KR102837532B1 - Method to solve the problem that file operation is blocked because of journal conflict - Google Patents

Abstract

A transaction method is disclosed, including a first step of creating a shadow page of the first group by copying pages of a first group among a plurality of transaction pages prepared by a running transaction, and a second step of transferring pages of a second group among the plurality of transaction pages and the shadow pages of the first group to a storage.

Description

METHOD TO SOLVE THE PROBLEM THAT FILE OPERATION IS BLOCKED BECAUSE OF JOURNAL CONFLICT

The present invention relates to computing technology, and more particularly, to technology for executing a transaction to ensure consistency of a file system.

Most conventional file systems, including EXT4 and XFS, use a mechanism called journaling to ensure file system consistency. Instead of reflecting file system changes immediately, they are recorded in a journal area, which is an area separate from the file system. When recording changes in the journal area, they are recorded in the form of a transaction, which is defined as a set of multiple write operations that are executed atomically. The changed contents are collected into blocks, which are the minimum units of writing and reading in storage devices, to form a transaction. The contents of a block are cached in volatile memory and are called pages. A transaction can be defined as a set of block writes or a set of page writes. The operation of saving a transaction in the journal area is called a commit. If a transaction is successfully committed to the journal area, the changes contained in the transaction are reflected to the original location in the file system. If a file system crashes, the changes contained in the transactions committed to the journal area are reflected to the original location in the file system.

A general file operation may attempt to change the content that is in the process of being committed to the journal area. This phenomenon is defined in the present invention as a journal conflict. When the journal conflict occurs, the file operation called by the user must wait until the writing of the block that caused the conflict is completed.

A page that caches the content that caused a journal conflict is defined as a conflict page. A conflict page cannot be modified by a normal file operation. A conflict page is a page that is being committed. Its content is being transferred to the journal area in the storage device. The content of the page when the commit started must be maintained to ensure the atomicity of the transaction. If a conflict page is modified by a normal file operation, the atomicity of the normal file operation may be broken. This is because only the content of the conflict page, which is part of the metadata modified by the normal file operation, can be committed by the transaction being committed.

The prior art makes a copy of the conflict page when committing a transaction. The page participating in the commit is not the conflict page, but a copy of the conflict page. Since the copy of the conflict page participates in the commit, normal file operations can freely modify the original page.

In the prior art, the entity that creates a copy of a conflict page is a general file operation. That is, the general file operation creates a copy of the conflict page. And in the prior art, the time at which the copy of the conflict page is created is immediately before the general file operation modifies the page. And in the prior art, the criterion for determining whether a copy of the conflict page is created is when the page to be modified is being committed by a commit operation and the page is immediately before being transferred to storage. The commit operation checks whether there is a copy created by the general file operation, and if there is a copy, it causes the copy to participate in the storage transfer. In this case, the prior art has a problem in that the general file operation that caused the journal conflict is suspended until the transaction commit is completed.

The present invention seeks to provide a technology that always maintains a copy of a block to be recorded in a journal area in order to resolve the phenomenon of file operation interruption due to journal collision. The present invention seeks to provide a technology that eliminates the phenomenon of general file operations being interrupted due to journal collision.

According to one aspect of the present invention, a method for resolving a phenomenon in which a file operation is interrupted due to a journal conflict can be provided. The paging method provided in this specification according to one aspect of the present invention can be referred to as Semantics-aware Shadow Paging. In this method, a page where conflicts frequently occur is determined, and a copy of a page where conflicts frequently occur is always made to participate in a transaction commit.

Figure 1 illustrates the working principle of a Semantics-aware Shadow Paging method provided according to one aspect of the present invention.

The above Semantics-aware Shadow Paging method divides metadata into Hot metadata and Warm metadata.

Hot metadata (110) is metadata that frequently causes collisions, and metadata that is modified by most file operations, such as super blocks and block bitmaps, belongs to hot metadata (110). When committing hot metadata (110), always make a copy.

