the time it takes for a dirty block to be written back to disk, and Cost _{access_miss} represents an access miss cost to be spent.

The elimination mechanism takes the band as the unit, and the Cost _band is used to represent the cost of eliminating the band, as shown in Equation 5:

Cost _band =Cost _clean *N _clean +Cost _dirty *N _dirty (5)

Cost _band =Cost _{write_back} *N _dirty +Cost _{access_miss} *N*P(access)

Among them, N _clean and N _dirty represent the number of clean and dirty blocks in the band, respectively, and N represents the number of all blocks in the band.

In the band-based management method, the candidate elimination band has a similar N value and is regarded as a fixed value; while in the fixed elimination algorithm and trace, P(access) is also regarded as a fixed value.

Therefore, it is concluded that the elimination cost of a band is positively related to the number of dirty blocks in it.

Time _band ∝Write _time *N _dirty (6)

Considering that the number of blocks in the candidate band is approximately the same, that is to say, the more the number of clean blocks, the less the number of dirty blocks. Therefore, the RCBF (Relatively Clean Band First) strategy is proposed, which combines the number of clean blocks in the band. As a positive factor, it is added to the calculation of W _band (b), which is the value of W in Equation 3.

In yet another embodiment of the present invention, a locality-aware-based SMR-SSD hybrid storage system cache replacement system is provided, which can be used to implement the above-mentioned locality-aware-based SMR-SSD hybrid storage system cache replacement method, specifically , the locality-aware-based SMR-SSD hybrid storage system cache replacement system includes a partition module, an elimination module, a sorting module, and a replacement module.

Among them, the division module divides the data of the cache block according to physical location, access recency and clean/dirty data;

The elimination module adopts the Band-based management method according to the physical location, establishes the elimination order of the bands according to the band weight, and selects the corresponding band for elimination according to the weight;

The sorting module uses the double locality comparison strategy DLC to sort the band weights of the cache blocks according to the access recency of the cache blocks, and places the k bands with the best spatial locality and the worst time locality at the end of the queue as candidates for elimination band;

The replacement module, according to the dirtyness of the cache block, adopts the clean band priority strategy to recalculate and sort the weights of the candidate bands for elimination, and select the band with the lowest weight for elimination; write the eliminated cache blocks back to the disk, and the freed SSD space is used as a new write. into the address space of the block to complete the cache replacement process.

To sum up, a method and system for cache replacement of SMR-SSD hybrid storage system based on locality awareness of the present invention improves the problem of low hit rate of traditional band-based algorithms, and can achieve hits similar to LRU in actual tests At the same time, the number of triggers of RMWs in the SMR-SSD hybrid storage system is significantly reduced, and the system performance is improved; in practical applications, the present invention can alleviate the phenomenon of abnormal performance fluctuations of SMR disks in standard storage scenarios, and expand SMR disks application areas, reducing the cost of data centers.

As will be appreciated by those skilled in the art, the embodiments of the present application may be provided as a method, a system, or a computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It will be understood that each flow and/or block in the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing device produce Means for implementing the functions specified in a flow or flow of a flowchart and/or a block or blocks of a block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions The apparatus implements the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.

These computer program instructions can also be loaded on a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process such that The instructions provide steps for implementing the functions specified in the flow or blocks of the flowcharts and/or the block or blocks of the block diagrams.

The above content is only to illustrate the technical idea of the present invention, and cannot limit the protection scope of the present invention. Any changes made on the basis of the technical solution according to the technical idea proposed by the present invention all fall within the scope of the claims of the present invention. within the scope of protection.

Claims (10)

