JPH08263380A - Disk cache control system - Google Patents

Abstract

PURPOSE: To improve the hit rate of a disk cache by determining the replace priority of a cache block based on access attribute management information. CONSTITUTION: A cache control part 5 reports access information to an access condition analysis means 7 at each time of the occurrence of access from a host computer 1 to a disk device 2. This information is fully recorded in the means 7. This recorded access information is analyzed at intervals of a certain time. At the time of analysis, access is classified with respect to each unit. The overall address area of the disk device 2 is divided by a proper size, and an accumulated value is calculated for each command in each section. Two sections are extracted from calculation results. That is, the section where the accumulated value is higher on the whole and the write operation occurs by a write command and the section where the accumulated value is higher on the whole and the write operation and the read operation equally frequently occur by write and read commands are extracted. A non-cache attribute table 8 is changed by the analysis of them.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a disk cache control system for improving the performance of a disk device by increasing the hit rate of the disk cache.

[0002]

2. Description of the Related Art A disk device requires seek time and rotation waiting time once accessed due to its mechanical structure, and the access time becomes much slower than main memory. Therefore,
By providing a cache memory on the disk controller, the data once accessed to the disk device is also stored in this cache memory, and when an access request to the same data is generated next time, the data in this cache memory is saved. By using it, direct access to the disk device itself is avoided, and the disk access time is shortened. However, since the capacity of the cache memory is generally smaller than the capacity of the disk device for cost reasons, when storing new data on the disk device in the cache memory, it is already stored in the cache memory. The need arises to eliminate this data. At this time, LRU (L
east Recently Used) or the like is used. This algorithm is such that the most recently accessed data has a higher priority of being resident in the cache memory. However, if it is simply managed by the LRU algorithm, the data having a certain access property will certainly stay in the cache memory for a long time, and as a result, the probability that it will exist in the cache memory when accessed by the host computer will increase. Get higher However, for data that has a different access property, data that does not need to be accessed will be placed in the cache memory for a long time, thus eliminating data that may be accessed in the future. There was a case.

For example, in a database system, log information is recorded separately from normal data for data preservation. Although this log information is also written to the disk device, due to the nature of the log information, most of them are write operations, and the access method is generally such that the addressing is sequentially and sequentially added. Further, normally, once written data is rarely read, and is only referred to at the time of data recovery in case of emergency. On the other hand, in update transaction processing represented by "bank inquiries" that randomly access a wide range of disk device addresses, once the read data has undergone some processing, it will be stored in the same location again. Often written back to. After being written back, it is assumed that this data is generally unlikely to be used because it randomly accesses a wide area.

Therefore, as in the case described above, useless data that is unlikely to be used in the future is left in the cache memory for a long time by the LRU algorithm,
Further, when the replacement process occurs, a cache block that is likely to be accessed from now is selected and eliminated. As a result, the hit rate of the cache decreased, leading to poor performance.

Therefore, for example, in Japanese Unexamined Patent Publication No. 2-35544, data that can be used again (hits) is examined by extending the stay time in the cache memory a little more, and the results of this investigation are reflected to improve the hit rate. This is provided with a normal data managing means for managing the cache memory by the LRU and a second data managing means for managing a predetermined amount of the data that has been expelled from the cache memory in the past.
Then, when the certain data is removed from the cache memory according to the LRU, it is checked whether or not the second data management means has expelled the data in the past, and if the history remains here, the LRU of this data is Even at the lowest level, the LRU of this data is changed to the highest level, the history of this data is deleted from the second data management means, and then other data is excluded.

As another method, for example, Japanese Patent Laid-Open No.
In 51342, the cache data to be excluded is determined by using not only the LRU but also the access count. This is to record not only the LRU but also the access count for each cache block, and for blocks with a certain access count or more, other blocks are excluded even if the LRU is the lowest. Is.

Further, a method of calculating statistical information of access from the host computer and incorporating this into cache management is also proposed. For example, JP-A-1-128145
Then, the access time and address information are registered for each of reading and writing to the disk device. Then, the accumulation time from writing to reading to the same individual address is calculated, the distribution is examined, a certain time T is determined from there, and the write data in the cache memory is eliminated when this time T has elapsed. is there.

[0008]

Since the conventional disk cache control is configured as described above, even if the data can be used by extending the stay time in the cache memory a little more, the cache memory can be used once in the past. There is a problem that it cannot be judged as the relevant data unless the data has the history excluded from the above.

Further, even if the data is frequently accessed until then, there is almost no possibility that this data will be accessed in the future due to the change of the business application. Since the access count is accumulated, there is a problem that data whose access count is still small despite being predicted to be frequently used is excluded from the cache memory.