Warm metadata (120) is metadata in which collisions do not occur frequently, and metadata that do not belong to hot metadata (110), such as inodes and directory blocks, belong to warm metadata (120). Among the warm metadata (120) blocks, copies are made only of blocks in which collisions occur relatively more frequently.

Warm metadata (120) may be metadata other than the Hot metadata (110), and among the Warm metadata (120), metadata in which collisions occur relatively more frequently may be referred to as Warm[H] metadata (121), and metadata in which collisions occur relatively less frequently may be referred to as Warm[L] metadata (122). The value of the journal collision frequency that distinguishes the Hot metadata (110) and the Warm[H] metadata (121) from each other may be defined as a first threshold, and the value of the journal collision frequency that distinguishes the Warm[H] metadata (121) and the Warm[L] metadata (122) from each other may be referred to as a second threshold. The second threshold may be less than or equal to the first threshold.

Warm metadata (120) blocks where collisions occur more frequently are selected through the LRU (Least Recently Used) algorithm (701).

Here, 'collisions occur frequently' may mean that the frequency with which collisions occur is greater than a given reference frequency, and 'collisions do not occur frequently' may mean that the frequency with which collisions occur is not greater than the given reference frequency.

For example, if the number of collisions for a specific type of metadata is greater than a predetermined value N1 while file operations for the specific type of metadata occur N times, the specific type of metadata can be considered as the Hot metadata (100) (N>N1). However, the definition of the frequency can be made in various ways, and the present invention is not limited to a specific method of defining the frequency.

FIG. 2 shows a commit process of a journal transaction to which a Semantics-aware Shadow Paging method is applied according to one aspect of the present invention.

The file system (12) first changes the status of the transaction to committing (S202) after starting the commit (S201). This is to prevent other file operations from accessing the transaction.

The above file system (12) prepares for data transmission of the transaction once conversion to a committing transaction is completed.

The data transmission of a transaction consists of three stages: the journal descriptor preparation stage (S210), the data to be transmitted preparation stage (S200; S221~S226), and the transmission stage (S230).

In the journal descriptor preparation step (S210), a journal descriptor block is created and information about data included in the transaction, i.e., the original block address of the block to be journaled, is stored.

In the data preparation step (S220) to be transmitted, the pages to be involved in the DMA transmission are set while traversing the blocks included in the transaction (S223). In the present invention, a step (S224) of determining whether a block is Hot metadata or Warm metadata, a step (S225) of determining whether it is Warm[H] metadata or Warm[L] metadata by referring to the collision frequency, etc., may be performed, and a step (S226) of generating a shadow page may be performed.

In the transmission step (S230), a storage device command is transmitted so that the contents of the page prepared in the data preparation step to be transmitted are transmitted to storage.

These steps can be executed by the file system.

In a method for resolving a file operation interruption phenomenon due to a journal collision provided according to one aspect of the present invention, the subject that creates a copy of a collision page is a commit operation. That is, the commit operation creates a copy of the collision page. And at this time, the point in time when the copy of the collision page is created is right before transmission starts during the commit operation. And at this time, the criterion for determining whether a copy is created is determined by an algorithm proposed and disclosed in this specification. That is, it can be determined using an algorithm that determines whether a page to be determined is a page of Hot metadata, a page of Warm[H] metadata, or a page of Warm[L] metadata. In the present invention, the terms 'Warm[H] metadata' and 'Warm[L] metadata' may be replaced with other words not described in this specification.

According to one aspect of the present invention, a transaction method provided may include a first step in which a file system of a host copies pages of a first group among a plurality of transaction pages prepared by a running transaction to create shadow pages of the first group; and a second step in which the file system transfers pages of a second group among the plurality of transaction pages and shadow pages of the first group to a storage.

At this time, the pages of the first group may be composed of all pages among the plurality of transaction pages that require copying.

At this time, the transmission may be characterized in that it starts after the first step is completed.

