JPH08115169A - Disk controller - Google Patents

Abstract

PURPOSE: To execute efficient destaging processing by selecting continuous areas to be destaged while referring to the quantity of dirty data in each continuous data at the time of destaging dirty data stored in a cache in the continuous areas. CONSTITUTION: In the case of executing the destaging processing 7401 of dirty data stored in a cache memory 5700 in continuous areas, a processor 5800 selects a cylinder group 5550 to be destaged 7402. In respect to the selection, the quantity of dirty data in each cylinder group 5550 included in dirty segment managing information 6100 is referred to 8200 and the cylinder group 5550 having the maximum dirty data quantity is selected. Thereby when the memory 5700 is filled with dirty data, the dirty data in the memory 5700 can be quickly swept off to a disk.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a disk control device, and more particularly to a disk control device with a cache for performing write-back control, in which a dirty data which is not reflected in the disk is accumulated in the cache at an early stage. The present invention relates to a disk control device which is effective for sweeping data to a disk and creating a free area on a cache.

[0002]

2. Description of the Related Art A head scheduling method in a disk control device is such that a SCAN algorithm for sequentially servicing requests in the disk radial direction, an FCFS algorithm (or FIFO algorithm) for servicing requests in the order of arrival, and a shortest seek time are required. There are known SSTF algorithms and the like that sequentially serve the following requests. A head scheduling method in such a disk control device is described in, for example, document [Teorey].

The superiority or inferiority of the head scheduling method in the disk control device is related to the relationship between the seek distance and the seek time in the disk device. The relationship between seek distance and seek time of the disk device is described in, for example, document [Hennessy].

A disk cache is known as a technique for improving the response performance of the disk subsystem. Furthermore, I / O is completed by simply writing the write data to the cache memory at the time of write access, and the data not reflected on the cache memory (generally called dirty data) is written to the disk asynchronously with the write access. Write back control is known.

Japanese Laid-Open Patent Publication No. 3-172946 discloses a method of sweeping dirty data on a cache memory to a disk in a disk subsystem with a cache for performing write-back control. In the method described in this publication, dirty data with the oldest access time, that is, in the LRU state, is the target of disk sweep. As an alternative method, the method based on the FCFS algorithm and the method based on the stay time in the cache of the truck are also mentioned.

[Teorey] TJTeorey and TBPinkerton: A
Comparative Analysis of Disk Scheduling Policies,
Communications of the ACM, March 1972, Vol.15, Number
3, pp.177-184 [Hennessy] Hennessy and Patterson: Computer Architecture, Nikkei BP, p.582

[0007]

As one method of efficiently executing the write-back control, it is conceivable to preferentially schedule a cylinder containing a lot of dirty data. However, in the conventional head scheduling method in the disk controller for performing the write-back control, it has not been taken into consideration in which cylinder of the disk the dirty data exists.

An object of the present invention is a disk control device with a cache for performing write-back control, in which the distribution of dirty data on cylinders is managed, and cylinders with a large amount of dirty data are given priority for service. By doing so, it is possible to provide a disk control device capable of performing an efficient destage processing position by destageing more dirty data with a smaller number of disk arm operations.

[0009]

The object of the invention is achieved by providing a table which manages the on-disk distribution of requests.

More specifically, this is achieved by providing a table for managing the amount of dirty data on the cache memory for each one or more consecutive cylinder groups (cylinder groups).

[0011]

The disk controller manages the amount of dirty data for each of one or more continuous cylinders (cylinder groups) of the disk using a table. Specifically, this table is updated as follows. The disk controller adds the dirty data amount of the corresponding cylinder group when the dirty data amount increases during the processing of the write command from the host computer. Further, the disk controller subtracts the dirty data amount of the cylinder group targeted for destage during the destage processing. At the time of destage processing, the dirty data amount of each cylinder group is referred to, the cylinder group to be destaged is selected, and the destage processing is performed on this cylinder group. As described above, the object of the present invention can be achieved.

[0012]

【Example】

(Embodiment 1) In this embodiment, at the time of destaging, a group of cylinders in which a large amount of dirty data is collected is found, and the dirty data belonging to the group of cylinders is preferentially destaged, so that a high performance It realizes stage processing.

Next, the components of the present invention and their outline will be described with reference to FIG. In FIG. 1, 5100 is a host computer, 5300 is a disk controller, and 5500.
Is a disk drive. Reference numeral 5200 is an I / O bus connecting the host computer 5100 and the disk controller 5300, and reference numeral 5400 is a disk access path. The disk controller 5300 manages the storage area of the disk drive 5500 by dividing it into a total of L one or more continuous cylinder groups (cylinder groups) 5550.

