KR20220138269A - Data storage device dynamically determinning whether data is hot or cold and operation method of the same - Google Patents

Abstract

A data storage device that dynamically determines whether data is hot/cold is disclosed. The data storage device that dynamically determines whether data is hot/cold according to various embodiments of the present disclosure may include: a host interface that communicates with the host device; a flash memory; and at least one processor. The processor performs: based on a stack distance of workload data, selecting a bloom filter to record an indicator for the workload data from among the plurality of bloom filters; based on at least one hash function, writing the indicator for the workload data to the selected bloom filter; obtaining a hot/cold decision value of the workload data based on the indicator for the workload data recorded in each of the plurality of bloom filters and a weight according to recency for each of the plurality of bloom filters; and determining whether the workload data is hot data/cold data based on the determination value.

Description

Data storage device that dynamically determines whether data is hot/cold and its operation method

The present invention relates to a data storage device for dynamically determining whether data is hot/cold, and more particularly, to a data storage device for dynamically determining whether data is hot/cold using a plurality of bloom filters.

Flash memory has advantages of low power, high performance, and high durability. Recently, as the price of flash memory has decreased and capacity has increased, a solid state drive (SDD), which is a storage device based on a NAND flash memory, is replacing an existing hard disk driver. SSD consists of several NAND flash chips, DRAM buffers and a controller, and it is possible to read/write operations to multiple chips at the same time by utilizing multi-channel and multi-way structures.

A solid state disk (SSD) stores information using a non-volatile NAND flash memory. SSD can input/output data at high speed without search time by random access, while mechanical delay or failure rate is significantly low. In addition, SSDs do not damage data due to external shocks, reduce heat, noise, and power consumption, and have the advantages of being compact and lightweight.

The core technology of SSD is FTL (Flash Translation Layer). Among the core functions of FTL, GC (Garbage Collection) and WL (Wear Leveling) greatly affect the performance and lifespan of an SSD.

GC and WL basically adopt the hot/cold data classification technique. However, in the case of the existing hot/cold data classification mechanism, it merely considers the frequency of data access in classifying hot/cold data.

As an example of the technology that is the background of the present invention, Republic of Korea Patent Registration No. 10-1311031 (September 24, 2013) relates to a method of efficiently accessing a plurality of storage devices with a distributed load, each of a plurality of data Disclosed is a method of classifying hot data and cold data based on an access time.

The present invention has been devised to solve the above problems, and the present invention considers recency in addition to the access frequency in determining whether data is hot/cold, so that the more recently accessed data is, the hotter it is. It aims to provide a way to define data.

In addition, the present invention provides a dynamic method by adaptively changing a criterion for classifying hot/cold data for each different workload data, unlike the conventional technology that statically divides hot/cold data. Data can be divided into hot/cold by (dynamic).

A data storage device for dynamically determining whether data is hot/cold according to various embodiments of the present disclosure includes a host interface that communicates with a host device, a flash memory, and a stack distance of workload data. select a bloom filter to record the indicator for the workload data among the bloom filters, and record the indicator for the workload data in the selected bloom filter, based on at least one hash function, and the plurality of bloom filters Obtain a hot/cold determination value of the workload data based on the weight according to the freshness for each of the indicator and the plurality of bloom filters for the workload data recorded in each, and based on the determination value, the work It may include at least one processor that determines whether the load data is hot data/cold data.

In the method of operating a data storage device for dynamically determining whether data is hot/cold according to various embodiments of the present disclosure, an indicator for the workload data among a plurality of bloom filters is displayed based on a stack distance of the workload data. The process of selecting a bloom filter to be recorded, the process of recording the indicator for the workload data to the selected bloom filter based on at least one hash function, the workload data recorded in each of the plurality of bloom filters A process of obtaining a hot/cold decision value of the workload data based on the weight according to the freshness of each of the indicator and the plurality of bloom filters, and hot data/cold data of the workload data based on the decision value It may include the process of determining whether or not

According to various embodiments of the present invention, in the present invention, in determining whether data is hot/cold, it is important to consider recency in addition to the access frequency, so that more recently accessed data can be defined as hot data. have.

In addition, the present invention provides a dynamic method by adaptively changing a criterion for classifying hot/cold data for each different workload data, unlike the conventional technology that statically divides hot/cold data. Data can be divided into hot/cold by (dynamic).