Further, the time T is determined by integrating the entire accumulation time distribution of the data having various attributes, and the data is stored in the cache memory within this time, so it is only written out and not read. Even for data,
The result is that the cache memory is allowed to stay for a certain period of time, and there is a problem that other data is excluded accordingly.

The present invention has been made in order to solve the above problems, and in consideration of the access form to the cache block in addition to the LRU algorithm, the replacement priority order or replacement timing of the cache block is set. It is an object of the present invention to provide a disk cache control method with a high hit rate that reflects the system operating status by making a decision.

[0012]

According to a first aspect of the present invention, there is provided a disk cache control system, which controls data transfer between a host computer and an external storage device, holds a part of transfer data, and the cache memory. In the external storage control means, the external storage control means comprises a cache management means for dividing and managing the cache block into an appropriate size and a cache control means for performing cache control based on the information of the cache management means. A cache attribute management unit that manages an access attribute from the host computer to the external storage device is provided, and the cache control unit performs cache block management based on the management information held by the cache management unit and the cache attribute management unit. is there.

According to a second aspect of the present invention, in the disk cache control system according to the first aspect, the cache attribute management means for managing the access attribute includes access command information, address information, length information and access form information, and a host It can be set from a computer.

According to a third aspect of the present invention, in the disk cache control system according to the first aspect, the external storage control means comprises access status analysis means, and the access status analysis means analyzes the access status from the host computer to the external storage device. However, the cache attribute management means is set based on the analysis result.

According to a fourth aspect of the present invention, in the disk cache control system according to the third aspect, the access status analyzing means analyzes that the access attribute to the external storage device is concentrated in a certain section by a write operation. This is a criterion.

According to a fifth aspect of the present invention, in the disk cache control method according to the third aspect, the access status analyzing means performs a data read operation from the host computer to the external storage device for an address having the same access attribute to the external storage device. The analysis judgment criterion is that the next access mode following is a data write.

According to a sixth aspect of the present invention, in the disk cache control method according to the second to fourth aspects, the priority change for the cache block replacement by the cache control means follows the cache block access from the host computer. This is done at the time when the cache block area of is accessed.

The seventh invention is the second or third or fifth invention.
In the disk cache control method according to the invention,
The priority change for the cache block replacement by the cache control means is performed at the time when the same cache block area is accessed subsequently to the cache block access from the host computer.

[0019]

In the disk cache control system according to the first aspect of the present invention, the cache control means divides the cache memory into cache blocks for management, and the cache attribute for managing access attributes from the host computer to the external storage device. The priority order of the cache blocks to be replaced is managed based on the management information held by the management means.

According to a second aspect of the present invention, in the disk cache control method according to the first aspect of the present invention, information of a cache attribute management means configured to include access command information, address information, length information and access form information is stored in a host computer. Can be set from.

According to a third aspect of the present invention, in the disk cache control system according to the first aspect, an access status analyzing means is provided, and the access status analyzing means analyzes the access status from the host computer to the external storage device,
The analysis result is set in the cache attribute management means.

According to a fourth aspect of the present invention, in the disk cache control method according to the third aspect, the access status analyzing means analyzes that the access attributes to the external storage device are concentrated in a certain section of the address by the write operation. Use as a criterion.

According to a fifth aspect of the present invention, in the disk cache control system according to the third aspect, the access status analyzing means performs a data read operation from the host computer to the external storage device for an address having the same access attribute to the external storage device. The analysis reference is that the next access mode following is a data write.

According to a sixth aspect of the present invention, in the disk cache control method according to the second to fourth aspects, the cache control means changes the priority order for replacing the cache block, following the cache block access from the host computer. This is performed when the cache block area of is accessed.

The seventh invention is the second or third or fifth invention.
In the disk cache control method according to the invention,
The cache control means changes the priority order for the cache block replacement at the time when the same cache block area is accessed subsequent to the cache block access from the host computer.

[0026]

【Example】

Example 1. An embodiment of the present invention will be described below with reference to the drawings. In FIG. 1, 1 is a host computer, 2 is a disk device that stores data from the host computer, 3 is a disk control device that controls data transfer between the host computer 1 and disk device 2, and 4 is the host computer 1 and disk device. Two
A cache memory 5 for holding a part of the data to be transferred between the host computer 1 and the disk device 2 judges whether or not the transfer data between the host computer 1 and the disk device 2 exists in the cache memory 4, and if there is, the data transfer from the cache memory 4 A cache control unit 6 for controlling the cache memory 4 and a cache management table 6 for managing the data in the cache memory 4 are used by the cache control unit 5 to determine whether the cache is hit or unhit. Further, 7 is an access status analyzing means for recording and analyzing an access history from the host computer, and 8 is a non-cache attribute table for recording the attribute of meaningless data even when cached.

