KR20060087958A - Method of cleaning the flash memory for minimizing overhead of erasing operation - Google Patents

Abstract

The present invention relates to a flash memory management method for minimizing the overhead of an erase operation. The present invention relates to a process of grouping data having similar update frequency when writing data to a flash memory, and unable to write current data when performing an erase operation. Information about the number of data blocks in the state and the number of times erased so far is used to find the Erase unit that is most suitable for erasing. In addition, the number of EUs currently in the free state is used to find the most suitable time for erasing and to perform erasing. As a result, the lifetime of the flash memory can be extended to the maximum, and the performance of the flash file system can be improved by minimizing the cost and the number of erase operations.

Flash memory, EU, erase, minimize overhead

Description

Method of cleaning the flash memory for minimizing overhead of erasing operation}

1 is an overview of a flash memory managed in EU units;

2 is a flow diagram of an embodiment of clustering data in accordance with the present invention;

3 is an overview for logically separating a plurality of regions;

4 illustrates an embodiment of clustering data according to an update frequency;

5 is a flowchart of an embodiment of moving data in accordance with the present invention;

6 is an overview of a process of moving data for each region;

7 is an example of a data management structure for each area;

8 is a flow diagram of one embodiment for performing an erase in accordance with the present invention;

9 is a flowchart of an embodiment of determining an erasure time point according to the present invention;

10 is a graph for explaining a determination process of erasing time.

* Explanation of symbols for main parts of the drawings

10: flash memory 11-1: EU header

11-2: Block Header 11-3: Data Block

71: EU listing 72: area table

73: Free EU Table

The present invention relates to a flash memory management method for minimizing the overhead of an erase operation, and more particularly, to manage a flash memory having a slow erase time and a limited number of times to erase data in consideration of wear leveling. It's about how to do it.

Recently, various embedded systems such as PDAs, HPCs, set-top boxes, and information appliances have been developed. Real-time operating systems and file systems are required to operate such systems. In this case, flash memory is mainly used as a storage medium for storing data due to the characteristics of an embedded system requiring easy portability, fast access time, and low power.

Flash memory has fast read operations but relatively slow write and erase operations, and has the following problems.

First, unlike general storage media, flash memory cannot be rewritten to an area where data has been written once. In other words, before writing new data, the flash memory must be erased, and unlike other storage media, the erase operation must be erased by a certain size erase unit (EU). This EU, also called a segment, will be used as a generic meaning of the erase unit used to manage flash memory regardless of the term.

Second, flash memory is limited in the number of times it can be erased. In other words, if you continue to use a certain EU, you will reach the end of its useful life and will no longer be usable. This results in a reduction in the size of the flash memory and the erasing process occurs more frequently to write new data, which is fatal to the use of the system.

Accordingly, the present invention has been made to solve the above problems, to minimize the overhead of the erase operation of the flash memory and to evenly distribute the EU in which the erase operation is performed to make the wear leveling efficiently, using a flash memory It is an object of the present invention to provide a method for improving the performance of various devices.

In order to achieve the above object, an embodiment of performing clustering of data according to the present invention includes: logically dividing a flash memory into a plurality of regions each having at least one EU; And processing data to be recorded in the area in which data having an update frequency similar to the update frequency of the corresponding data is recorded in the flash memory.

Each logically divided area may be set to record data having a different update frequency than that of other areas.

At this time, when data belonging to a certain area is updated, the updated data may be recorded in an adjacent area set to store data having a higher update frequency than the current area to which the corresponding data belongs.

When erasing is performed for an EU belonging to a certain area, the data of the EU may be recorded in an adjacent area set to store data having a lower update frequency than the current area to which the EU belongs. In this case, the valid data may be configured to be recorded in another EU belonging to the current area without moving to another area when the time spent in the current area is less than or equal to a predetermined value.

The flash memory may further include a free EU that does not belong to the logically divided region.

At this time, if the erasing is performed for any EU belonging to each of the logically divided areas, the deleted EU may be configured to belong to the free EU.

In addition, when there is not enough EU space in the area to record data, one of the free EUs may be included as an EU belonging to the corresponding area, and data may be recorded in the EU.

