KR20200092019A - Data storage system of on-volatile memory apparatus and method of the same - Google Patents

Abstract

The present invention relates to a data management system of a nonvolatile memory device capable of significantly reducing the write performance and WAF by reducing unnecessary copybacks due to frozen pages where write operations do not occur any more, and a method thereof. According to an embodiment of the present invention, the data management system of the nonvolatile memory device comprises: a nonvolatile memory including a plurality of memory blocks composed of a plurality of pages; a host that monitors data pages to determine whether the data pages are frozen pages where update operations have not occurred, and when a page on which a write request is to be performed is a frozen page, sends the write request including a flag indicating a frozen state to the nonvolatile memory; and a controller storing the page into a frozen block, when the flag indicating the frozen state is included in the write request received from the host.

Description

DATA STORAGE SYSTEM OF ON-VOLATILE MEMORY APPARATUS AND METHOD OF THE SAME

The present invention relates to a data management system and method of a nonvolatile memory device, and more particularly, to reduce unnecessary copyback due to a frozen page in which a write operation no longer occurs in a nonvolatile memory device, thereby reducing write performance. And a data management system and method for a nonvolatile memory device that can significantly reduce WAF.

Recently, the paradigm of the computer environment has been shifted to ubiquitous coumpting, which enables computer systems to be used anytime, anywhere. Due to this, the use of portable electronic devices such as mobile phones, digital cameras, and notebook computers is rapidly increasing. Such a portable electronic device generally uses a memory device, and the memory device is used as a main storage device or an auxiliary storage device of the portable electronic device.

The memory device has the advantages of excellent stability and durability because there is no mechanical driving unit, and an information access speed is very fast and power consumption is low. Data storage devices having these advantages include Universal Serial Bus (USB), memory cards with various interfaces, and solid state drives (SSDs).

As large files such as music and videos are played in portable electronic devices, a memory device is also required to have a large storage capacity. In order to secure a large storage capacity, a memory device uses a memory device having a high density of memory cells, for example, a flash memory device that is one of nonvolatile memory devices.

Flash memory devices do not support data overwrite due to structural features. That is, it is impossible to update the data of the programmed state page (the basic unit in which data is stored and used in the same sense as'memory cell'). Therefore, in order to program data in the flash memory, an erase operation must be preceded. This is called an erase-before-program operation. That is, the page in the programmed state of the flash memory device must be returned to the initial state or the erased state before data is programmed.

On the other hand, in the online transaction processing (OLTP) database and the like, even though there are quite a lot of freeze pages in which no more write operations actually occur, there is no concept of a freeze page or a technique for separately processing the freeze page in a flash memory device. In the case of the actual TPC-C benchmark, there are over 80% of frozen pages.

When a random writes, such as OLTP, drives a frequent workload on a flash memory device (eg, SSD), the freeze page is unnecessarily copied back in the garbage-collection stage, causing excessive write amplification factor (WAF). As a result, there is a problem that severely adversely affects the performance and life of the storage device.

Accordingly, there is a need to develop a technology capable of dramatically reducing write performance and WAF in an OLTP workload by drastically reducing unnecessary copyback due to frozen pages.

Prior art related to this is Korean Patent Registration No. 10-1829625 (Invention name: a method of simultaneously transmitting data and power wirelessly and a transmitting device and a receiving device using the same).

An object of the present invention is to provide a data management system and method for a non-volatile memory device that can significantly reduce write performance and WAF by reducing unnecessary copyback due to a frozen page in which no more write operations occur. will be.

The problem to be solved by the present invention is not limited to the problem(s) mentioned above, and another problem(s) not mentioned will be clearly understood by those skilled in the art from the following description.

The data management system of a nonvolatile memory device according to an embodiment of the present invention monitors a nonvolatile memory and a data page including a plurality of memory blocks composed of a plurality of pages, and a frozen page in which an update operation has not occurred. If it is determined whether or not the page to be written to the nonvolatile memory is a frozen page, a host transmitting a write request including a flag indicating a frozen state, and a frozen state in a write request received from the host When the flag is included, a controller that stores the corresponding page in a frozen block is included.