FIG. 2 is a diagram showing the configuration of the cache management table 6, and shows the configuration of the cache memory 4 of the 4-way set associative system. Also, 6
0 is block information for managing 4 sectors of the disk device 2 as one cache block.

FIG. 3 is a diagram showing the structure of the block information 60. In the figure, 600 is a block valid flag indicating whether the data of this block is valid, 601 writes the data from the host computer 1 to the data of this block, and has not yet written to the disk device 2. A block update flag indicating that there is data, 604
Is the block address of the data held by the block information 60, and also indicates the upper part of the sector address of the disk device 2. Further, 605 is a unit address for identification when there are a plurality of disk devices 2, and 6
06 is an L for indicating the access frequency of this block information
It is a RU flag.

FIG. 4 is a diagram showing the structure of the non-cache attribute table 8. 81 is command information indicating a non-cached access command, 82 is address information indicating a non-cached access start address, and 83 is address information. Length information indicating a limit value of 82, and 84 is a log information bit indicating whether or not each entry is a log area.

FIG. 5 is a flow chart for explaining the cache control at the time of writing according to the present invention, and FIGS. 6 and 7 are explanatory diagrams for making it easier to understand the embodiment of the present invention.

Next, the operation will be described with reference to FIGS. 1 to 7. First, the basic usage of the non-cache table will be described. Host computer 1 to disk device 2
It is assumed that there is a write request for 2 sectors from the address 10000 (hereinafter referred to as access (a)). The cache control unit 5 searches the cache management table 6 and checks the block valid flag 600 in the block information 60 to check whether or not the data of the request address exists in the cache memory 4. As a result, if there is the corresponding data in the cache memory 4, the data is written to this block (step S13). If the block does not exist in the cache memory 4, the presence or absence of an unused block is checked (step S3), and if the block does not exist, the LRU removes the data of the lowest block from the cache memory. Perform processing to do so.
Therefore, if the LRU has changed the lowest block data (step S4), this block is written back to the disk device 2 (step S5), and an empty block is allocated (step S6). Write data to that area. Next, the cache control unit refers to the non-cache attribute table 8 (step S14) and determines whether or not this access (a) has a certain attribute (step S15).

Here, for example, this access request (a)
Is a write to address 10000. Then, the non-cache attribute table 8 is referenced. here,
The non-cache attribute table 8 has information of a write access from the address 20000 with a length of 10000, and a write to this area will not be used later even if it is cached (for example, the log data 84 of the database system). Data area. First, since the command attribute 81 of the non-cache attribute table 8 is written as "write", this access request (a) is applicable. Next, referring to the address attribute 82 of the non-cache attribute table 8, since the access request (a) this time is for the address 10000, it does not apply. Therefore, the cache control unit 5
Changes the LRU flag of this block to the highest level as usual (step S11).

Next, it is assumed that there is a write request (hereinafter referred to as an access request (b)) for 2 sectors from the address 25000 of the disk device 2 from the host computer 1. The cache control unit 5 searches the cache management table 6 and determines whether or not the data of the requested address exists in the cache memory 4 by checking the block valid flag 600 in the block information 60. Thereby, if there is the corresponding data in the cache memory 4, the data is written in this block. Cache memory 4
If not, step S3 and step S4 are executed if necessary, and the LRU writes the lowest block to the disk device (step S5) to allocate an empty block (step S6). Write data to that area. Next, the cache control unit 5 refers to the non-cache attribute table 8 (step S14) and determines whether or not this access request (b) has a certain attribute (step S15). First, since the command attribute 81 of the non-cache attribute table 8 is written as write, this access request (b) is applicable. Next, referring to the address attribute 82 of the non-cache attribute table 8, since this access request (b) is an access to the address 25000, this time it is applicable. Finally, the length information 83 of the non-cache attribute table 8 is referred to, and since the length information 83 indicates 10000, all the access requests (b) this time satisfy the condition.
As described above, the cache control unit 5 changes the LRU flag 606 of the block corresponding to the access request (b) this time to "lowest", which is the opposite of the normal one (step S16).

Further, it is assumed that the host computer 1 makes an access request for 2 sectors from the address 40000 of the disk device 2 (hereinafter referred to as access request (c)). Similar to the above, first, the cache management table is searched, and if it is not in the cache memory 4, the LRU replaces the lowest block. That is, at this time, the target of replacement is the write block for the address 25000 of the access request (b), even though the access request (b) was most recently accessed, and the access request is issued before this. It is not the write block of the address 10000 (a). In this way, an access corresponding to the non-cache attribute table 8 (for example, log data of the database system) is preferentially replaced from the cache.