1 is a block diagram of a data storage device according to an embodiment of the present invention.
2 is an architecture of a flash system according to an embodiment of the present invention.
3 illustrates a framework for a plurality of bloom filters according to an embodiment of the present invention.
4 is a block diagram of a data storage device according to an embodiment of the present invention.
5 and 6 are flowcharts of a method for determining hot/cold data according to an embodiment of the present invention.
7 is a flowchart illustrating a method of operating a storage device for dynamically determining whether data is hot/cold according to an embodiment of the present invention.

Hereinafter, the principle of operation of a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings. In addition, when it is determined that a detailed description of a related well-known function or configuration may obscure the gist of the present disclosure in describing an embodiment of the present invention, the detailed description thereof will be omitted. And the terms used below are terms defined in consideration of functions in the present invention, which may vary depending on the intention or custom of the user or operator. Therefore, the definitions of the terms used should be interpreted based on the contents and corresponding functions throughout this specification.

1 is a block diagram of a data storage device according to an embodiment of the present invention.

Hereinafter, the data storage device 1 will be described by taking a solid state drive (SSD) as an example for convenience of description, but the usage example will not be limited thereto. For example, an electronic device including a NAND or a cache, for example, a smart phone, a PC, etc. may also be included in the data storage device 1 .

In addition, various embodiments of the present invention, which will be described later, may be applied to various data processing technologies, such as a storage tiering technology for tying storage of different performance.

Referring to FIG. 1 , the data storage device 1 may include a host interface 11 , a RAM buffer 12 , a flash memory 13 , and a controller 14 .

The host interface 11 may transmit a signal received from the host to the controller 12 .

The host interface 11 may be configured to connect the data storage device 1 with a host. The host interface 11 may be compatible with Peripheral Component Interconnect-Express (PCIe), Serial Attached SCSI (SAS), Serial AT Attachment (SATA), or other configurations known in the art.

In addition, the data storage device 1 may be connected to the host through a network fabric such as Ethernet, Fiber Channel, InfiniBand, and other network fabrics. In this case, the data storage device 1 may be compatible with NVMe (Non-Volatile Memory express, non-volatile memory) through the fabrics (NVMe-oF).

The RAM buffer 12 may store a mapping table of data stored in the NAND. For example, since the NAND constituting the SSD is composed of blocks and pages, an address mapping that substitutes a logical address on the operating system and a physical address of the NAND is required. This mapping relationship is stored in the RAM buffer 12 and stored in the controller 2 ) can be provided.

In addition, the RAM buffer 12 may temporarily store data so that processed data can be transmitted according to a mode supported by the CPU and the bus system.

The flash memory 13 is a nonvolatile semiconductor storage device. The flash memory 13 may read/write data under the control of the controller 14 .

For example, the flash memory 13 may have a form in which several NAND flashes are connected in parallel. Examples of the flash memory 14 are SLC (Single Level Cell), MLC (Multi Level Cell). There are Triple Level Cell (TLC), Quad Level Cell (QLC), and Penta Level Cells (PLC) types.

The controller 14 may perform read, write, life management, and the like through the data storage device 1 . The controller 14 may perform the above-described address mapping in a Flash Translation Layer (FTL), and may perform wear leveling and garbage collection to maintain long-term performance and improve lifespan of NAND.

Here, the FTL is a firmware layer that converts an address (logical address) given by the operating system into a flash address (physical address). In the garbage collection, unused areas are accumulated while using the data storage device 1, and these areas are collected and deleted. Wear leveling is the operation of changing the physical address itself connected to the logical address by modifying the contents of the address mapping table.

The controller 14 may include a processor 141 , a flash controller 142 , and a buffer manager 143 .

The processor 141 may read/execute various commands required by the host.

The processor 141 may include one or more processing cores. For example, the processor 141 may include a plurality of ARM processing cores such as cortex processing cores (Advanced RISC Machine processing cores).

For example, the processor 141 may be configured to operate according to embedded firmware pre-stored in the NOR flash.

The flash controller 142 may be configured to perform various types of maintenance and other tasks. Also, the flash controller 142 may be configured to manage input/output of data. For example, the flash controller 142 may be configured to provide wear leveling, block picking, garbage collection and encryption, and may be configured to provide a flash translation layer and mapping, and the like.

Also, the flash controller 142 may connect to various groups included in the flash memory using one or more channels.