Preferably, the non-volatile memory includes a normal block storing data in which an update operation occurs, and a frozen block storing a frozen page in which an update operation has not occurred. can do.

Preferably, the host may determine whether to freeze the page based on the schema or transaction logic of the database engine.

Preferably, the controller determines whether a free page exists in a freeze block of the non-volatile memory when the write request includes a flag indicating a freeze state, and if the free page exists, freezes the frozen page. In the free page, if a free page does not exist, garbage-collection is performed on blocks excluding the freeze block, a free block is generated after selecting a victim block, and the free block is allocated as a freeze block and freezes. You can save the page.

Preferably, the controller writes to the frozen block for a false frozen page that is incorrectly judged to be a frozen page, and then writes the false frozen page to the frozen block after rewriting. The records may be invalidated, and if necessary, garbage collection may be performed on a frozen block according to the garbage collection policy.

In a data management method of a nonvolatile memory device according to another embodiment of the present invention, the host determines whether a page to be written to the nonvolatile memory is a frozen page, and requests writing to the controller, wherein the controller responds to the write request. Determining whether a flag indicating a frozen state exists, the controller stores a corresponding page in a frozen block of a non-volatile memory when a flag indicating a frozen state exists in the write request, and a flag indicating a frozen state And if it does not exist, storing the page in a general block of the non-volatile memory.

Preferably, when a flag indicating a frozen state exists in the write request, the step of storing the page in a frozen block of the nonvolatile memory may include determining whether a free page exists in the frozen block of the nonvolatile memory, If a free page exists, the frozen page is stored in a frozen block, and if a free page does not exist, garbage-collection is performed on blocks excluding the frozen block, and a free block is generated after selecting a victim block And, it may include the step of allocating the generated free block as a frozen block and storing a frozen page.

Preferably, when there is no flag indicating a frozen state in the write request, the page may be stored in a general block of the nonvolatile memory.

According to the present invention, when the host knows that a write operation no longer occurs for a specific page Pi, it leaves the “F” (frozen) state of the page and writes (Pi, By requesting'F'), the corresponding page is processed separately in the FTL of the flash memory device, thereby reducing unnecessary copyback due to a frozen page in which no further write operation occurs, and dramatically reducing write performance and WAF. Can be reduced to Accordingly, there is an effect of dramatically increasing the performance and life of the flash memory device.

Meanwhile, the effects of the present invention are not limited to the above-mentioned effects, and various effects may be included within a range obvious to those skilled in the art from the following description.

1 is a view for explaining a data management system of a nonvolatile memory device according to an embodiment of the present invention.
2 is a view for explaining a write operation method when it is not a frozen page according to an embodiment of the present invention.
3 is a diagram for describing a write operation in the case of a frozen page according to an embodiment of the present invention.
4 is a diagram for explaining garbage-collection in the case of a frozen page according to an embodiment of the present invention.
5 is a view for explaining a data management method of a nonvolatile memory according to an embodiment of the present invention.
6 is a flowchart illustrating a method of storing a frozen page in a frozen block according to an embodiment of the present invention.
7 is a flowchart illustrating a method of allocating a property of a free block when garbage collection occurs according to an embodiment of the present invention.

The present invention can be applied to various changes and can have various embodiments, and specific embodiments will be illustrated in the drawings and described in detail. However, this is not intended to limit the present invention to a specific embodiment, it should be understood to include all modifications, equivalents, or substitutes included in the spirit and scope of the present invention. In describing each drawing, similar reference numerals are used for similar components.

Terms such as first, second, A, and B may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from other components. For example, the first component may be referred to as a second component without departing from the scope of the present invention, and similarly, the second component may be referred to as a first component. The term and/or includes a combination of a plurality of related described items or any one of a plurality of related described items.

When an element is said to be "connected" or "connected" to another component, it is understood that other components may be directly connected to or connected to the other component, but there may be other components in between. It should be. On the other hand, when a component is said to be "directly connected" or "directly connected" to another component, it should be understood that no other component exists in the middle.