At this time, the transaction method may further include, before the first step, a step in which the file system prepares the plurality of transaction pages by executing the running transaction; a step in which the file system initiates a commit; and a step in which the file system changes the state of the transaction to a commit state. At this time, after the second step, the file system may further include a step in which the file system terminates the commit.

At this time, the pages of the second group may be stored in the main buffer included in the host, and the shadow pages of the first group may be stored in the shadow buffer included in the host.

At this time, the plurality of transaction pages are stored in the main buffer included in the host, and the pages of the second group may be composed of the remaining pages excluding the pages of the first group among the plurality of transaction pages.

At this time, the pages of the first group may include pages of metadata in which a journal collision frequency is greater than a predetermined first threshold value, and the pages of the second group may include pages of metadata in which a journal collision frequency is less than a predetermined first threshold value.

At this time, the first step may include: a step in which the file system generates a journal descriptor; a step in which the file system determines, for each transaction page, whether the transaction page is a page of metadata in which a journal collision occurs at a frequency greater than or equal to the first threshold; and a step in which the file system copies the transaction page only if the transaction page is a page of metadata in which a journal collision occurs at a frequency greater than or equal to the first threshold.

Alternatively, the pages of the first group include pages of metadata having a journal collision frequency greater than or equal to a first threshold value, the pages of the second group include pages of metadata having a journal collision frequency less than or equal to a second threshold value, and the pages of the first group further include pages of metadata having a journal collision frequency less than or equal to the first threshold value and greater than or equal to the second threshold value, wherein the second threshold value may be less than the first threshold value.

At this time, the first step may include: a step in which the file system generates a journal descriptor; a step in which the file system determines, for each transaction page, whether the transaction page is a page of metadata in which a journal collision occurs at a frequency greater than or equal to the second threshold; and a step in which the file system copies the transaction page only if the transaction page is a page of metadata in which a journal collision occurs at a frequency greater than or equal to the second threshold.

According to one aspect of the present invention, a host including a processing unit; and a communication unit; may be provided. The processing unit is configured to execute a file system of the host. The file system is configured to execute a first step of copying pages of a first group among a plurality of transaction pages prepared by a running transaction to create a first group of shadow pages; and a second step of transmitting pages of a second group among the plurality of transaction pages and the first group of shadow pages to a storage.

At this time, the transmission starts after the first step is completed.

At this time, the file system may be further configured to perform, prior to the first step, a step in which the file system prepares the plurality of transaction pages by executing the running transaction; a step in which the file system initiates a commit; and a step in which the file system changes the state of the transaction to a commit state. And the file system may be further configured to perform, after the second step, a step in which the file system terminates the commit.

According to another aspect of the present invention, a computing device including a host and storage may be provided. The host is configured to execute a file system. And the file system is configured to execute a first step of copying pages of a first group among a plurality of transaction pages prepared by a running transaction to create a first group of shadow pages; and a second step of transferring pages of a second group among the plurality of transaction pages and the first group of shadow pages to a storage.

At this time, the pages of the second group may be stored in the main buffer included in the host, and the shadow pages of the first group may be stored in the shadow buffer included in the host.

The paging method provided according to one aspect of the present invention can be applied to all software using journaling, including file systems using journaling mechanisms such as EXT4 and XFS, and database systems using journaling mechanisms such as SQLite and MySQL, and also can be applied to all commercial products equipped with software using journaling, including embedded devices such as data center servers and smartphones, and all services using the same.

The present invention is applicable to all software using journaling. Since journaling is the most basic technology that maintains the consistency of the file structure used by software, journaling is an essential technology for file management. Software using journaling ranges from small applications installed in embedded devices to systems installed in large data centers. The present invention can improve the quality of various services from embedded devices to servers by improving journaling performance.

According to the present invention, in order to solve the phenomenon of file operation interruption due to journal collision, a technology can be provided that always maintains a copy of a block to be recorded in a journal area. In addition, a technology can be provided that eliminates the phenomenon of general file operations being interrupted due to journal collision.