From here, the non-cache attribute table 8
The operation of changing the definition of from the host computer will be described. It is assumed that a meaningless area that is cached systematically in the disk cache has changed during system operation.
Then, the host computer issues a non-cache attribute change command (hereinafter referred to as command (4)) to the disk control device 3. The disk controller 3 changes the contents of the non-cache attribute table according to the information in this command (4). For example, it is assumed that the instruction (4) has an instruction to de-write the write command from the address 10000 to the length 20000. The disk controller compares the instruction information of this command (4) with the contents of the address information 82 and the length information 83 of the non-cache attribute table 8. If the address range indicated by the instruction (4) matches any of the entries of the non-cache attribute table 8 or if a part of the areas match, the contents of this entry are changed. On the contrary, if there is no match, a new entry is added and registered. In this way, the host computer knows the useless area even if the host computer caches it in the disk control device in advance, and the cache control is executed so that the access to the useless area even if the cache is performed is preferentially replaced from the cache. It

From here, the non-cache attribute table 8
The principle of changing the definition of (1) by the disk controller itself judging from the operating condition will be described. Each time the host computer accesses the disk device, the cache control unit 5 notifies the access status analysis unit 7 of the access information. The access status analysis means 7 records the information one by one. This access information includes the unit number, address, command, etc. of the accessed disk device. Also,
The access status analysis means analyzes this access record at regular intervals. In the analysis, access is first classified by unit. Next, the entire address area of the disk device is divided into appropriate sizes, and the accumulation is calculated for each command for each section. For example, FIG. 6 is a graph for easy understanding of the result of calculation of accumulation for one disk device, where the horizontal axis represents the sector address when accessing the disk device and the vertical axis represents the number of accesses. Then, the access status analysis means 7 determines 2 from this result.
Extract one feature. One feature is that the cumulative value is much higher than the whole, and it is almost a write command. In FIG. 6, it corresponds to the portion (1). Another one
One of the characteristics is that the cumulative value is relatively high compared to the whole, and the ratio of the command corresponds to the part (2) of FIG. 6 in the area where the write operation and the read operation occur to the same extent. From these analysis results, it can be seen that the region corresponding to the first feature corresponds to the log region described above and the region corresponding to the second feature corresponds to the random region described above. Next, the access status analysis means sets these areas in the non-cache attribute table 8 as non-cache attributes. One is an area obtained by combining an area corresponding to the above-mentioned part (1) as a log area in the non-cache attribute table and an area having an appropriate size thereafter, and the log information bit 84.
To enable. In this case, the command attribute 81 is “write”, the address attribute 82 is 10000, and the length information 8
3 is set to 10000, and the log information bit is set to "1". Not only the log area of the analysis result but also the area after the log area is added because the log area is sequentially accessed and the area beyond this is presumed to be the log area. The other is to set an area corresponding to the above-mentioned part (2) as a random access area. In this case, the command attribute 81 is “write”, the address attribute 82 is 30,000, the length information 83 is 20,000, and
Set "0" to the log information bit. Command attribute 8
The reason why "write" is set as 1 instead of "read" will be described later. In this way, the access status analysis means 7 analyzes the access status from the host computer and changes the contents of the non-cache attribute table 8.

Next, an operation will be described in the case where the timing for changing the LRU to the lowest order is not performed at the time of access, but at the time of access to the area corresponding to the cache block next to the accessed cache block. First, it is assumed that the host computer 1 makes a write request for 2 sectors from the address 15000 of the disk device 2. This access is for the non-cache attribute table 8
Since this corresponds to the corresponding block (corresponding to the log area of (1) in FIG. 6), the LRU of this block may be the lowest rank at this point. However, if one block has 4 sectors as in the present example, FIG.
As shown in, the remaining area portion is present in (2) of this block. Since the log area is often accessed sequentially, there is a high possibility that the next log access will write to the rest of the block in FIG. Therefore,
If the LRU is set to the lowest level at this point, this block may be replaced by the next log access. Therefore, if the log bit is valid even if an access corresponding to the log area of the non-cache attribute table 8 occurs, the LRU is not immediately set to the lowest rank at that time, and as shown in (3) of FIG. At the time when the access occurs, that is, when the access moves to the next block (m + 1),
The LRU of the cache block m + 1 corresponding to the previous access (2), not the cache block m + 1 corresponding to the access (3), is controlled to be the lowest. By doing so, it is possible to effectively deal with a case where a plurality of accesses occur in the same block.