The buffer manager 143 may be configured to provide a buffer between the host interface 130 and the flash controller 150 to facilitate input/output requests. For example, the buffer manager 143 may include an integrated SRAM as well as a DRAM controller coupled to the RAM buffer 12 .

The configuration of the data storage device 1 has been described in detail in the above description. Hereinafter, a method of dynamically determining whether data is hot/cold according to various embodiments of the present invention will be described in detail with reference to FIGS. 2 to 7 . Hereinafter, it is assumed that the method of dynamically determining whether hot/cold is operated by the controller 14 of the data storage device 1 .

2 is an architecture of a flash system according to an embodiment of the present invention.

Since a general FTL configuration is known, a detailed description thereof is omitted here.

The FTL according to an embodiment of the present invention may include a hot/cold data identifier (Hot DATA Identifier).

The hot/cold data identification unit 21 may perform hot/cold data identification using a plurality of bloom filters.

Here, although there are various ways of defining hot data and cold data, in general, hot data may be defined as frequently modified data, and cold data may be defined as data that is not frequently modified.

If hot data and cold data are stored in the same page, whenever the hot data is changed, the cold data must be copied together with the hot data in the Read-Modify-Write operation. Also, for wear leveling, cold data must be continuously moved to another page. Cold data and hot data must be separated as much as possible so that the above-described GC can be efficiently processed.

For example, the hot/cold data identification unit 21 performs a process of selecting a bloom filter to record an indicator for workload data among a plurality of bloom filters based on a stack distance of the workload data. can Also, the hot/cold data identification unit 21 may record an indicator for the workload data in the selected bloom filter based on at least one hash function.

In this case, the hot/cold data identification unit 21 determines the hot/cold determination value of data based on an indicator for workload data recorded in each of the plurality of bloom filters and a weight according to the freshness of each of the plurality of bloom filters. can be obtained Also, the hot/cold data identification unit 21 may determine whether the workload data is hot data/cold data based on the obtained determination value.

3 illustrates a framework for a plurality of bloom filters according to an embodiment of the present invention.

Specifically, FIG. 3 shows a framework for a method of determining (identifying) hot/cold data using a plurality of bloom filters.

Referring to FIG. 3 , V independent bloom filters and K independent hash functions are defined. Each of the V bloom filters may consist of M bits for recording K hash values.

For example, when a first write request is received in the Flash Translation Layer (FTL), the corresponding LBA may be hashed by K hash functions. The K hash output value can be written as 1 in the corresponding bit position of the first bloom filter. In this case, the output values of the hash function range from 1 to M, and each hash output value corresponds to each bit position of the M-bit bloom filter.

When a second write request is received, a second bloom filter may be selected, and the K hash output value may be written as 1 to the corresponding bit position of the second bloom filter. Here, the bloom filter selection method may be a round robin fashion.

Each of the aforementioned bloom filters may be deleted or reset when all M bits are written or according to a predefined time period.

The hot/cold data determination method using a plurality of bloom filters according to an embodiment of the present invention does not use a bloom filter counter for counting the number of LBA appearances. That is, in the hot/cold data determination method using a plurality of bloom filters according to an embodiment of the present invention, the LBA frequency may be stored by checking each of the plurality of bloom filters and counting whether a specific LBA has been recorded.

For example, when a specific LBA is input, the controller 14 may select one of a plurality of bloom filters and write 1 to a corresponding bit position. If the hash output value of the specific LBA is recorded in the selected bloom filter, the controller 14 may check sequentially (round robin method) until it finds another bloom filter that does not record the hash output value of the specific LBA. If another bloom filter that does not record the hash output value of a specific LBA is found, the controller 14 may record the hash output value of the specific LBA.

4 is a block diagram of a data storage device according to an embodiment of the present invention.

Referring to FIG. 4 , the data storage device 1 ′ may include a workload data analysis unit 41 , a bloom filter selection unit 42 , and a hot/cold data classification unit 43 .

The workload data analysis unit 41 calculates a stack distance (SD) for the input workload data. The workload data analyzer 41 may calculate the stack distance whenever workload data is input. For example, the workload data may be defined as a logical block address (LBA).

The stack distance may be defined as the number of unique objects among multiple objects appearing between two identical objects. For example, assuming a list of objects 'a b c c c d e e e a b', the unique object in the object 'b c c c d e e e' appearing between the same 'a' objects is 'b c d e'. Accordingly, the stack distance of the array of objects may be defined as 4 (SD(a)=4).