The terms used in this application are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions, unless the context clearly indicates otherwise. In this application, terms such as “include” or “have” are intended to indicate that a feature, number, step, operation, component, part, or combination thereof described in the specification exists, and that one or more other features are present. It should be understood that the existence or addition possibilities of fields or numbers, steps, operations, components, parts or combinations thereof are not excluded in advance.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person skilled in the art to which the present invention pertains. Terms, such as those defined in a commonly used dictionary, should be interpreted as having meanings consistent with meanings in the context of related technologies, and should not be interpreted as ideal or excessively formal meanings unless explicitly defined in the present application. Does not.

1 is a view for explaining a data management system of a non-volatile memory device according to an embodiment of the present invention, FIG. 2 is a view for explaining a write operation method when it is not a frozen page according to an embodiment of the present invention, 3 is a view for explaining a write operation in the case of a frozen page according to an embodiment of the present invention, and FIG. 4 is a view for explaining garbage-collection in the case of a frozen page according to an embodiment of the present invention.

Referring to FIG. 1, a data management system of a nonvolatile memory device according to an embodiment of the present invention includes a host 100, a controller 200, and a nonvolatile memory 300. In addition, the data management system of the nonvolatile memory device may further include other components such as an interface unit in addition to the above components.

The host 100 monitors a data page to determine whether a frozen page has not been updated, and if a page to be written to the nonvolatile memory 300 is a frozen page, it means a frozen state. Send a write request containing a flag. At this time, the host 100 may determine the freezeability (whether or not a page is frozen) of the page through information possessed by the host 100. That is, the host 100 may determine whether the page is frozen based on the schema or transaction logic of the database engine. For example, the host 100 may determine whether the page is frozen based on the number of records in the page. In this case, the host 100 may determine that the page is a frozen page when the number of records inside the page of the table matches the maximum number of records or the number of updates.

The freeze page occurs when a write operation occurs in the nonvolatile memory 300 as data is added to a certain data page, and a specific condition is satisfied (for example, there is no space to add more data to the corresponding page, and there is no space for adding data). If deletion/change does not occur in the future), it may mean a page in which no further write operation occurs.

As such, the determination of whether to freeze the page is made in the application of the host 100, not inside the SSD. When writing the page determined to be the frozen page, the host 100 transmits a “frozen page” hint (a flag indicating a frozen state) to be recognized by the FTL 220 in the controller 200.

The host 100 may be an application, and the application may include a DB engine, an operating system (OS), and a program operating as a foundation of the operating system.

Hereinafter, when the host 100 is a Database engine, a description will be given of a method for the Database engine to determine a frozen page. It is assumed that the Database engine understands the specific table schema characteristics and the semantics of transaction programs that add records and update records to the table, so that it is possible to know when a particular page of the table is "frozen." For example, according to the logic of the Order_Line table, New_Order, and Delievery transaction of TPC-C, which is a representative database benchmark, after all records in a specific page belonging to the Order_Line table are updated, the page is no longer written. Since it is certain, it can be judged to be frozen. As an example of the host 100, in the case of the Databse engine, more specifically, since the page header stores information about the number of records in the page, the Database engine can detect the table when the page is evicted from the buffer. If the number of records in the page matches the maximum number of records or the number of updates, it can be determined as a frozen page. The DB engine determines the freezing point in the manner described above, and transmits the frozen hint (a flag indicating a frozen state) to the controller 200, and the controller 200 receives and processes the frozen hint.

In conclusion, the host 100 leaves the “F” (frozen) state of the page at a time when it knows that a write operation does not occur for a specific page Pi, and when it requests to write to the nonvolatile memory 300 You can request it with "Write(Pi,'F')". In “Write(Pi,'F'),'F' may be a flag indicating a frozen state.