According to an embodiment of the present invention, an optimal erasing operation may include: verifying, for each EU of a flash memory, information about the number of data blocks in a state in which the current data cannot be written (Invalid state) and the number of times of erasing so far; ; Finding the EU most suitable for erasing by calculating the number of data blocks in the Invalid state and information on the number of times of erasing using a predetermined equation; And performing erasing on the found EU.                         

The operation may be configured to use the ratio of the data block in the Invalid state to the maximum number of data blocks constituting each EU, and the ratio of the current erase count to the maximum possible erase count.

In addition, the calculation may be configured to reflect a certain weight in order to be able to set which part of the number of data blocks in the invalid state and the number of times erased so far is important.

According to an embodiment of the present invention, an embodiment of determining when to perform erasing includes: setting one or more reference values for the number of free EU (EU) states in which data can be recorded at present; And deleting the predetermined number of EUs according to a result of comparing the current number of Free EUs with the reference value.

At this time, the reference value is N1, N2 (N1 < N2), and if the current number of Free EU is greater than or equal to N2, no erase operation is performed. If the current number of Free EU is greater than or equal to N1 and less than N2, (N2-N1). ) Can be configured to perform erase for EU. In addition, when the number of free EUs is less than N1, erasing may be prioritized to secure EU space for recording data.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Referring to FIG. 1, the flash memory 10 may be divided into a plurality of EUs that are units of erase. Each EU may be composed of a plurality of data blocks. In general, an EU header 11-1 and a block header 11-2 are recorded in the first data block. The EU header 11-1 includes information such as erase count, time information, EU status, EU identification information (ID), etc. The block header 11-2 includes a data block status, a file identifier, and a block identifier. It includes information such as data length, time information, and update count. Each other data block 11-3 is a place for storing actual data.

On the other hand, when erasing is performed in the EU, all data blocks of the EU become 'Free' state, and when valid data is stored, the corresponding data block becomes 'Valid' state. When updating the data in 'Valid' state, it is impossible to save the data in the existing saved location directly due to the characteristics of flash memory. Therefore, save it in another free state data block and change the previous data block to 'Invalid' state. In addition, the erased data block immediately becomes 'Invalid' state. When deleting the EU, the invalid data block existing in the EU is erased as it is, and the data block in the 'Valid' state is deleted after moving to another EU.

In this case, it is very important to reduce the number of times of erasing the EU since the write operation is basically slower than the read operation and there is no available space, so performing the erase operation is seen as a write operation externally. Therefore, when erasing the EU, it is desirable to select the largest number of invalid data blocks.

An embodiment of grouping data according to the present invention will be described with reference to FIG. 2. First, a flash memory is divided into a plurality of regions logically divided (S21). Here, each 'region' may include one or more EUs as in the example shown in FIG. 3, and each region may be set to record data having a different range of update frequencies.

When data is updated or erased and data is recorded in the flash memory, the data is processed to be recorded in areas in which data having an update frequency similar to that of the corresponding data is recorded in each area (S22 and S23).

An example of a process of performing step S23 will be described with reference to FIG. 4. Each area may be set as an area for recording data having a higher update frequency from the lowest area (area # 1) to the highest area (area #n). have. That is, data with the lowest update frequency is recorded in the lowest region (area # 1), and data with the highest update frequency is recorded in the highest region (area #n).

Then, the data belonging to each area are processed to be recorded by moving to the upper area or the lower area according to the update frequency, so that the data having similar update frequency are clustered. For example, when data is updated, the data is updated more frequently than other data that has not been updated, so the data is moved and recorded in the upper region of the current region, and data that stays in one region for a long time is updated compared to other data. The frequency can be regarded as low, so it is recorded by moving to the lower area of the current area.

However, since there is no further subregion in the lowest region, recording is made between its own EU, and since there is no further upper region in the highest region, recording is made between its own EU.

An embodiment of performing step S23 shown in FIG. 2 will be described in more detail with reference to FIG. 5.

When the data is updated (S51), the updated data is recorded in the EU of the adjacent area set to store data having a higher update frequency than the area to which the data currently belongs (S52).