Figure 1 illustrates the working principle of a Semantics-aware Shadow Paging method provided according to one aspect of the present invention.
FIG. 2 shows a commit process of a journal transaction to which a Semantics-aware Shadow Paging method is applied according to one aspect of the present invention.
Figure 3 illustrates a concept in which a host writes information to storage according to one embodiment.
Figures 4a to 4e illustrate execution criteria for transactions allowed by a file system in one embodiment.
Figure 5 shows an example of a process in which two different transactions occur and are terminated.
FIG. 6 is a flowchart illustrating a transaction execution method provided according to one embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described with reference to the attached drawings. However, the present invention is not limited to the embodiments described in this specification and may be implemented in various other forms. The terminology used in this specification is intended to help understanding of the embodiments and is not intended to limit the scope of the present invention. In addition, the singular forms used below also include the plural forms unless the phrases clearly indicate the opposite meaning.

Figure 3 illustrates a concept in which a host writes information to storage according to one embodiment.

The host (10) and the storage (20) may each be computing devices that operate by supplying power from a power supply. The host (10) and the storage (20) may exchange data and commands through one or more transmission channels (30). The transmission channels (30) may be wireless channels or wired channels. The host (10) and the storage (20) may share power provided from one power supply, or may each receive power from two different power supplies.

The above host (10) may include a CPU, memory, power supply, and communication device.

The above storage (20) may include a controller (21), volatile memory (22), and non-volatile memory (23).

The above host (10) can transmit various commands and data to the storage (20) through the above transmission channels (30). The commands may include a write command.

The controller (21) of the storage (20) can store data received from the transmission channels (30) in the volatile memory (22) based on the command received from the transmission channels (30). The data stored in the volatile memory (22) can be stored in the nonvolatile memory (23) according to the rule followed by the controller (21). The data stored in the volatile memory (22) can be deleted when the power supplied to the storage (20) is cut off, but the data stored in the nonvolatile memory (23) is not deleted even when the power supplied to the storage (20) is cut off.

The above host (10) can execute an application (11) and a file system (12). The above application (11) and the above file system (12) may be executed by executing predetermined command codes stored in a memory accessed by the host (10) by a CPU included in the host (10).

In one embodiment, the application (11) may be a program that is executed or terminated by a user providing user input through a user interface provided by the host (10).

In one embodiment, the file system (12) may be a program that is automatically executed by the host (10) when power is applied to the host (10) or a reset is performed.

The above application (11) can send various system calls to the file system (12). The file system (12) can execute a task corresponding to the system call.

The host (10) can execute a transaction. It may take a certain amount of time from the time a particular transaction starts until it ends. At a particular point in time, the host (10) can perform one transaction.

The application (11) can control the start and commit of a transaction. And the application (11) can control one or more operations to be executed during the transaction. As shown in Fig. 3, the application (11) can transmit system calls including a start call, a set of operation calls, and a commit call to the file system (12).

A specific transaction is started by the start call, commands to be transmitted from the host (10) to the storage (20) are prepared by the set of operation calls, and the prepared commands can be transmitted to the storage (20) through the transmission channels (30) by the commit call.

Meanwhile, the start of a specific transaction running in the file system (12) may not require a transaction start call from the application (11). The start of a transaction covered in one embodiment of the present invention may not require a transaction start call from the application (11).

Figures 4a to 4e illustrate execution criteria for transactions allowed by a file system in one embodiment.

The above file system (12) can execute a running transaction (71) and a commit transaction (72).

When one running transaction (71) ends, another running transaction (71) can be executed. Also, when one commit transaction (72) ends, another commit transaction (72) can be executed.

However, another running transaction (71) cannot be executed before one running transaction (71) ends. Also, another commit transaction (72) cannot be executed before one commit transaction (72) ends.

However, it is possible for another commit transaction (72) to be executed before one running transaction (71) ends, and another running transaction (71) can be executed before one commit transaction (72) ends.