Because the stack distance can be calculated only when the history of all objects that appear (or are accessed) is stored, the memory usage is large, and in particular, in the case of the cache mechanism, the computational complexity is very high because SD must be obtained for each object. In general, there is a doubly linked data structure for implementing a stack.

)를 계산할 수 있다. 최신성 가중치(

)는 다음의 수학식 1에 기초하여 계산될 수 있다.The workload data analysis unit 41 calculates a freshness weight (

) can be calculated. Freshness weight (

) can be calculated based on Equation 1 below.

는 최신성 가중치,

는 스택 거리,

는 평균 스택 거리,

이다.here,

is the freshness weight,

is the stack distance,

is the average stack distance,

to be.

The workload data analysis unit 41 may give greater weight to freshness than frequency in the case of workload data in which recency is more important, that is, when the stack distance is relatively small through workload analysis. have.

Conversely, when the stack distance is relatively small, the workload data analyzer 41 may give a greater weight to the frequency than to the freshness.

The bloom filter selection unit 42 may select a bloom filter to record an indicator for workload data among a plurality of bloom filters based on the freshness weight. Here, the indicator may be defined as a display value for the bit of the bloom filter corresponding to the hash output value of the workload data. Also, the bloom filter selection unit 42 may be defined as a freshness weight allocator.

로 설정하고, 복수의 블룸필터 중 가장 최근에 리셋된 블룸필터 외의 적어도 하나의 블룸필터의 선택 확률은

로 설정하여, 복수의 블룸필터 중 지시자를 기록할 블룸필터를 선택할 수 있다.As an example, the bloom filter selection unit 42 determines that the selection probability of the most recently reset bloom filter among the plurality of bloom filters is

, and the selection probability of at least one bloom filter other than the most recently reset bloom filter among the plurality of bloom filters is

By setting to , it is possible to select a bloom filter to record an indicator among a plurality of bloom filters.

The hot/cold data classification unit 43 may classify and store whether the workload data is hot data or cold data based on the workload data and number information of the selected bloom filter. Here, the hot/cold data classification unit 43 may use a plurality of bloom filters.

The hot/cold data classification unit 43 obtains a hot/cold determination value of the workload data based on an indicator for the workload data recorded in each of the plurality of bloom filters and a weight according to the freshness of each of the plurality of bloom filters can do.

Specifically, the hot/cold data classification unit 43 may obtain a hot/cold determination value based on Equation 2 below.

는 핫/콜드 판정 값,

는 n번째 BF에 상기 워크로드 데이터가 기록된 경우 1, 아닌 경우 0,

는 n번째 BF에 대한 기정의된 exponential 함수의 출력 값이다.here,

is the hot/cold judgment value,

is 1 when the workload data is written to the nth BF, 0 otherwise,

is the output value of the predefined exponential function for the nth BF.

은 복수의 블룸필터 각각에 대한 최신성에 따른 가중치이다. 즉, 복수의 블룸필터 중 최근에 리셋된 블룸필터일수록 최신의 워크로드 데이터에 대한 지시자를 기록하고 있으므로, 복수의 블룸필터 중 최근에 리셋된 블룸필터일수록 최신성에 따른 가중치가 크게 설정될 수 있다.Each of the plurality of bloom filters may be reset according to a predefined period. In this case, the

is a weight according to the freshness of each of the plurality of bloom filters. That is, since a recently reset bloom filter among a plurality of bloom filters records an indicator for the latest workload data, a more recently reset bloom filter among a plurality of bloom filters may have a greater weight according to freshness.

은 상기 수학식 1의

와 동일한 함수로 정의될 수도 있다.For example, the

is in Equation 1 above

It can also be defined as the same function as

As an example, it is assumed that four bloom filters are used. Also, it is assumed that an indicator indicating workload data input to only the first bloom filter and the second bloom filter among the plurality of bloom filters is recorded.

를 계산할 수 있다.In this case, the hot/cold data classification unit 43 determines the hot/cold determination value based on the following Equation (3).

can be calculated.

The hot/cold data classification unit 43 may determine whether the workload data is hot data/cold data based on the obtained determination value.

For example, the hot/cold data classification unit 43 determines that the workload data is hot data when the hot/cold determination value is greater than a predefined threshold value, and the hot/cold determination value is smaller than the predefined threshold value. , the workload data can be determined as cold data.