That is, if any data is updated, the data cannot be written directly to the data block because of the characteristics of the flash memory. Therefore, the data must be written to another data block. For this reason, the data can be regarded as having a higher update frequency than other data that is not updated. Referring to FIG. 6A, assuming that data of "HCHCCCHHHCCHCCHH" is recorded in any EU of the current area 61, and data indicated by "H" is updated, the data indicated by "H" is regarded as having a high update frequency. The data can be stored in the EU of the next area 62, which stores data of a higher update frequency.

On the other hand, if the erasing is performed for a certain EU (S53), since valid data of the valid state recorded in the EU must be preserved, the EU must be deleted after transferring the valid data to another EU. In this case, in order to more accurately perform clustering of data, the valid state data recorded in the EU to be erased may be reflected in the current area.

That is, the latest update time of the data of the valid state recorded in the EU to be erased is checked (S54), and it is determined whether a predetermined reference time elapses from the latest update time (S55).

When it is determined in step S55 that the reference time has not elapsed since the last update time, the process is transferred to another EU belonging to the current area 63 as shown in FIG. 6B (S56). In FIG. 6B, 'V' display data means data in a valid state. However, as a result of the determination in step S55, when more than a reference time has elapsed since the last update time, the probability of the update being low is high. Therefore, the update frequency is lower than that of the region 64 to which the current one belongs, as shown in FIG. 5C. The data is recorded in the EU of the adjacent area 65 that stores data (S57).

That is, the degree of variability of the data may be determined according to the update frequency of the data, but it may appropriately reflect the situation in which the variability decreases as the time the data stays in the current area increases.

As described above, the number of times of erasing can be reduced by separating data that frequently occurs from data that does not occur frequently and clustering data having similar update frequencies. In addition, since data that is frequently updated is highly likely to change again, the EU, which has most of such data, has the most invalid data block. As a result, the erase operation on the EU can reduce the process of moving the data block in the valid state to another EU, thereby freeing up a lot of available space and improving performance.

Meanwhile, the flash memory may be configured to include an EU in a state in which data can be recorded without belonging to any region other than the EU belonging to logically divided regions as in the above-described embodiments. Hereinafter, such an EU will be referred to as a free EU.

Referring to FIG. 7, the flash memory includes logically divided regions (regions # 1 to #n) and free EU 75 which are not included in any regions.

The EU list 71 is a list that holds information for identifying the EU assigned to each area. The area table 72 contains information about each EU, such as information about the EU assigned to each area, and Active EU information indicating the EU (ie, EU having a free data block) to which the current data block can be allocated in the area. Table that holds. At this time, all data block allocation is made in the active EU 74. In addition, the Free EU table 73 is a table for holding information about the Free EU.

In the embodiment shown in FIG. 7, when erasing is performed for a certain EU belonging to each of the logically divided regions, the EU may be configured to belong to the free EU without being in the corresponding region anymore. That is, the Free EU list 73 manages the EU on which the erasing is performed as the Free EU.

In addition, if there is not enough space to record data in the EU of the current region, one of the free EUs may be included as an EU belonging to the corresponding region, and the data may be configured to be recorded in the EU. That is, when each active EU does not have enough space for data storage, one EU is allocated from the free EU and the assigned EU is set as the active EU.

An embodiment of effectively selecting an EU to be an object of an erase operation according to the present invention will be described in detail with reference to FIG. 8.

First, when the erasing is executed, information about the number of data blocks in the Invalid state and the number of times of erasing is confirmed for each EU of the flash memory (S81).

The data block number of the invalid state and the current erase count information identified in step S81 are calculated according to a predetermined formula to find the EU most suitable for erasing (S82), and then the erase is performed on the found EU (S83).

At this time, the information of the number of data blocks in the invalid state and the information of the number of times of erasing are different from each other. And use the ratio of the current erase count to the maximum possible erase count.

An embodiment of finding an EU most suitable for erasing in step S82 will now be described in detail.

When the EU is deleted, the data block in the valid state existing in the EU must be moved to another EU and deleted. Therefore, it is preferable to select and delete the EU with the largest data block in the invalid state. In other words, the lowest erase operation cost is selected and erased.