That is, as shown in Fig. 4a, it is allowed for two different running transactions (71) to be executed with a time difference. Similarly, as shown in Fig. 4b, it is allowed for two different commit transactions (72) to be executed with a time difference.

However, as shown in Fig. 4c, it is not allowed for two different running transactions (71) to overlap on the time axis. Similarly, it is not allowed for two different commit transactions (72) to overlap on the time axis.

However, as shown in Fig. 4e, it is allowed for one running transaction (71) and one commit transaction (72) to overlap on the time axis.

Figure 5 shows an example of a process in which two different transactions occur and are terminated.

In Figure 5, the time on the right is the time that occurred first, and the time on the left is the time that occurred later.

The file system (12) can execute the first running transaction in the time period t11 to t12 and the first commit transaction in the time period t13 to t14. As a result, the first transaction can be completed.

The file system (12) can execute the second running transaction in the time period t21 to t22 and execute the second commit transaction in the time period t23 to t24. As a result, the second transaction can be completed.

At this time, the file system (12) allows the second running transaction to be executed while the first commit transaction is being executed.

FIG. 6 is a flowchart illustrating a transaction execution method provided according to one embodiment of the present invention.

In FIG. 6, a box with reference number 51 represents a main buffer (51) that stores pages to be processed in the running transaction, a box with reference number 52 represents a shadow buffer (52) that copies and stores pages to be shadowed among the pages stored in the main buffer (51), and a box with reference number 53 represents a pointer buffer (53) that stores pointers indicating the locations of pages to be transferred in a commit transaction corresponding to the running transaction.

The above main buffer (51), the above shadow buffer (52), and the above pointer buffer (53) may be referred to as a first set of buffers, a second set of buffers, and a third set of buffers, respectively.

The square of reference number 54 represents a page buffer (54) in which a page of the Hot metadata is stored among the main buffer (51), the circle of reference number 55 represents a page buffer (55) in which a page of the Warm[H] metadata (121) is stored among the main buffer (51), and the diamond of reference number 56 represents a page buffer (56) in which a page of the Warm[L] metadata (122) is stored among the main buffer (51).

The above page buffer (54), page buffer (55), and page buffer (56) may be expressed as a hot page buffer (54), a warm [H] page buffer (55), and a warm [L] page buffer (56), respectively, or may be expressed as a first page buffer (54), a second page buffer (55), and a third page buffer (56).

The hatched rectangle of reference number 57 represents a shadow page buffer (57) in which a hot shadow page created by duplicating a page of hot metadata stored in the main buffer (51) is stored. The shadow page buffer (57) may belong to the shadow buffer (52).

The hatched circle of reference number 58 represents a shadow page buffer (58) in which a Warm shadow page created by duplicating a page of the Warm[H] metadata (121) stored in the main buffer (51) is stored. The shadow page buffer (58) may belong to the shadow buffer (52).

The above shadow page buffer (57) and the above shadow page buffer (57) may be expressed as a hot shadow page buffer (57) and a warm shadow page buffer (58) or a first shadow page buffer (57) and a second shadow page buffer (58), respectively.

In Figure 6, the vertical line extending from the left side is the time axis, where the upper part indicates the time that occurred first and the lower part indicates the time that occurred later.

Step (S100) is a step in which the first running transaction is executed. While the first running transaction is executed, pages to be processed in the first running transaction are stored in the main buffer (51).

In the example of FIG. 6, while the first running transaction is being executed, the page (P ₀ ) stored in the Hot page buffer (54) is the Hot metadata (110), the page (P 1 ) stored in the Warm[H] page buffer ( ₅₅ ) is the Warm metadata (120) in which collisions frequently occur, and the page (P ₂ ) stored in the Warm[L] page buffer (56) is the Warm metadata (120) in which collisions do not frequently occur.

Step (S200) represents a step in which a first commit transaction corresponding to the first running transaction is executed. Step (S200) illustrated in Fig. 6 corresponds to step (S200) illustrated in Fig. 2. In Fig. 6, for convenience, steps (S201), (202), and (S240) of Fig. 2 are omitted from the illustration.