According to various embodiments of the present invention described above, the data storage device 1 ′ dynamically performs hot/cold determination on workload data by calculating a stack distance and a hot/cold determination value for each input workload data. can

5 and 6 are flowcharts of a method for determining hot/cold data according to an embodiment of the present invention.

Hereinafter, since the contents of FIG. 4 are applied to the contents of FIGS. 5 and 6, overlapping descriptions will be omitted for convenience of description. 5 and 6 should be interpreted as a continuous process.

Referring to FIG. 5 , the controller 14 may receive workload data ( S501 ). The controller 14 may obtain a stack distance for the input workload data ( S502 ).

The controller 14 may determine a freshness weight for the workload data based on the obtained stack distance (S503).

가 결정되면, 복수의 블룸필터 각각에 대한 선택 확률을 결정할 수 있다. The controller 14 may select a bloom filter to record an indicator for the workload data from among a plurality of bloom filters based on the freshness weight ( S504 ). For example, the controller 14 may weight the freshness

If is determined, a selection probability for each of the plurality of bloom filters may be determined.

로 설정하고, 복수의 블룸필터 중 가장 최근에 리셋된 블룸필터 외의 적어도 하나의 블룸필터의 선택 확률은

로 설정하여, 복수의 블룸필터 중 지시자를 기록할 블룸필터를 선택할 수 있다.As an example, the bloom filter selection unit 42 determines that the selection probability of the most recently reset bloom filter among the plurality of bloom filters is

, and the selection probability of at least one bloom filter other than the most recently reset bloom filter among the plurality of bloom filters is

By setting to , it is possible to select a bloom filter to record an indicator among a plurality of bloom filters.

The controller 14 may determine whether the bloom filter reset period has been reached (S505). For example, when the reset period of the selected bloom filter is satisfied (S505 - Yes), all data of the selected bloom filter may be initialized (S506). As another example, when the reset period of the selected bloom filter is not satisfied (S505 - No), the hashing process for recording the indicator for the workload data may be started without data initialization of the bloom filter.

The controller 14 may obtain a hash output value (hash value) for the workload data based on at least one hash function (S507).

Also, the controller 14 may set the position corresponding to the hash output value on the selected bloom filter to 1 to record the input of the workload data (S508).

)을 획득할 수 있다(S509).In addition, the controller 14 determines the hot/cold determination value (

) can be obtained (S509).

When the hot/cold determination value is greater than the predefined threshold (S510-Yes), the controller 14 may classify the selected workload data as cold data (S511). Alternatively, when the hot/cold determination value is not greater than the predefined threshold (S510-No), the controller 14 may classify the selected workload data as hot data (S512).

According to various embodiments of the present invention described above, in determining whether data is hot/cold, by considering recency in addition to the frequency of access as important, the more recently accessed data can be defined as hot data. .

In addition, unlike the existing technology that statically classifies hot/cold data, it is dynamic by adaptively changing the criteria for classifying hot/cold data for each different workload data. can distinguish between hot/cold data.

7 is a flowchart illustrating a method of operating a storage device for dynamically determining whether data is hot/cold according to an embodiment of the present invention.

Referring to FIG. 7 , in a method of operating a storage device for dynamically determining whether data is hot/cold, a bloom filter for recording an indicator for workload data among a plurality of bloom filters is selected based on a stack distance of the workload data. A process of selecting (S710), a process of recording an indicator for workload data in the selected bloom filter based on at least one hash function (S720), an indicator and a plurality of workload data recorded in each of the plurality of bloom filters A process of obtaining a hot/cold decision value of the workload data based on a weight according to the freshness of each bloom filter of ( S730 ) and a process of determining whether the workload data is hot data/cold data based on the decision value (S740) may be included.

The above-described method of operating a storage device for dynamically determining whether data is hot/cold includes a process of determining a freshness weight for the workload data based on the stack distance of the workload data, and Based on the plurality of bloom filters, the method may further include selecting a bloom filter to record the indicator for the workload data.

In an embodiment of the present invention described above, the freshness weight may be obtained based on the following equation.

는 최신성 가중치,

는 스택 거리,

는 평균 스택 거리.here,

is the freshness weight,

is the stack distance,

is the average stack distance.