5800 is a disk controller 530
A processor in 0, 5600 is a control memory, and 5700 is a cache memory. On the control memory 5600,
Dirty segment management information 6100 for managing the dirty data of each cylinder group 5550 and the latest seek destination address 6400 of the disk drive 5500.
Is included. The processor 5800 is a command processing program 7101 for processing a read / write command from the host computer 5100, a cache memory 570.
The destage processing 7401 for sweeping dirty data on 0 to the disk is executed. The destaging process 7401 includes step 7402 of selecting a cylinder group 5550 to be destaged, and the selected cylinder group 55.
A destage processing step 7403 for 50 is included.
The dirty segment management information 6100 includes the amount of dirty data for each cylinder group 5550.

The outline of the present invention will be described below. Upon receiving the write command, the command processing 7101 updates the dirty data amount of the corresponding cylinder group 5550 according to the content of the command (8000). Step 740: Select cylinder group 5550 for destage
2 refers to the dirty data amount for each cylinder group 5550 included in the dirty segment management information 6100 (8
200) and selects the cylinder group 5550 with the largest amount of dirty data. In step 7403 of performing the destage process for the cylinder group 5550, the destage process (8400) of the cylinder group 5550 selected in step 7402 is performed, and the cylinder group 555 in the dirty segment management information 6100.
The dirty data amount of 0 is changed (8300).

As a result, it becomes possible to perform the destage process in consideration of the distribution of dirty data on the cylinder.

Hereinafter, FIGS. 2 to 7, FIG. 9 and FIGS.
The details of this embodiment will be described with reference to FIG.

For simplicity, the description will be made assuming that one disk drive is connected to the disk controller, but the present invention can be applied to any number of disk drives.

In this embodiment, the cache memory is divided into fixed length areas for management. This area is called a cache segment. The storage area of the disk drive can be regarded as a one-dimensional address space. When the storage area of the disk drive is divided into areas having the same length as the cache segment and handled, this area on the disk drive is called a disk segment.

In this embodiment, the symbol L is used to represent the total number of cylinder groups.

FIG. 2 shows the configuration of the control table on the control memory 5600. A hit / miss determination table 6000 is used to determine a data hit / miss. The structure of the hit / miss determination table 6000 is shown in FIG. 610
0 is dirty segment management information for managing the cache segment in the dirty state. The structure of the dirty segment management information 6100 is shown in FIG. 6200 is clean / free cache segment management information for managing cache segments in the clean state and the free state. Clean / empty cache segment management information 62
The configuration of 00 is shown in FIG. 6300 is a cache management table for managing cache segments. The structure of the cache management table 6300 is shown in FIG. 6400 is the latest seek destination address of the disk drive 5500. The latest seek address is
Disk controller 5300 is disk drive 5
When issuing a command with a seek operation to 500,
It can be managed by writing the cylinder address of the seek destination in the area 6400.

FIG. 10 shows a hit / miss judgment table 6000.
Is shown. Hit-miss judgment table 6000
Has one entry for each disk segment, and the entry includes a hit-miss determination flag 6010 and a cache segment number 6020 when the disk segment is assigned to the cache memory.

FIG. 11 shows the dirty segment management information 61.
00 configuration is shown. The dirty segment management information 6100 includes a dirty data amount 6110 for each cylinder group and a pointer 6120 for the queue of the dirty cache segment for each cylinder group for each L total of cylinder groups.

FIG. 12 shows the structure of the clean / free segment management information 6200. The clean / free segment management information 6200 is composed of a pointer 6210 of the queue of the cache segment in the clean state and a pointer 6220 of the queue of the cache segment in the free state.

FIG. 13 shows the cache management table 6300.
Is shown. Cache management table 6300
Has an entry for each cache segment. The entry includes a busy flag 6310 indicating whether the cache segment is in use, a dirty flag 6320 indicating whether it is in a dirty state, a latest reference time 6330, a corresponding disk segment address 6340, a cache memory address 6350 of the cache segment, Backward pointer 6 in the queue to which the cache segment is connected
Including 360.

Next, the operation of the disk device in response to a read / write command from the host computer to the disk device will be described.

FIG. 3 is a flowchart of the command processing program 7101 executed by the disk controller 5300. The command processing program is called subsequent to the arrival of a command from the host computer (step 7101), determines whether the command is read or write (step 7102), makes a hit-miss determination (step 7103 and step 7107), and reads. /
Read hit processing (step 7104) and read miss processing (step 7) depending on whether write or hit / miss
109), write hit processing (7108), or write miss processing (7110), and then destage processing is started (step 7105) and the processing ends (step 7106).