In this case, the erase operation cost of the EU may be expressed as a cost of moving a data block in a valid state from another EU selected as an erase operation to another EU and a cost of deleting the EU, as shown in Equation 1 below.

Cost (TE) = Cost (VRM) + Cost (ER)

Here, Cost (TE) is a cost of erasing operation, Cost (VRM) is a cost of moving a valid data block from another EU to another EU, and Cost (ER) is a cost of erasing the selected EU. At this time, the data block in the valid state, the data block in the invalid state, the data block in the free state, and the size of the EU are called V, I, F, and E, respectively, and read, write, and erase operations are performed in the flash memory. When the time required is r, w, and e, respectively, Equation 1 may be expressed as Equation 2 below.

Cost (TE) = V (r + w) + E * e

At this time, r is a very small value compared to w, and since w and e have similar values, it can be expressed as in Equation 3 below.

Cost (TE) = (V + E + F) * w

Since E and w are values determined according to the characteristics of the flash memory, the erase operation cost can be expressed by Equation 4 below.

Cost (TE) = (2E-I) * w

In other words, in order to select the EU with the lowest erase operation cost, the EU with the most invalid data block must be selected.

On the other hand, in order to perform an efficient erase, in addition to the erase operation cost, the erase count that determines the lifetime of the flash memory must be considered. That is, it is preferable to select the erasing target EU so that the wear leveling is performed over the entire area of the flash memory. A method of selecting an erase operation target EU in consideration of the number of erase operations will be described.

The probability P (k) at which the k-th EU is selected for the erase operation is inversely proportional to the cost of the erase operation and the current erase count. When the erase count reaches the maximum eraseable count, P (k) should be reduced.

If the maximum number of erasable times for an EU is E _{maximum erasable number} and the current erasing number is E _{current erasing frequency} , the remaining erasing number may be expressed as Equation 5 below.

E _{number of remaining erases} = E _{maximum number of erase possible} -E _{number of current erases}

Then, since P (k) is inversely proportional to the erase operation cost and is proportional to the remaining erase counts, when P (k) is expressed as in Equation 6 or Equation 7 below, the EU having a small P (k) value is expressed as the EU to be erased. Can be selected. In this case, a certain weight α may be applied to determine whether to give a greater weight to the erase operation cost or a larger weight to the erase count.

At this time, the unit of the erase operation cost and the number of erase times are different from each other. One way to solve this problem is to use a ratio with respect to the maximum value for each. The maximum value for the erase operation cost is the cost of erasing the EU when all data blocks in the EU are valid. Each ratio may be expressed as Equation 8 and Equation 9 below.

Using Equations 8 and 9, P (k) may be expressed as Equations 10 and 11 below.

Now, by determining the weight α, it is possible to determine the target EU for the erase operation.

At this time, if the current erase count approaches the maximum value, the P (k) value must be increased to exclude the erase operation. That is, if the number of erases remaining for the maximum value approaches zero, the P (k) value should increase. Therefore, if α is inversely proportional to the number of erases to the maximum value, it can be finally expressed as in Equation 12 below.

An embodiment of determining a start time of an erase operation according to the present invention will now be described. When using flash memory, once data has been written to the EU, it can only be written back after the EU is erased. Therefore, when and which EU is erased throughout flash memory has a significant impact on performance.

Referring to FIG. 9, one embodiment of determining an erasing time point according to the present invention will be described. First, a reference value for the number of free EUs that are maintained in a state capable of recording current data is set (S91). Hereinafter, an example in which N1 and N2 (N1 <N2) are set as reference values in step S91 will be described. In this case, N1 may have a meaning of a minimum value necessary for stable operation, and N2 may have a meaning of a maximum value.

If the number of free EUs is greater than or equal to N2, the free EU to record data may be regarded as sufficiently secured. Therefore, the erase operation that may affect performance may not be executed (S92 and S93).

However, if the number of Free EUs is greater than or equal to N1 and less than N2, the Free EUs are secured by performing the erasure on a certain number of EUs, although there is a Free EU to record data, but not enough (S94, S95). ). At this time, it is preferable to process the number of EUs to be erased by N2-N1.

