KR20190003091A - Device and method on file system journaling using atomic operation - Google Patents

Abstract

The present invention relates to a device and a method for file system journaling using an atomic operation. More specifically, the device and method: receive a write request and a plurality of pieces of information including additional information for an atomic operation from an application; determine whether an atomic write request exists in the received write request; record the plurality of pieces of information in a write buffer in a storage if the atomic write request exists as a result of determination; and control to commit a mapping table of a flash translation layer (FTL) when the plurality of pieces of information are completely recorded in the write buffer. Accordingly, consistency problems of a file system can be resolved by efficiently using the storage.

Description

Technical Field [0001] The present invention relates to a file system journaling apparatus and method using an atomic operation,

The present invention relates to a journaling apparatus and method, and more particularly, to an apparatus and method for journaling a file system using an atomic operation that supports atomic operations without using a separate journal area to reduce the load of journaling.

A file system is a component of an operating system essential for using a computer, and the operation of such a file system is composed of a plurality of write operations for file and directory creation. Specifically, it modifies file and directory entries, modifies data bitmaps, inode bitmaps and inode blocks, and the data file system is an essential component of an operating system , And the operation of such a file system consists of several write operations for file and directory creation. Specifically, it modifies file and directory entries, modifies data bitmaps, inode bitmaps and inode blocks, and writes data to the data area. The important thing is that all requests for writing should be written, or none should be written, so that consistency is maintained. If the system is unintentionally stopped during these multiple write operations, or if the computer is turned off due to a power outage, some data may be written, and the file system may be in an abnormal state. For example, if only the inode is written and no data is written, the garbage data can be read.

To solve this problem, there is a file system check (fsck) function for checking the consistency of the file system. fsck will check if the existing file system has not been unmounted properly at the time the file system is mounted and will check the metadata of all the inodes and bitmap data in the storage device, Solve the problem. However, due to the increase in the capacity of the storage device, it takes a lot of time to examine all the metadata. Therefore, there is a limit to use in a big data or a cloud environment.

Another technique for maintaining consistency in multiple write requests in the file system is journaling. Journaling techniques are a common technique for improving the consistency of file systems. Journaling performance and lifetime issues are considered as important issues. This journaling technique will be described in detail as follows.

The journaling technique does not perform in-place updates directly to the data storage device for multiple write requests, but rather writes to an area called a journal. When the journal for all writes is completed, commit. Thereafter, a checkpoint is made to update the contents of the journal to the actual location of the storage device. Therefore, even if the system abruptly ends, the system can always perform a consistent state, since undo or redo can be performed depending on whether or not the system is committed. Specifically, undoing will roll back to a consistent state, and redoing will change to a new, consistent state. Also, there is no need for a full scan such as fsck, since only the journal area needs to be examined rather than the entire storage device. This will be described with reference to the drawings.

FIG. 3 is a diagram illustrating a specific structure of a conventional journaling file system, Ext3 and Ext4, and FIG. 4a is a diagram illustrating a conventional journaling process. As shown in FIG. 3 and FIG. 4A, the journal is composed of a journal super block managing journal devices and journal logs. The journal log has a transaction structure, and internally consists of a descriptor, changed data / inode / bitmap, and a commit block.

When a file or directory contents changes, the associated multiple write requests (for example, data write, inode write, bitmap write request) are collected in the transaction and written to the journal area. Then, when the commit is done, checkpoints to the actual storage device. That is, if a commit is recorded in the transaction, it means that all the changed contents have been written, and the checkpointing redo is done. On the other hand, if the commit is not written, it means that the problem occurred while the multiple writes are going on, and the undo is done.

However, existing journaling techniques cause performance degradation because data is written twice to the storage device (once in the journal area and once in the actual storage device during checkpointing). The bigger problem is the increase in the number of writes to the storage device with the life limitation such as the flash memory (flach memory), and the life span greatly decreases. To solve this problem, an order mode journaling technique for only recording metadata in journaling has been proposed, but there is still a problem of write amplification.

Korean Patent Registration No. 1529651 Korean Patent Registration No. 1533042 U.S. Patent No. 5,959,520

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and an object of the present invention is to solve a file system consistency problem that may occur due to a sudden power shutdown and system collision.

It also aims to improve the performance of the system and prevent the degradation of the flash memory lifetime without any transactions, journal descriptor blocks, additional writes and commits, checkpointing, etc. of existing journaling file systems.