The write hit process 7108 and the write miss process 7110 include a process of updating the dirty data amount 6110 for each cylinder group when the dirty data amount increases. The write hit process and the write miss process are developed in FIGS. 4 and 5, respectively.

The read hit process and the read miss process are the same as those known in the art.

FIG. 4 is a flowchart of the write hit process. When the write hit process is called (step 7201), it is determined whether the hit cache segment is in the clean state (this case is called a clean hit) or the dirty state (this case is called a dirty hit. Of course, it is not a clean hit). Judge (step 72)
02). If it is a clean hit, the dirty data amount increases, so the dirty data amount 6110 of the corresponding cylinder group is incremented (step 720).
3), write clean hit processing (step 720)
4) is performed, and the process ends (step 7206).
If it is not a clean hit, the dirty data amount does not change, so a write dirty hit process is simply performed (step 7205), and the process ends (step 72).
06).

FIG. 5 is a flowchart of the write miss process. In the case of a write miss, a cache segment in a clean state or an empty state is newly allocated, so that the amount of dirty data on the cache memory always increases. When the write miss process is called (step 73)
01), write miss processing is performed (step 730).
2), the amount of dirty data for the corresponding cylinder group 6
110 is decremented (step 7303), and the process is ended (step 7304).

FIG. 6 is a flowchart of the destage process in this embodiment. When the destaging process is activated (step 7401), the cylinder group to be destaged is selected (step 7402), the destaging process is subsequently performed on the selected cylinder group (step 7403), and the process is terminated (step 7403). Step 7
404). The processes of step 7402 and step 7403 are expanded into FIGS. 7 and 9, respectively.

FIG. 7 is a flow chart for realizing the method of selecting the cylinder group having the maximum dirty data amount. i is a work variable indicating the cylinder group number. j is a variable that holds the number of the cylinder group with the largest dirty data amount among the cylinder group numbers 1 to j, and x is the dirty data amount of cylinder group j. In step 7502, 0 is set to the variable x,
Substitute 1 for the variable i. It is checked in step 7503 if the cylinder group i has dirty data. If there is no dirty data, the process jumps to step 7507. If there is dirty data, the dirty data amount of the cylinder group i is obtained from the table 6100 and set in y, and the values of y and x are compared (step 750).
5). If the value of y is large, the cylinder group i is the cylinder group having the maximum dirty data amount up to that point, the value of x is changed to the value of y, and the value of j is set to i (step 7506), and the process proceeds to step 7507. move on. Step 7
If the value of y is less than or equal to the value of x in 505, nothing is done and the process jumps to step 7507. In step 7507,
It is checked whether i has reached the maximum value L of the cylinder group number. If i is not equal to L, the value of i is incremented by 1 (step 7510) and the process returns to step 7503 to continue the process. If the value of i is equal to L, it means that processing for all cylinder groups has been completed.
The cylinder group j is the cylinder group having the largest dirty data amount, the cylinder group j is set as the destage target (step 7508), and the process ends (step 7509). The above is the method of selecting the cylinder group with the largest amount of dirty data.

FIG. 9 is a flowchart showing the destage process for the cylinder group. Step 760
After performing the destage process for the cylinder group selected in 2, the dirty data amount 6110 of the cylinder group is decremented in step 7603, and the process ends (step 7604).

According to the method described in the present embodiment, in a cached disk device which performs write-back control, when destaging is performed, a group of consecutive cylinders having the largest amount of dirty data is selected and the cylinders of that cylinder are selected. By performing the destaging process for the group first, an efficient destaging process can be realized.

(Embodiment 2) In this embodiment, at the time of destage, a group of cylinders having a minimum value obtained by dividing the seek time by the amount of dirty data is found, and the dirty data belonging to the group of cylinders is given priority. By performing the destage, the high-performance destage processing is realized.

Since the process is the same as that of the first embodiment except the process of step 7402 for selecting the cylinder group to be destaged, only the process of step 7402 will be described.