In addition, if the current number of Free EU is less than N1, the available memory may be regarded as inadequate, and thus, the erase operation is first performed with respect to other operations (S96). If the number of Free EUs is less than N1 even after the erasing is performed in step S96, an arbitrary operation for notifying a shortage of available memory, such as outputting a warning message, may be performed (S97, S98).

Referring to FIG. 10, when N2 or more Free EUs are secured, an erase operation does not occur. When the number of Free EUs is between N2 and N1, N2-N1 EUs are used to secure N2 or more Free EUs. Clears. If the number of free EUs is less than or equal to N1, the erase operation is performed first and the number of free EUs after N1 is N1. If it is below, it can be judged that there is no more available space.

The present invention is not limited to the above-described embodiments and can be variously modified and implemented by those skilled in the art without departing from the technical spirit of the present invention.

When the embodiments of the present invention are used in a file system using a flash memory, the following effects can be obtained.

1) By selecting the EU for the erase operation in consideration of the wear leveling, the life of the flash memory can be extended as much as possible.

2) The performance of flash file system can be improved by minimizing the cost and number of erase operations, which is the biggest overhead when using flash memory as a storage medium.

3) The present invention can be applied to the minimum modification of the flash file system using the conventional simple erasing method, thereby increasing the performance of the flash file system.

Claims (14)

Logically dividing the flash memory into a plurality of regions each having at least one erase unit (EU); And A method of managing flash memory for minimizing the overhead of an erasing operation, including processing data to be written in the flash memory so that data having an update frequency similar to the update frequency of the data can be recorded in an area in which the data is recorded. . The method of claim 1, And each of the logically divided regions is configured to record data having an update frequency in a range different from that of other regions. The method of claim 2, When data belonging to a certain area is updated, the flash memory management method for minimizing the overhead of an erase operation is configured to write to an adjacent area set to store data having a higher update frequency than the current area to which the data belongs. . The method of claim 2, When erasing is performed for an EU belonging to an area, the erase operation over is configured to write valid data of the corresponding EU in an adjacent area set to store data having a lower update frequency than the current area to which the EU belongs. How to manage flash memory to minimize heads. The method of claim 4, wherein And writing the valid data to another EU belonging to the current area when the valid data stays in the current area is less than or equal to a predetermined value. The method according to any one of claims 1 to 5, The flash memory management method for minimizing the overhead of an erase operation, characterized in that it further comprises a free EU (EU) which does not belong to the logically divided area. The method of claim 6, And erasing is performed on any EU belonging to each of the logically divided regions, and the erased EU is configured to belong to a free EU. The method of claim 6, When there is not enough EU space in the area to record data, one of the free EUs is included as an EU belonging to the corresponding area, and flash memory management for minimizing the overhead of the erase operation, characterized in that it is configured to write data to the EU. Way. For each EU of the flash memory, checking information on the number of data blocks in a state where the current data cannot be written (Invalid state) and the number of times of erasing so far; Finding the EU most suitable for erasing by calculating the number of data blocks in the Invalid state and information on the number of times of erasing using a predetermined equation; And And performing an erase operation on the found EU. The method of claim 9, The operation may be performed by using a ratio of data blocks in an invalid state to a maximum number of data blocks constituting each EU, and a ratio of a current erase count to a maximum possible erase count. How to manage flash memory for minimization. The method of claim 10, The operation of the flash memory for minimizing the overhead of the erase operation, characterized in that configured to reflect a certain weight. Setting one or more reference values for the number of free EUs in which data can be recorded; And And erasing a predetermined number of EUs according to a result of comparing the current number of Free EUs with the reference value. The method of claim 12, The reference value is composed of N1 and N2 (N1 <N2). If the current number of free EUs is greater than or equal to N2, no erase operation is performed. If the current number of free EUs is greater than or equal to N1 and less than N2, (N2-N1) Flash memory management method for minimizing the overhead of erase operations, characterized in that it is configured to perform erase for the EU. The method of claim 13, If the number of the current free EU is less than the N1, the flash memory management method for minimizing the overhead of the erase operation, characterized in that configured to attempt to perform the highest priority for the (E2-number of the current free EU) as EU. .

Images