It is another object of the present invention to provide an apparatus and a method for efficiently allocating additional storage space and utilizing space without using a journal area in a storage apparatus.

It is also intended to improve the write performance by reducing the number of writes required when a write operation is performed in a storage device.

The objects of the present invention are not limited to the above-mentioned objects, and other objects not mentioned can be clearly understood by those skilled in the art from the following description.

In order to achieve the above object, a file system journaling apparatus using an atomic operation according to the technical idea of the present invention receives a plurality of information including additional information for an atomic operation together with a write request from an application program A determination unit for determining whether or not there is an atomic write request among the received write requests, and a controller for determining whether there is an atomic write request among the received write requests, A write unit that writes data in a write buffer and a commit control unit that controls a mapping table of a flash translation layer to be committed when the plurality of pieces of information are all written in the write buffer, And a control unit.

In this case, the additional information for the atomic operation is information added to a frame information structure (FIS) interface and is a plurality of data including a request type and a logical block addressing (LBA).

And a non-performing unit that does not perform at least one write request corresponding to the serial number when the determined result is a result that there is no atomic write request.

And the commit control unit informs the recording unit of the completion result indicating that the plurality of pieces of information are all written in the write buffer, through an interrupt, when the commit is completed.

According to an aspect of the present invention, there is provided a method for journaling a file system using an atomic operation, the method comprising: receiving a plurality of information including additional information for an atomic operation together with a write request from an application program; A determination step of determining whether the determination unit determines whether there is an atomic write request during the received write request; if the result of the determination is a result of an atomic write request, To a storage internal write buffer, and when the commit control unit records the plurality of pieces of information in the write buffer, a mapping table of a Flash translation layer (FTL) and a commit control step of controlling to commit the table.

In this case, the additional information for the atomic operation is information added to a frame information structure (FIS) interface and is a plurality of data including a request type and a logical block addressing (LBA).

And performing no write request corresponding to the serial number when the determined result is a result that there is no atomic write request.

And the commit control step informs the recording unit of the completion result indicating that the plurality of pieces of information are all recorded in the write buffer, via an interrupt, when the commit is completed.

According to the file system journaling apparatus and method using the atomic operation as described above,

First, there is an effect of solving the file system consistency problem that may occur due to sudden power shutdown and system collision.

Second, it improves the system performance and prevents the degradation of the flash memory lifetime, without transaction, journal descriptor block, additional write and commit, and checkpointing of the existing journaling file system.

Third, there is an effect that an additional storage space can be secured and space utilization can be efficiently performed by using a storage device without using a journal area in a storage device.

Fourth, there is an effect that write performance can be improved by reducing the number of write operations required when a write operation is performed in a storage device.

1 is a block diagram of a file system journaling apparatus using an atomic operation according to an embodiment of the present invention.
2 is a flowchart illustrating a file system journaling method using an atomic operation, according to an embodiment of the present invention.
3 is a diagram showing a specific structure of Ext3 and Ext4 which are conventional journaling file systems.
4A shows a conventional journaling process.
FIG. 4B illustrates a file system journaling process using an atomic operation, according to an embodiment of the present invention. FIG.
5 illustrates a structure of a Flash translation layer (FTL) layer according to an embodiment of the present invention.
6 is a diagram illustrating the operation of an atomic write operation in a storage device, according to one embodiment of the present invention.
7A is a diagram illustrating a conventional Frame Information Structure (FIS) interface;
FIG. 7B illustrates a host / SSD frame information structure (FIS) interface for atomic writing, according to an embodiment of the present invention. FIG.

In order to fully understand the present invention, operational advantages of the present invention, and objects achieved by the practice of the present invention, reference should be made to the accompanying drawings and the accompanying drawings which illustrate preferred embodiments of the present invention. The features and advantages of the present invention will become more apparent from the following detailed description based on the accompanying drawings. Prior to this, terms and words used in this specification and claims are to be interpreted relative to the technical idea of the present invention based on the principle that the inventor can properly define the concept of the term to describe it in the best way of its own invention It should be interpreted in terms of meaning and concept. In the following description, well-known functions and constructions are not described in detail to avoid obscuring the subject matter of the present invention.

FIG. 1 is a block diagram of a file system journaling apparatus using atomic computation according to an embodiment of the present invention. Referring to FIG. As shown in FIG. 1, a file system journaling apparatus using an atomic operation includes a receiving unit 10 for receiving a plurality of pieces of information including additional information for an atomic operation together with a write request from an application program 110 A determination unit 20 for determining whether there is an atomic write request among the received write requests, if the result of the determination indicates that there is an atomic write request, A writing unit 220 for writing a plurality of pieces of information in the writing buffer 210 and a flash translation layer 220 for writing in the writing buffer 210, And a commit control unit 40 for controlling the mapping table of the source node to commit.