로 설정되고, 복수의 블룸필터 중 가장 최근에 리셋된 블룸필터 외의 적어도 하나의 블룸필터의 선택확률은

로 설정될 수 있다.In addition, in the above-described embodiment of the present invention, the selection probability of the most recently reset bloom filter among the plurality of bloom filters is

is set to , and the selection probability of at least one bloom filter other than the most recently reset bloom filter among the plurality of bloom filters is

can be set to

는 아래의 수학식을 만족할 수 있다.In one embodiment of the present invention described above, the

may satisfy the following equation.

In addition, in the above-described embodiment of the present invention, the hot/cold determination value may be obtained based on the following equation.

는 핫/콜드 판정값,

는 n번째 BF에 상기 워크로드 데이터가 기록된 경우 1, 아닌 경우 0,

는 n번째 BF에 대한 기정의된 exponential 함수의 출력값이다.here,

is the hot/cold judgment value,

is 1 when the workload data is written to the nth BF, 0 otherwise,

is the output of the predefined exponential function for the nth BF.

In one embodiment of the present invention described above, if the hot/cold determination value is greater than the predefined threshold value, the workload data is determined as hot data, and if the hot/cold determination value is less than the predefined threshold value, the work Load data may be determined as cold data.

은 복수의 블룸필터 각각에 대한 최신성에 따른 가중치이고, 복수의 블룸필터 중 최근에 리셋된 블룸필터일수록 크다.In one embodiment of the present invention described above, the

is a weight according to the freshness of each of the plurality of bloom filters, and the recently reset bloom filter among the plurality of bloom filters is larger.

In one embodiment of the present invention described above, each of the plurality of bloom filters may be reset according to a predefined period. Also, the workload data may be defined as a logical block address (LBA).

Meanwhile, the above-described method of operating a storage device for dynamically determining whether data is hot/cold according to various embodiments of the present invention is implemented as a computer-executable program code, and thus various non-transitory computer readable media (non-transitory computer readable media) are used. medium) and may be provided to each server or device to be executed by the processor.

For example, in a non-transitory computer-readable medium storing computer instructions for performing an operation of the electronic device when executed by a processor of an electronic device, the operation may include: The process of selecting a bloom filter to record the indicator for the workload data from among the bloom filters of , the process of recording the indicator for the workload data in the selected bloom filter based on at least one hash function, the plurality of A process of obtaining a hot/cold determination value of the workload data based on the weight according to the freshness of each of the indicator and the plurality of bloom filters for the workload data recorded in each bloom filter of It may include a process of determining whether the workload data is hot data/cold data based on it.

For example, on the basis of the stack distance of the workload data, the process of selecting a bloom filter to record an indicator for the workload data from among a plurality of bloom filters, based on at least one hash function, the indicator for the workload data The process of recording in the selected bloom filter, the process of obtaining a hot/cold determination value of the workload data based on an indicator for the workload data recorded in each of the plurality of bloom filters and the weight according to the freshness of each of the plurality of bloom filters and a non-transitory computer readable medium storing a program for performing a process of determining whether the workload data is hot data/cold data based on the determination value.

The non-transitory readable medium refers to a medium that stores data semi-permanently, rather than a medium that stores data for a short moment, such as a register, cache, memory, etc., and can be read by a device. Specifically, the above-described various applications or programs may be provided by being stored in a non-transitory readable medium such as a CD, DVD, hard disk, Blu-ray disk, USB, memory card, ROM, and the like.

While the embodiments of the present invention have been shown and described, various changes in form and details may be made by those skilled in the art without departing from the spirit and scope of the embodiments as defined by the appended claims and their equivalents. you will understand that you can

Data storage: 1, 1'
Host Interface: 11
RAM Buffer: 12
Flash memory: 13
Controller: 14
Processor: 141
Flash controller: 142
Buffer Manager: 143
Workload Data Analysis Department: 41
Bloom filter selector: 42
Hot/Cold Data Classification: 43

Claims (21)

A data storage device for dynamically determining whether data is hot/cold, comprising:
a host interface for communicating with a host device;
flash memory; and
Based on the stack distance of the workload data, select a bloom filter to record the indicator for the workload data from among a plurality of bloom filters,
Based on at least one hash function, writing the indicator for the workload data to the selected bloom filter,
Obtaining a hot/cold determination value of the workload data based on the weight according to the freshness for each of the indicator and the plurality of bloom filters for the workload data recorded in each of the plurality of bloom filters,
and at least one processor to determine whether the workload data is hot data/cold data based on the determination value.
According to claim 1,
the at least one processor,
determine a freshness weight for the workload data based on the stack distance of the workload data;
Selecting a bloom filter to record the indicator for the workload data from among the plurality of bloom filters, based on the freshness weight.
3. The method of claim 2,
The freshness weight is obtained based on the following equation, a data storage device.