In this embodiment, a method of selecting a cylinder group to be destaged based on a concept different from that of the first embodiment will be described. This is a method in which the seek time from the current head position to the cylinder group having dirty data is divided by the amount of dirty data, and the cylinder group having the smallest value is destaged first. With this method, the amount of dirty data that can be destaged per unit seek time is the largest in one seek operation. The seek time can be calculated by using a function that approximates the relationship between the seek distance and the seek time. FIG. 8 shows a flowchart of this cylinder group selection method. In FIG. 8, i
Is a variable indicating the number of the cylinder group being processed, j is the number of the cylinder group selected by the time of the search, and x is the minimum (seek time / dirty data amount) value up to that time. When this procedure is read (step 7501), x is the maximum value of the processing system and i is 1
Is substituted (step 7511), and it is checked whether or not the cylinder group i has dirty data (step 75).
03), jump to step 7507 if there is no dirty data, and step 7512 if there is dirty data.
Then, the process proceeds to step a and the dirty data amount of the cylinder group i is obtained from the table 6100 and substituted into the variable a. Next, the latest seek destination address 6400 is regarded as the current head position, the seek time to the cylinder group i is obtained, and the seek time is substituted for the variable t (step 7513), and the value (t / t) is set for the variable y.
a) is substituted (step 7514), the values of the variable y and the variable x are compared (step 7515), and if the value of y is small, the value of x is updated to the value of y, and the value of j is updated to i. Then (step 7506), the process proceeds to step 7507. If the value of y is not smaller than the value of x in step 7515, the process jumps to step 7507. If the value of the variable i reaches the maximum value L of the cylinder group number in step 7507, the cylinder group j is destaged (step 7508), and the process ends (step 7509).
If the value of the variable i has not reached the maximum value L in step 7507, i is incremented by 1 (step 7510) and the process jumps to step 7503 to continue the processing.

According to the method described in the present embodiment, in the disk device with cache for performing write-back control, when destaging is performed, the cylinder group having the smallest value obtained by dividing the required seek time by the dirty data amount is selected. Then, by performing the destage process for the cylinder group first, the efficient destage process can be realized.

[0040]

In a disk device with a cache for performing write-back control, when the present invention is applied when the cache memory is full of dirty data, the dirty data in the cache memory can be swept to the disk earlier than in the prior art. You can

[Brief description of drawings]

FIG. 1 is an overall configuration diagram showing an outline of the present invention.

FIG. 2 is a diagram showing a configuration of a control table on a control memory.

FIG. 3 is a flowchart of command processing.

FIG. 4 is a flowchart of a write hit process.

FIG. 5 is a flowchart of a write miss process.

FIG. 6 is a flowchart of destage processing.

FIG. 7 is a flowchart showing a first method of selecting a cylinder group to be destaged.

FIG. 8 is a flowchart showing a second selection method of a destage target cylinder group.

FIG. 9 is a flowchart of destage processing for a cylinder group.

FIG. 10 is a diagram showing the structure of a hit / miss determination table.

FIG. 11 is a diagram showing a structure of dirty segment management information.

FIG. 12 is a diagram showing the structure of clean / free segment management information.

FIG. 13 is a diagram showing a configuration of a cache management table.

[Explanation of symbols]

5100 ... Host computer, 5300 ... Disk controller, 5500 ... Disk drive,
5600 ... Control memory, 5700 ... Cache memory, 5800 ... Processor, 6100 ... Dirty segment management information, 6110 ... Dirty data amount of each cylinder group, 6400 ... Latest seek destination cylinder address, 7101 ... Command processing program, 74
01 ... Destage processing program, 7402 ... Step for selecting cylinder group to be destaged, 7
403 ... Destage processing step for the selected cylinder group.

Claims (4)

[Claims] 1. A disk controller with a cache for connecting a disk device and performing write-back control, comprising:
For each one or more continuous areas included in the disk device,
Means for managing the amount of dirty data that is write data that is present in the cache and that has not been written in the continuous area; and when destaging the dirty data in the cache to the continuous area, for each continuous area 2. A disk control device comprising means for selecting a continuous area to be destaged by referring to the amount of the dirty data of. 2. The disk control device according to claim 1, further comprising means for selecting a continuous area having a maximum amount of the dirty data when selecting the continuous area to be destaged. The disk controller. 3. A disk controller with a cache for connecting a disk device and performing write-back control, comprising:
For each one or more continuous areas included in the disk device,
Means for managing the amount of dirty data that is write data that is present in the cache and has not been written in the continuous area; means for managing the latest seek destination address; and the dirty data on the cache for the continuous area. The disk control device further comprises means for selecting a continuous area to be destaged by referring to the dirty data amount for each continuous area and the latest seek destination address when destaging. 4. The disk control device according to claim 3, wherein when the continuous area to be destaged is selected, the continuous area seek time is obtained using the latest seek destination address, The value obtained by dividing the seek time to the continuous area by the amount of the dirty data in the continuous area is
A disk controller comprising means for selecting a continuous area having a minimum positive value.