At this time, the additional information for the atomic operation is information added to the Frame Information Structure (FIS) interface, and is a plurality of data including a request type and a logical block addressing (LBA).

And an unexecution unit that does not perform at least one write request corresponding to the serial number when the determined result is a result that there is no atomic write request.

The commit control unit 40 informs the recording unit 30 of the completion result indicating that the plurality of pieces of information are all recorded in the write buffer 210 through an interrupt.

FIG. 2 is a block diagram illustrating a file system journaling method using an atomic operation according to an embodiment of the present invention. Referring to FIG. 2, a file system journaling method using an atomic operation is a method in which a receiving unit 10 sends a plurality of pieces of information including additional information for an atomic operation together with a write request from an application program 110 A determination step S20 of determining whether there is an atomic write request among the received write requests, a step of determining whether the result of determination is an atomic write request A recording step S30 of recording the received plurality of pieces of information in an internal write buffer 210 of the storage device 200, And a commit control step S40 for controlling the mapping table to be written in the flash translation layer (FTL) 220 when the write buffer 210 is completely written.

At this time, the additional information for the atomic operation is information added to the Frame Information Structure (FIS) interface, and is a plurality of data including a request type and a logical block addressing (LBA).

The method may further include a step (S35) of not performing at least one write request corresponding to the serial number when the determined result is a result of no atomic write request.

The commit control step S40 informs the recording unit 30 of the completion result indicating that the plurality of pieces of information are all recorded in the write buffer 210 through an interrupt.

The present invention will be described in detail as follows.

The first is that the present invention supports atomic operations. This is a function that guarantees that everything is written or not written for multiple multiple write requests. This is similar to the atomic commands provided by the central processing unit (CPU), for example the Test_and Set or the Exchange command. For example, in the case of atomic Exchange commands, it is guaranteed that read and write operations at two different memory locations will be performed atomically. That is, both memory locations are guaranteed to have their contents changed or not changed. Atomic operations on multiple writes on a storage device ensure that all write requests are processed or none are processed.

The second is that the present invention proposes a new journaling scheme using atomic operations at the storage level. FIGS. 4A and 4B are diagrams comparing a conventional journaling process and a journaling process according to an embodiment of the present invention. 4A, a conventional journaling technique collects a plurality of related write requests (e.g., data write, inode write, and bitmap write requests) into a transaction when a file or directory contents is changed, We write this in the Journal area. Then, when the commit is done, check-pointing to the actual data area. On the other hand, as shown in FIG. 4B, the journaling process according to an embodiment of the present invention collects related multiple write requests into an atomic write request. Therefore, the number of writes can be reduced to one, rather than two, and the life of the storage device can be improved.

The operation of each part will be described in detail as follows.

5 is a diagram illustrating a structure of a flash translation layer (FTL) layer according to an embodiment of the present invention. As shown in FIG. 5, a new interface and firmware inside the storage device are required to provide atomic write operations at the storage device level. In the present invention, solid state drives (SSD) composed of a flash memory are mainly used. There is a firmware called an FTL (Flash translation layer) inside the SSD, which can be software-enabled to support atomic operations on the storage device.

6 is a diagram illustrating an operation of an atomic write operation in a storage device according to an embodiment of the present invention. As shown in FIG. 6, the process of processing atomic writing in the storage device is described as an example. For example, suppose that the data A, B, and C are stored. And the user (TXT) has sent a request to atomically change this data to A ', B', C '. When a host sends a write request to an SSD, it sends a request type, a logical block addressing (LBA), and additional information for atomic operations along with the new data. In FIG. 6, an A ', B', and C 'request is defined as an atomic operation, and a' Write command number / Total write command number ' Additional information exists. For example, A '= 1/3 means that all three are atomic writes, and A' is the first write.

Write requests are written to the write buffer inside the SSD. At this time, the FTL (Flash translation layer) checks whether there is an atomic request during the write request, and if so, it is processed only when the write requests are present in the write buffer. That is, when A ', B', and C 'are all in the write buffer, they are processed. In consideration of the characteristics of the flash memory, the processing of the write request proceeds to an out-place update. That is, it is written to a new location, not to the location of existing data. When all writing is completed, the FTL mapping table is committed. This ensures atomicity, in which everything is written or none written.