here,

is the freshness weight,

is the stack distance,

is the average stack distance.
4. The method of claim 3,
The selection probability of the most recently reset bloom filter among the plurality of bloom filters is

is set, and the selection probability of at least one bloom filter other than the most recently reset bloom filter among the plurality of bloom filters is

data storage device, which is set to .
4. The method of claim 3,
remind

is a data storage device that satisfies the following equation.

According to claim 1,
and the hot/cold determination value is obtained based on the following equation.

here,

is the hot/cold judgment value,

is 1 when the workload data is written to the nth BF, 0 otherwise,

is the output of the predefined exponential function for the nth BF.
7. The method of claim 6,
the at least one processor,
If the hot/cold determination value is greater than a predefined threshold, the workload data is determined as hot data. If the hot/cold determination value is less than a predefined threshold, the workload data is converted to cold data. Judging, data storage.
7. The method of claim 6,
remind

silver,
A weight according to the freshness for each of the plurality of bloom filters,
A data storage device, wherein the bloom filter recently reset among the plurality of bloom filters is larger.
The method of claim 1,
wherein each of the plurality of bloom filters is reset according to a predefined period.
According to claim 1,
The workload data is LBA (Logical Block Address), a data storage device.
In the method of operating a data storage device for dynamically determining whether data is hot/cold,
selecting a bloom filter to record an indicator for the workload data from among a plurality of bloom filters based on a stack distance of the workload data;
recording the indicator for the workload data in the selected bloom filter based on at least one hash function;
obtaining a hot/cold determination value of the workload data based on the indicator for the workload data recorded in each of the plurality of bloom filters and a weight according to the freshness of each of the plurality of bloom filters; and
and determining whether the workload data is hot data/cold data based on the determination value.
12. The method of claim 11,
determining a freshness weight for the workload data based on the stack distance of the workload data; and
Selecting a bloom filter to record the indicator for the workload data from among the plurality of bloom filters based on the freshness weight; further comprising a method of operating a data storage device.
13. The method of claim 12,
The method of operating a data storage device, wherein the freshness weight is obtained based on the following equation.

here,

is the freshness weight,

is the stack distance,

is the average stack distance.
14. The method of claim 13,
The selection probability of the most recently reset bloom filter among the plurality of bloom filters is

is set, and the selection probability of at least one bloom filter other than the most recently reset bloom filter among the plurality of bloom filters is

The method of operation of the data storage device, which is set to .
14. The method of claim 13,
remind

is a method of operating a data storage device that satisfies the following equation.

12. The method of claim 11,
The method of operating a data storage device, wherein the hot/cold determination value is obtained based on the following equation.

here,

is the hot/cold judgment value,

is 1 when the workload data is written to the nth BF, 0 otherwise,

is the output of the predefined exponential function for the nth BF.
17. The method of claim 16,
If the hot/cold determination value is greater than a predefined threshold, the workload data is determined as hot data. If the hot/cold determination value is less than a predefined threshold, the workload data is converted to cold data. A method of operating a data storage device, further comprising a process of determining.
17. The method of claim 16,
remind

silver,
A weight according to the freshness for each of the plurality of bloom filters,
The method of operating a data storage device, wherein the bloom filter recently reset among the plurality of bloom filters is larger.
12. The method of claim 11,
Each of the plurality of bloom filters is reset according to a predefined period, the operating method of the data storage device.
12. The method of claim 11,
The workload data is LBA (Logical Block Address), the operating method of the data storage device.
A non-transitory computer-readable medium storing computer instructions for performing an operation of the electronic device when executed by a processor of an electronic device, the operation comprising:
selecting a bloom filter to record an indicator for the workload data from among a plurality of bloom filters based on a stack distance of the workload data;
recording the indicator for the workload data in the selected bloom filter based on at least one hash function;
obtaining a hot/cold determination value of the workload data based on the indicator for the workload data recorded in each of the plurality of bloom filters and a weight according to the freshness of each of the plurality of bloom filters; and
and determining whether the workload data is hot data/cold data based on the determination value.

Images