1. The SMR-SSD hybrid storage system cache replacement method based on locality perception, is characterized in that, comprises the following steps: Divide cache block data by physical location, access recency, and clean/dirty data; According to the physical location of the cache block, the Band-based management method is adopted, the band elimination order is established according to the band weight, and the corresponding band is selected for elimination according to the weight; According to the access recency of the cache block, the double locality comparison strategy DLC is used to sort the band weights of the cache block, and the k Bands with the best spatial locality and the worst temporal locality are placed at the end of the queue as candidate elimination bands; According to the clean/dirty data of the cache block, the clean band priority strategy is used to recalculate and sort the weights of the candidate bands for elimination, and select the band with the lowest weight for elimination; write the eliminated cache blocks back to the disk, and the freed SSD space is used as a new write. block address space to complete the cache replacement process. 2. the SMR-SSD hybrid storage system cache replacement method based on locality perception according to claim 1, is characterized in that, double locality comparison strategy DLC is specifically: Distinguish the cache blocks in the band according to the access frequency and access recency; calculate the temporal locality weight of the cache blocks in the band according to the calculation formula of time locality; according to the calculation formula of spatial locality, calculate the cache blocks in the band The spatial locality weight of the band is calculated; the total weight performance of the band is calculated and sorted. 3. the SMR-SSD hybrid storage system cache replacement method based on locality perception according to claim 2, is characterized in that, time locality weight W _TPL (b) is specifically calculated as follows: W _TPL (b)=∑F(cur_time-reference_time) Among them, F(x) is the time locality of a single block, cur_time is the current timestamp, and reference_time is the timestamp when it was accessed. 4. the SMR-SSD hybrid storage system cache replacement method based on locality perception according to claim 2, is characterized in that, spatial locality weight W _SPL (b) is specifically calculated as follows: W _SPL (b)=SPL _band (b) where F(x) is the temporal locality of a single block, and SPL _band (b) is the number of cache blocks in a single band. 5. the SMR-SSD hybrid storage system cache replacement method based on locality perception according to claim 2, is characterized in that, the total weight W _band (b) of band is: W _band (b)=αW _SPL (b)-βW _TPL (b)+γW Among them, W is a numerical value related to clean blocks and dirty blocks, and α, β, and γ are coefficients. 6. The locality-aware-based SMR-SSD hybrid storage system cache replacement method according to claim 1, wherein, in the relatively clean band-priority RCBF strategy, the cost of band elimination, Cost _band , is positive with the number of dirty blocks eliminated therein. related, specifically: Time _band ∝Write _time *N _dirty Among them, Time _band is the time required to eliminate the band, Write _time is the time required to write back the cache block, and N _dirty is the number of dirty blocks. 7. the SMR-SSD hybrid storage system cache replacement method based on locality perception according to claim 6, is characterized in that, the elimination cost Cost _clean of clean block and the elimination cost Cost _dirty of dirty block are as follows: Cost _clean =Cost _{access_miss} *P(access) Cost _dirty =Cost _{write_back} +Cost _{access_miss} *P(access) Among them, Cache _clean and Cache _dirty represent the time it takes for a block to be eliminated from the cache, P(access) represents the probability that the block is re-read, Cost _{write_back} represents the time it takes for a dirty block to be written back to disk, and Cost _{access_miss} represents an access miss cost to be spent. 8. The SMR-SSD hybrid storage system cache replacement method based on locality perception according to claim 7, is characterized in that, the cost of eliminating band Cost _band is: Cost _band =Cost _{write_back} *N _dirty +Cost _{access_miss} *N*P(access) Among them, N _clean and N _dirty represent the number of clean and dirty blocks in the band, respectively, and N represents the number of all blocks in the band. 9. the SMR-SSD hybrid storage system cache replacement method based on locality perception according to claim 1, is characterized in that, Band-based management mode is specifically: When inserting a cache block, calculate the band to which the cache block belongs based on the offset of the cache block; build a hash table according to the offset of the cache block and the band to which the cache block belongs; calculate the band weight, and establish the band elimination order according to the calculated band weight; when the SSDcache space is insufficient , select a suitable band in the band elimination order for elimination according to the weight. 10. A locality-aware-based SMR-SSD hybrid storage system cache replacement system, comprising: The division module divides the data of cache blocks according to physical location, access recency and clean/dirty data; The elimination module, according to the physical location of the cache block, adopts the Band-based management method, establishes the elimination order of the bands according to the band weight, and selects the corresponding band for elimination according to the weight; The sorting module uses the double locality comparison strategy DLC to sort the band weights of the cache blocks according to the access recency of the cache blocks, and places the k bands with the best spatial locality and the worst time locality at the end of the queue as candidates for elimination band; The replacement module, according to the clean/dirty data of the cache block, adopts the clean band priority strategy to recalculate and sort the weights of the candidate bands for elimination, and select the band with the lowest weight for elimination; write the eliminated cache blocks back to the disk, and the freed SSD space is used as The address space of the newly written block completes the cache replacement process.

Images