Finally, the operation is performed when the timing of changing the LRU to the lowest is not at the access time, that is, when the first access is a data read and the next access is a data write at the time of this data write operation. explain. The data in the random area as shown in (2) of FIG. 6 is once read by the host computer like a bank accounting transaction and subjected to some processing.
It is likely to be written again. Therefore, if the LRU is immediately set to the lowest level at the time of access by this read,
There is a possibility that this block will be replaced before the above rewrite access is performed. Therefore, the command information 81 of the non-cache attribute table 8 is set to "write" instead of read.
Then, when the host computer 1 subsequently makes a write request for 2 sectors from the address 40000 of the disk device 2, the cache control unit 5 causes the non-cache table 8
Since the address information 82, the length information 83, and the log bit information 84 all match, including the command information 81, the LRU of this block is changed to the lowest. By controlling in this manner, rewrite access to the same block can be cache-hit even in the above random area data.

[0039]

As described above, according to the present invention, the replacement priority of the cache block is determined based on the access attribute management information in addition to the LRU algorithm of the cache block, so that unnecessary data is cached. There is an effect that the hit rate of the disk cache can be improved without storing it in the memory for a long time.

Further, since the access attribute management information can be set from the host computer, it is possible to realize a cache system suitable for the operating mode, and it is possible to realize a disk cache with a high hit ratio without depending on the system. There is an effect.

Further, since the external storage control means itself can change the access attribute management information, it is possible to realize a cache system that sensitively reflects the system operating status.
There is an effect that the hit rate of the disk cache can be improved.

Further, the sequential access area is estimated as the access attribute, and the data which is less effective even if the continuous access is predicted to be cached is discharged from the cache memory, so that other data can be stored for a long time. Since the data can be stored in the cache memory, there is an effect that the hit rate of the disk cache can be improved.

Further, the random access area by the read operation and the write operation is estimated from the access status, and the data corresponding to the area is discharged from the cache memory after the update processing is completed. In this case, it is possible to eliminate the waste of reading the data block again in the cache memory for writing because it is discharged, and it is possible to improve the hit rate of the disk cache.

[Brief description of drawings]

FIG. 1 is a block diagram showing a disk control device according to an embodiment of the present invention.

FIG. 2 is a configuration diagram showing a cache management table according to an embodiment of the present invention.

FIG. 3 is a configuration diagram showing block information according to an embodiment of the present invention.

FIG. 4 is a configuration diagram showing a non-cache attribute table according to an embodiment of the present invention.

FIG. 5 is a flowchart showing cache control at the time of writing according to an embodiment of the present invention.

FIG. 6 is a graph showing an analysis of access from a host according to an embodiment of the present invention.

FIG. 7 is a schematic diagram showing access from a host according to an embodiment of the present invention.

[Explanation of symbols]

1 host computer, 2 disk device, 3 disk control device, 4 cache memory, 5 cache control unit, 6 cache management table, 7 access status analysis means, 8 non-cache attribute table.

Claims (7)

[Claims] 1. A cache memory that controls data transfer between a host computer and an external storage device and holds a part of the transfer data, and a cache management that manages the cache memory by dividing it into cache blocks of an appropriate size. Means and cache control means for performing cache control based on the information of the cache management means, wherein the external storage control means manages an access attribute from the host computer to the external storage device A disk cache control method comprising: management means, wherein the cache control means performs cache block management based on management information held by the cache management means and the cache attribute management means. 2. The cache attribute management means for managing the access attribute includes access command information, address information, length information and access form information, and is set from a host computer. Disk cache control method. 3. The external storage control means comprises an access status analysis means, and the access status analysis means analyzes the access status from the host computer to the external storage device and makes the cache attribute management means based on the analysis result. The disk cache control system according to claim 1, wherein the disk cache control method is set. 4. The access status analysis means is characterized in that the access attribute to the external storage device is based on the analysis judgment that the access addresses are concentrated in a certain section by a write operation. The disk cache control method described in item 3. 5. The access status analyzing means confirms that the next access form is data write after the data read operation from the host computer to the external storage device for the same address having the same access attribute to the external storage device. 4. The disk cache control method according to claim 3, wherein the analysis criterion is used. 6. The priority change for cache block replacement by the cache control means is performed at the time when the next cache block area is accessed subsequent to the cache block access from the host computer. The disk cache control system according to any one of claims 2 to 4, wherein: 7. The priority change for cache block replacement by the cache control means is performed at the time when the same cache block area is accessed subsequent to the cache block access from the host computer. 6. The disk cache control system according to claim 2, 3, or 5.