More specifically, the FTL checks the write identification number of the blocks received from the write buffer of the storage device and the number of write operations. If the number of operations with the same identification number is measured and equal to the number of required operations of the write command, the write operation is performed to the actual flash memory. On the other hand, if the number of operations with the same identification number is smaller than the number of necessary operations, a problem arises and a write operation is missed. Therefore, the atomicity can not be maintained and the write command of the identification number is canceled. If you have finished writing to the flash memory storage space, modify the mapping table to link the LBA and the physical block address (PBA) in the FTL. Upon completion of the write operation, the storage device notifies the host of the write success through an interrupt.

FIG. 7A is a diagram illustrating a conventional FIS interface, and FIG. 7B is a diagram illustrating a host / SSD frame information structure (FIS) interface for atomic writing according to an embodiment of the present invention. As shown in FIG. 7A, the conventional FIS interface is composed of an instruction, an address (LBA), an FIS type, and the like. On the other hand, as shown in FIG. 7B, information for atomic operation is added to the conventional FIS interface in the FIS interface for the atomic writing host and the SSD AHCI (Advanced Host Controller Interface) protocol of the present invention. At this time, the added information is composed of an atomic operation ID (Atomic ID) and a write command number / total write command number. This additional information can be implemented using FIS command extension fields used by SATA and PCIe.

Thus, writing and modifying files and directories in the file system requires multiple writes, which must be atomically processed to maintain consistency. However, the conventional file system guarantees atomic operation by separately writing and checkpointing in a separate journal area. On the other hand, the journaling technique of the present invention does not require a separate journal area. The journaling technique of the present invention guarantees atomicity by processing multiple writes with atomic operations provided by a storage device.

Specifically, the journaling scheme proposed in the present invention does not use a journal descriptor block, a transaction, or the like used in a conventional file system. Instead, it assigns the same atomic operation ID to multiple write commands and adds serial number / total write number information when sending each write request to the SSD. It then determines whether to redo or not by looking at the response of the atomic write operation.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, It will be appreciated by those skilled in the art that numerous changes and modifications may be made without departing from the invention. It is therefore intended that all such pertinent modifications and variations be considered within the scope of the present invention.

10: Receiving unit 20:
30: recording unit 40:
100: Host 110: Application (Application)
200: Storage 210: Write buffer:
220: Flash translation layer (FTL)

Claims (8)

A receiving unit for receiving a plurality of pieces of information including additional information for an atomic operation together with a write request from an application;
A determining unit for determining whether there is an atomic write request among the received write requests;
A write unit for writing the received plurality of pieces of information in a storage internal write buffer when the determined result is a result of an atomic write request; And
And a commit control unit for controlling a mapping table of a flash translation layer (FTL) to be committed when the plurality of pieces of information are all written in the write buffer. File system journaling device using computation. The method of claim 1, wherein the additional information for the atomic operation comprises:
Wherein the information is added to a Frame Information Structure (FIS) interface and is a plurality of data including a request type and a logical block addressing (LBA). 3. The method of claim 2, wherein if the determined result is a result of no atomic write request,
And not performing at least one write request corresponding to the serial number of the file system. The apparatus of claim 1, wherein the commit control unit comprises:
When the commit is completed, notifies the write unit of the completion result indicating that the plurality of pieces of information are all written in the write buffer, through an interrupt. Receiving a plurality of pieces of information including additional information for an atomic operation together with a write request from an application;
A determination step of determining whether the determination unit determines that there is an atomic write request among the received write requests;
A recording step of recording the received plurality of pieces of information in a storage internal write buffer when the result of the determination is a result of an atomic write request; And
And a commit control step of controlling the commit control unit to commit a mapping table of a flash translation layer (FTL) when the plurality of pieces of information are all recorded in the write buffer A file system journaling method using an atomic operation. 6. The method of claim 5, wherein the additional information for the atomic operation comprises:
Wherein the information is added to a Frame Information Structure (FIS) interface and is a plurality of data including a request type and a logical block addressing (LBA). 7. The method of claim 6, wherein if the determined result is a result of no atomic write request,
Further comprising: a non-performing step of not performing at least one write request corresponding to the serial number by the non-performing unit. 6. The method of claim 5, wherein the commit control step comprises:
And when the commit is completed, notifying the write unit of the completion result indicating that the plurality of pieces of information has been completely written in the write buffer to the write unit through an interrupt.

Images