In step (S210), the file system (12) prepares a journal descriptor.

In step (S220) of Fig. 2, the initial value of variable i is set. In step (S220) of Fig. 6, the value of variable i is also initialized. In the example of Fig. 6, since there are a total of three pages stored in the main buffer (51) in the first running transaction, variable i is initialized to 3.

In step (S220), when i=2, the page (P ₂ )(=P _i=2 ) corresponding to i=2 is a page of the Warm[L] metadata (122) where collisions do not occur frequently. In one embodiment of the present invention, the page of the Warm[L] metadata (122) is not replicated. Therefore, the page (P ₂ )(=P _i=2 ) is not replicated in the shadow buffer (52). At this time, a pointer (ptr2) pointing to the location of the original page (P ₂ ) stored in the main buffer (51) is stored in the pointer buffer (53).

The value of i can decrease with each step.

In step (S220), when i=1, the page (P ₁ ) (=P _i=1 ) corresponding to i=1 is a page of the Warm[H] metadata (121) where collisions frequently occur. In one embodiment of the present invention, a Warm shadow page that is a copy of the original page of the Warm[H] metadata (121) is created. The copied Warm shadow page copies the page stored in the Warm[H] page buffer (55) and stores it in the Warm shadow page buffer (58). As a result, in step (S220), the original of the page of the Warm[H] metadata (121) is maintained in the state of being stored in the main buffer (51), and the copy of the page of the Warm[H] metadata (121) is stored in the shadow buffer (52). At this time, a pointer (ptr1) indicating the location of the Warm shadow page (P ₁ ), which is a copy of the page of the Warm[H] metadata (121), is stored in the pointer buffer (53).

In step (S220), when i=0, the page (P ₀ ) (=P _i=0 ) corresponding to i=0 is the page of the Hot metadata (100). In one embodiment of the present invention, a Hot shadow page that is a copy of the page of the Hot metadata (100) is created. The copied Hot shadow page is stored in the Hot shadow page buffer (57) of the shadow buffer (52). As a result, in step (S220), the original of the page of the Hot metadata (100) is maintained in the state of being stored in the main buffer (51), and the copy of the page of the Hot metadata (100) is stored in the shadow buffer (52). At this time, a pointer (ptr0) indicating the position of the Hot shadow page (P ₀ ), which is a copy of the page of the Hot metadata (100), is stored in the pointer buffer (53).

In step (S220), preparation of data to be transmitted can be completed by completing the main buffer (51), the shadow buffer (52), and the pointer buffer (53).

In step (S230), a storage device command is transmitted so that the contents of the page prepared in step (S220) are transmitted to the storage. Specifically, in one embodiment, the file system (12) can read pages stored at locations pointed to by the pointers of the pointer buffer (53), and the file system (12) can transmit a command to the storage (20) to write the read pages to the storage (20).

In the example presented in Fig. 6, the original page (P ₂ ) stored in the Warm[L] page buffer (56) pointed to by the pointers of the pointer buffer (53), the shadow page (P ₀ ) stored in the Hot shadow page buffer (57), and the shadow page (P ₁ ) stored in the Warm shadow page buffer (58) can be transmitted in step (S230).

Even if, during step (S230), another second running process changes the original page (P 0 ) stored in the Hot page buffer (54) of the main buffer ( ₅₁ ) or the original page (P 1 ) stored in the Warm [H] page buffer (55) of the main buffer ( ₅₁ ), it is guaranteed that the atomicity of the data transmitted to the storage (20) in step (S230) is not damaged.

One embodiment of the present invention is characterized in that, before transmission of a set of pages for which atomicity must be guaranteed begins, replication of pages satisfying a predetermined condition among the set of pages is completed. In addition, for the replicated pages, the replicated pages are transmitted instead of the original pages.