The nonvolatile memory 300 is composed of a plurality of memory blocks, and each memory block may be composed of a plurality of pages. At this time, the page represents the minimum unit in which data is stored, and the block represents the unit in which the set number of pages are collected. Depending on the matching relationship between the logical address of data and the physical address of memory, data may be stored in units of pages or blocks, and the number of each page and the amount of data stored in each page may be set by the controller 200. One memory block may be composed of 4, 16, 128, and 256 pages. Also, depending on the type of use, a block in which data is stored or to store data may be divided into a normal block 310 and a frozen block 320. That is, the non-volatile memory 300 may include a normal block 310 storing data in which an update operation occurs, and a freezing block 320 storing a frozen page in which an update operation has not occurred. . In the present invention, the block storing the page on which the update operation occurs is limited to the normal block 310, but may include a cold block or the like.

The controller 200 is a configuration that controls the nonvolatile memory 300. When a request for writing an input file is received from the host 100, the controller 200 controls the nonvolatile memory 300 to record the input file in a desired block and page. .

When the write request received from the host 100 includes a flag indicating a frozen state, the controller 200 stores the corresponding page in the frozen block 320. At this time, the controller 200 checks whether there are more free pages than the write requested capacity in the freeze block 320, and if there are free pages, stores the pages requested to be written in the freeze block 320. do. If a free page does not exist in the freeze block 320, the controller 200 performs garbage-collection on the remaining blocks 310 except for the freeze block 320, thereby performing a sacrificial block. By selecting, a new free block is generated, and the generated free block is allocated as a frozen block, and the page requested to be written is stored in the allocated frozen block 320. Here, the free block means a block in which no pages are stored and no attributes are designated. Garbage-collection refers to scanning a memory block to determine the memory block with the smallest valid page count as a victim block, copying a valid page included in the victim block, and deleting the data of the victim block.

On the other hand, even if the host 100 determines the frozen page incorrectly and delivers a hint about the frozen page, the controller 200 stores a false frozen page in the frozen block 320. However, since the False Frozen Page is not a frozen page, rewriting may occur after being written to the freezing block 320. In this case, the controller 200 invalidates the previous record after writing the false freeze page to the freeze block 320, and if the freeze block 320 is also invalidated according to the garbage collection policy, the garbage collection (garbage) collection(GC)). As such, the frozen page is stored in the freezing block 320 even in the case of rewriting. Then, the frozen page may be incorrectly determined and updated. At this time, when performing garbage-collection, the controller 200 must process frozen blocks and frozen pages so as not to mix with general blocks and general pages in order to prevent mixing of frozen pages and general pages.

The controller 200 includes a host interface 210, a flash translation layer (FTL) 220, and a memory interface 230.

The host interface 210 may be an interface for communication between the host 100 and the controller 200.

The flash translation layer 220 may be a module that manages the nonvolatile memory 300. The non-volatile memory 300 has an advantage of providing a fast read speed at a relatively low unit price. However, in order to write data to the nonvolatile memory 300, an erase (erase) operation is preceded, and a unit (block) of data that is erased is larger than a unit (page) of data to be written. Due to this feature of the non-volatile memory 300, it is difficult to use an existing file system for a hard disk as it is, and the flash conversion layer 220 has a compatible operation between the file system and the non-volatile memory 300. It can be enabled.

The flash translation layer 220 performs garbage-collection according to the garbage-collection policy when there is no empty space in the input file write request of the host 100 and thus garbage-collection needs to be performed. For example, when the page to be written is a frozen page, garbage collection is performed on the frozen block 320 including the most invalid pages among the frozen blocks 320. For another example, when the page to be written is a normal page, garbage collection is performed on the normal block 310 including the most invalid pages among the normal blocks 310.

In order to distinguish between the normal block 310 and the frozen block 320, the flash translation layer 220 divides the block pool through flags of attributes of the block. Blocks with the same flag are considered to be a block pool with the same stream.

In the initial state, all of them are set to the normal block 310, and writing is performed, and as many frozen blocks 320 as necessary are generated according to the amount of frozen hints. When free space is insufficient, garbage-collection occurs. At this time, an existing block 310 or a frozen block 320 may be generated depending on which pool the required space is.