A transaction method provided according to one embodiment of the present invention may include a first step in which a file system of a host copies pages of a first group among a plurality of transaction pages prepared by a running transaction to create shadow pages of the first group; and a second step in which the file system transfers pages of a second group among the plurality of transaction pages and shadow pages of the first group to a storage.

Here, the pages of the first group may include pages of the Hot metadata (110). And the pages of the second group may include the Warm[L] metadata (122).

Additionally, the pages of the first group may further include pages of the Warm[H] metadata (121).

At this time, the pages of the first group may be composed of all pages among the plurality of transaction pages that require copying. Here, all pages that require copying include pages of the Hot metadata (110). In addition, all pages that require copying may further include pages of the Warm [H] metadata (121).

At this time, the transaction method may further include, prior to the first step, a step in which the file system prepares the plurality of transaction pages by executing the running transaction. Here, the prepared plurality of transaction pages may be stored in the main buffer (51).

At this time, the pages of the first group may include pages of metadata having a journal collision frequency greater than or equal to a first threshold value, and the pages of the second group may include pages of metadata having a journal collision frequency less than or equal to the first threshold value. At this time, the first step may include: a step in which, for each of the transaction pages, the file system determines whether the transaction page is a page of metadata having a journal collision frequency greater than or equal to the first threshold value; and a step in which, for each of the transaction pages, the file system copies the transaction page only when the transaction page is a page of metadata having a journal collision frequency greater than or equal to the first threshold value. Here, the pages of the first group are pages of the Hot metadata (110), and the pages of the second group are pages of the Warm[H] metadata (121) and pages of the Warm[L] metadata (122).

At this time, the pages of the first group include pages of metadata (=Hot metadata) having a journal collision frequency greater than or equal to a first threshold value; the pages of the second group include pages of metadata (=Warm[L] metadata) having a journal collision frequency less than or equal to a second threshold value; and the pages of the first group may further include pages of metadata (=Warm[H] metadata) having a journal collision frequency less than or equal to the first threshold value and greater than or equal to the second threshold value. At this time, the second threshold value is less than the first threshold value. At this time, the first step may include, for each transaction page, a step in which the file system determines whether the transaction page is a page of metadata having a journal collision frequency greater than or equal to the second threshold value; and a step in which the file system copies the transaction page only if the transaction page is a page of metadata having a journal collision frequency greater than or equal to the second threshold value.

By using the embodiments of the present invention described above, those who belong to the technical field of the present invention will be able to easily perform various changes and modifications within the scope that does not depart from the essential characteristics of the present invention. The content of each claim of the patent claims can be combined with other claims that do not have a citation relationship within the scope that can be understood through this specification.

[Sasa]

This invention was created with the support of the following research project supported by the Government of the Republic of Korea.

1. Research project 1

*Ministry Name: Ministry of Science and ICT

*Task ID Number: 20180005490031001

*Research Project Name: SW Computing Industry Core Technology Development Project

*Research Project Name: Development of an Expandable Order-Guaranteed Operating System for Manycore Large-Capacity Memory

*Organization: Korea Advanced Institute of Science and Technology

*Research management specialist organization: Information and Communications Technology Planning and Evaluation Institute

*Research period: 2018.04.01.~2022.12.31

2. Research project 2

*Ministry Name: Ministry of Science and ICT

*Task ID: 2020R1A2C300852511

*Research Project Name: Basic Research Project in Science and Engineering

*Research Project Name: Research on a Non-Competitive Scalable I/O Subsystem for Ultra-Low-Latency Storage Devices (2020)

*Research Management Specialist Organization: National Research Foundation of Korea

Claims (15)

A first step in which the host's file system creates a first group of shadow pages by copying a first group of pages consisting of all pages that need to be copied among a plurality of transaction pages prepared by a running transaction; and
A second step in which the file system transfers pages of the second group among the plurality of transaction pages and shadow pages of the first group to storage;
Including,
Transaction method. A transaction method according to claim 1, characterized in that the transmission is initiated after the first step is completed. In the first paragraph,
Before the above step 1,
A step in which the above file system prepares the plurality of transaction pages by executing the running transaction;
The above file system, the step of starting a commit; and
The above file system changes the state of a transaction to a commit state;
Including more,
After the above second step,
The step where the above file system terminates the commit;
Including more,
Transaction method. In the first paragraph,
The pages of the second group are stored in the main buffer included in the host,
The shadow pages of the first group are stored in a shadow buffer included in the host.
Transaction method. In the first paragraph,
The pages of the first group above include pages of metadata for which the frequency of journal collision occurrence is greater than a predetermined first threshold value,
The pages of the second group above include pages of metadata for which the frequency of journal collision occurrence is less than a predetermined first threshold,
Transaction method. In paragraph 5,
The above first step is,
The above file system creates a journal descriptor;
For each of the above transaction pages, the file system determines whether the transaction page is a page of metadata whose frequency of journal conflict occurrence is greater than or equal to the first threshold; and
For each of the above transaction pages, the file system copies the transaction page only if the transaction page is a page of metadata whose frequency of journal conflict occurrence is greater than or equal to the first threshold;
Including,
Transaction method. In the first paragraph,
The pages of the first group above include pages of metadata for which the frequency of journal collision occurrence is greater than a predetermined first threshold value,
The pages of the second group above include pages of metadata for which the frequency of journal collision occurrence is less than a predetermined second threshold,
The pages of the first group further include pages of metadata in which the frequency of journal collision occurrence is less than the first threshold and greater than or equal to the second threshold,
The above second threshold is smaller than the above first threshold.
Transaction method. A first step in which the host's file system creates a first group of shadow pages by copying pages of a first group among a plurality of transaction pages prepared by a running transaction; and
A second step in which the file system transfers pages of the second group among the plurality of transaction pages and shadow pages of the first group to storage;
Including,
The above multiple transaction pages are stored in the main buffer included in the host,
The pages of the second group are composed of the remaining pages among the plurality of transaction pages, excluding the pages of the first group.
Transaction method. delete A first step in which the host's file system creates a first group of shadow pages by copying pages of a first group among a plurality of transaction pages prepared by a running transaction; and
A second step in which the file system transfers pages of the second group among the plurality of transaction pages and shadow pages of the first group to storage;
Including,
The pages of the second group above include pages of metadata for which the frequency of journal collision occurrence is less than a predetermined second threshold,
The above first step is,
The above file system comprises a step of creating a journal descriptor; and
For each of the above transaction pages, the file system copies the transaction page only if the transaction page is a page of metadata whose journal conflict frequency is greater than or equal to the second threshold;
Including,
Transaction method. As a host,
Processing unit; and communication unit; including;
The above processing unit is configured to execute the file system of the host,
The above file system,
A first step of creating a first group of shadow pages by copying pages of a first group consisting of all pages that need to be copied among a plurality of transaction pages prepared by a running transaction; and
A second step in which the file system transfers pages of the second group among the plurality of transaction pages and shadow pages of the first group to storage;
characterized by being configured to execute,
Host. A host, characterized in that in claim 11, the transmission is started after the first step is completed. In Article 11,
The above file system,
Before the above step 1,
A step in which the above file system prepares the plurality of transaction pages by executing the running transaction;
The above file system, the step of starting a commit; and
The above file system changes the state of a transaction to a commit state;
is designed to run more,
After the above second step,
The step where the above file system terminates the commit;
is supposed to run more,
Host. A computing device comprising a host and storage,
The above host is configured to run a file system,
The above file system,
A first step of creating a first group of shadow pages by copying pages of a first group consisting of all pages that need to be copied among a plurality of transaction pages prepared by a running transaction; and
A second step in which the file system transfers pages of the second group among the plurality of transaction pages and shadow pages of the first group to storage;
characterized by being configured to execute,
Computing device. In Article 14,
The pages of the second group are stored in the main buffer included in the host,
The shadow pages of the first group are stored in a shadow buffer included in the host.
Computing device.