When the frozen block space is insufficient in the garbage-collection process, a free block is generated and allocated as Frozen, and in the case of a garbage-collection caused by insufficient existing block space, the existing Block 310 is allocated.

Which pool the required space depends on the nature of the write causing the garbage-collection. That is, if the hint of the garbage-collection-induced write is frozen, the frozen block is required, and if there is no frozen hint in the garbage-collection-induced write, it is determined that the space of the existing block is required.

After dividing the block into two pools as described above, the flash conversion layer 220 writes to the free space of the frozen block 320 when the hint of the write request is frozen and writes the corresponding address in the mapping table. do. If the hint of the write request is not frozen, the flash translation layer 220 writes to the free space of the existing block and writes the corresponding address in the mapping table.

The flash conversion layer 220 determines the type of the page when the host 100 writes data. At this time, the flash conversion layer 220 may determine that the page is a frozen page when a hint indicating a frozen state exists in the flag of the page requested to be written.

The flash translation layer 220 maps the logical address generated by the host 100 to the virtual address of the normal block 310 when the page type is not a frozen page, and stores the page in the mapped virtual address. .

For example, when a write request of “Write (Pi, NULL)” is received, as shown in FIG. 2, since the write request does not include the'F' flag indicating a frozen state, the flash conversion layer 220 may open the corresponding page. It is stored in the general block 310.

Further, the flash translation layer 220 maps the logical address generated by the host 100 to the virtual address of the freezing block 320 when the page type is a frozen page, and stores the frozen page in the mapped virtual address. do.

For example, when a write request of "Write (Pi,'F')" is received, as shown in FIG. 3, since the write request includes a'F' flag indicating a frozen state, the flash conversion layer 220 includes the corresponding page. Is determined as a frozen page, and is stored in the frozen block 320. At this time, the flash conversion layer 220 determines whether a free page exists in the freeze block 320 and, if a free page exists in the freeze block 320, stores the freeze page in the freeze block 320 as shown in FIG. .

If a free page does not exist in the freeze block 320, the flash conversion layer 220 performs garbage-collection on blocks other than the freeze block 320 as shown in FIG. 4 to perform a new free block (A). The free block A is allocated as a freeze block, and a freeze page is stored.

The memory interface 230 may be an interface for communication between the memory controller 200 and the nonvolatile memory 300.

In the system configured as described above, the host (DB engine, operating system, or general application) 100 “F” (frozen) of the page at the time when it knows that no more write operation occurs for a specific page Pi. When a request is made to write to the flash memory device, the state is left and the corresponding page is separately processed by the FTL 220 of the flash memory device by requesting Write (Pi,'F'). This has the effect of increasing the performance and life of the storage device.

5 is a diagram for describing a data management method of a nonvolatile memory according to an embodiment of the present invention.

Referring to FIG. 5, the host determines whether a page to be written to the nonvolatile memory is a frozen page (S510), and makes a write request to the memory controller (S520). At this time, when the host requests to write a frozen page, a flag indicating a frozen state may be included in the write request and transmitted. For example, when the flag indicating the frozen state is'F', the host may transmit Write (Pi,'F') when requesting to write a specific page Pi. Also, when a host requests to write a page other than a frozen page, "Write (Pi, NULL)" may be delivered.

When step S520 is performed, the controller determines whether a flag indicating a frozen state exists in the write request (S530).

As a result of the determination in step S530, if a flag indicating a frozen state exists in the write request, the controller stores the page in the frozen block of the nonvolatile memory (S540). For a detailed description of a method of storing a frozen page in a frozen block, refer to FIG. 6.

If, as a result of the determination in step S530, a flag indicating a frozen state does not exist in the write request, the controller stores the page in a general block of non-volatile memory (S550).

6 is a flowchart illustrating a method of storing a frozen page in a frozen block according to an embodiment of the present invention.

Referring to FIG. 6, when a write request for a frozen page is received (S610), the controller determines whether a free page exists in the frozen block of the nonvolatile memory (S620). At this time, the controller can also determine whether the free page is larger than the size of a frozen page.

As a result of the determination in step S620, if a free page exists in the frozen block, the controller stores the frozen page in the frozen block (S630).

If, as a result of the determination in step S620, there is no free page in the frozen block, the controller performs garbage-collection on the remaining blocks other than the frozen block (S640), selects a victim block and generates a free block ( S650).

Then, the controller allocates the generated free block as a frozen block (S660), and stores the frozen page in the frozen block (S670).

7 is a flowchart illustrating a method of allocating a property of a free block when garbage collection occurs according to an embodiment of the present invention.

Referring to FIG. 7, when a garbage-collection occurs (S710), the controller selects a sacrificial block to generate a free block (S720).

Then, the controller determines whether the hint of the write request that caused the garbage-collection is frozen (S730).

As a result of the determination in step S730, if the hint of the write request is frozen, the controller allocates a free block as a frozen block (S740).

If, as a result of the determination in step S730, the hint of the write request is not frozen, the controller allocates a free block as a basic block (S750).

So far, the present invention has been focused on the preferred embodiments. Those skilled in the art to which the present invention pertains will understand that the present invention can be implemented in a modified form without departing from the essential characteristics of the present invention. Therefore, the disclosed embodiments should be considered in terms of explanation, not limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the equivalent range should be interpreted as being included in the present invention.

100: host
200: controller
210: host interface
220: flash conversion layer
230: memory interface
300: non-volatile memory
310: normal block
320: freeze block

Claims (8)

A non-volatile memory including a plurality of memory blocks composed of a plurality of pages;
The data page is monitored to determine whether a frozen page has not occurred, and if a page to be written to the nonvolatile memory is a frozen page, a write request including a flag indicating a frozen state is issued. The sending host;
When a write request received from the host includes a flag indicating a frozen state, a controller that stores the corresponding page in a frozen block
Including, data management system of a non-volatile memory device. According to claim 1,
The nonvolatile memory may include a normal block storing data in which an update operation occurs, and a frozen block storing a frozen page in which an update operation has not occurred. Non-volatile memory device data management system. According to claim 1,
The host,
Data management system of a non-volatile memory device, characterized in that to determine whether the page is frozen based on the schema or transaction logic of the database engine. According to claim 1,
The controller,
When a flag indicating a frozen state is included in the write request, it is determined whether a free page exists in the frozen block of the nonvolatile memory, and if a free page exists, the frozen page is stored in the frozen block, and the free page exists If not, performing a garbage-collection on blocks other than the frozen block, selecting a victim block, generating a free block, and allocating the generated free block as a frozen block to store a frozen page Data management system for non-volatile memory devices. According to claim 1,
The controller,
For a false frozen page that is incorrectly judged as a frozen page, after writing to the frozen block and rewriting, when the false frozen page is written to the frozen block, the previous record is invalidated, A data management system of a nonvolatile memory device characterized in that garbage collection is performed on a frozen block according to a garbage collection policy if necessary. Determining whether a page to be written to a nonvolatile memory by a host is a frozen page and requesting writing to the controller;
Determining whether a flag indicating a frozen state exists in the write request;
The controller stores the page in a freeze block of a nonvolatile memory when a flag indicating a frozen state exists in the write request, and when a flag indicating a freeze state does not exist, the controller stores the page in a general block of the nonvolatile memory. Steps to save the page
Data management method of a non-volatile memory device comprising a. The method of claim 6,
When a flag indicating a frozen state exists in the write request, the step of storing the page in the frozen block of the nonvolatile memory may include:
Determining whether a free page exists in a frozen block of the nonvolatile memory;
If a free page exists, the frozen page is stored in a frozen block, and if the free page does not exist, garbage-collection is performed on blocks excluding the frozen block, and a free block is generated after selecting a victim block. And storing the free pages by allocating the generated free blocks as freeze blocks. The method of claim 6,
When there is no flag indicating a frozen state in the write request, storing the page in a general block of the non-volatile memory, the data management method of the non-volatile memory device.

Images