JPH02239348A - Access system for disk cache memory - Google Patents

Abstract

PURPOSE:To shorten a transfer time by dividing the data area of magnetic disks at every data, allocating independent device addresses to respective division areas and simultaneously transferring plural data concerned from respective areas of the magnetic disks and a disk cache memory through plural transfer routes by means of the designation of the device addresses. CONSTITUTION:The magnetic disks 8a-8d stored in a disk device 3 are divided in plural areas. The disk 8a is divided into three areas A, B and C, for example. Such division is decided by the setting on hardware. The disk device is constituted in such a way. A read/write instruction with respect to a record on an address A is given from the route 4a and it is given from the route 4b with respect to an address B. When both records exist on the cache memory 6, both data are simultaneously transferred by using two routes. One of the records exists on the memory 6, and transfer of the existed one is immediately executed when the other record does not exist. Then, the others are searched by using a head.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a disk cache memory access method for effectively accessing a magnetic disk and a disk cache memory in a computer system. (Prior Art) Magnetic disk devices have been developed as large-capacity external storage devices in computer systems.

However, while central processing units have become faster with advances in semiconductor technology, magnetic disk drives involve mechanical operations to read and write data, making it difficult to increase the speed due to their nature, and the speed difference between the two is It was only expanding. As a result, a situation has arisen in which the performance of the entire computer system is degraded.

In order to solve this situation, a disk cache device having a cache memory made of a semiconductor memory element is installed between the central processing unit and the magnetic disk device, and data accessed by the central processing unit is transferred to the cache memory. The data access time apparently generated by the mechanical operation of the magnetic disk device has been reduced by placing the magnetic disk drive on the top of the disk.

FIG. 2 shows an example of a computer system using such a disk cache device. (1) is a central processing unit,
(2) is a disk cache device, and (3) is a magnetic disk device. In the example in Figure 2, the central processing unit (1)
and the disk cache device are connected by two tree paths (4a) and (4b). Each route is connected to a magnetic disk control unit (5a) inside the disk cache device (2).
), (5b). The magnetic disk control units (5a) and (5b) are the disk cache device (2).
Each head drive mechanism (7a) to (7a) of the magnetic disk device (3) is connected to the cache memory (6) in the
7d), read and write data on the magnetic disks (8a) to (8d). Usually, one head drive mechanism is equipped with a plurality of magnetic heads, and accordingly, a plurality of magnetic disks are attached to one spindle and rotated by a motor. One device address "A" to "D" is assigned to each head drive mechanism, and the magnetic disk controllers (5a) and (5b) simultaneously operate different head drive mechanisms, for example (7a) and (
7b) can be used, but the same head drive mechanism cannot be used to read and write data at the same time. Next, the operation will be explained. When a program on the central processing unit (1) reads or writes data on the magnetic disk device (3), it uses either path (4a) or (4b) and first specifies the device address to drive the head. Select mechanism. Next, the cylinder address and head address are sent to specify the track on the magnetic disk to be read and written by the head drive mechanism. Furthermore, after specifying the record number of the record to be read or written on this track, a data read or write command is sent to transfer the data.

At this time, if a specific record specified by the central processing unit (1) exists on the cache memory (6), the magnetic disk control unit (5a) or (5b) uses the head drive mechanism. data is transferred only between the central processing unit (1) and the cache memory (6). However, if the record specified by the central processing unit (1) does not exist on the cache memory (6), the magnetic disk control unit (5a) or (5b)
Select the head drive mechanism specified by the device address, move it to the specified cylinder address, select the magnetic head of the specified head address, search for the specified record, and then move it to the central processing unit (1). Transfer data between. Even if the specified record exists on the cache memory (6), if the command of the central processing unit (1) is a write command, it is the same as if the record did not exist on the cache memory (6). There is also a method of writing data directly onto a magnetic disk using a head drive mechanism.

In any case, if the data transfer between the central processing unit (1) and the cache memory (6) is enough, the data must be transferred directly to the magnetic disk. Compared to the above, the moving time of the head drive mechanism, the time to search for the target record on the target track, etc. are eliminated, and the time required to start data transfer, that is, the data access time, can be significantly shortened. Records are automatically replaced in the disk cache device according to a certain algorithm so that there is always a high probability that the record requested by the central processing unit exists on the cache memory (6). control to shorten the access time for data.

Therefore, from the time when the program on the central processing unit (1) needs to read and write data on the magnetic disk device (3) until the end of reading and writing of the target data,
There are the following factors, including the data access time mentioned above. (I) Device waiting time The time spent waiting for the target device address to become free because it is already in use.

(II) Route latency central processing unit (1) and disk cache device (2)
The amount of time it takes to wait for a path to become free because all paths between are in use.

(III) Data access time After the command arrives at the disk cache device (2), it searches for the target record on the cache memory (6) or moves the head drive mechanism to search for the target record on the magnetic disk. The time it takes to start data transfer.

(IV) Data transfer time: The time it takes to actually transfer data between the central processing unit (1) and the disk cache device (2).

(V) Overhead time The time required for exchanging instructions, etc., in addition to data transfer, between the central processing unit (1) and the disk cache device (2).

The disk cache device aims to reduce the data access time of (III) among these elements (1) to (V), but in order to further improve the performance of the computer system, it is necessary to improve the remaining four elements. must also be shortened.

Therefore, regarding the reduction of waiting time for route (II),
A method is being taken to increase the number of paths that can be used simultaneously between the central processing unit (1) and the disk cache device. Furthermore, in order to reduce the data transfer time of (rV), efforts have been made to improve the data transfer speed of the path between the central processing unit (1) and the disk cache device (2), and the overhead time of (V) has been reduced. Regarding shortening, efforts are being made to improve the instruction set and reduce the number of interactions other than data transfer between the central processing unit (1) and the disk cache device (2). Regarding (I), reducing the latency time of the device, the disk cache device itself (as a result of reducing the access time for nB data, the entire system uses one device address for one data read/write) Since the time is shortened, the time spent waiting for the device address to become available is shortened, resulting in a large improvement effect.

(Problem to be Solved by the Invention) However, in the conventional disk cache device described above, for example, when a read/write command for a certain record on the device address “A′ is issued from the path (4a), the other system’s path ( 4
In b), if the device address “A゜” has already been used to read another record, the target record actually exists on the cache memory (6) and is immediately read from the path (4).
Although it is possible to transfer data using a)
There has been a problem in that commands from route (4a) are not accepted because they receive a device visit response. In addition, other system routes (
Even if you do not use the device address ゛A゜ from 4b),
For example, when the target record of an instruction previously issued from the path (4a) of the own system does not exist in the cache memory (6), and therefore the head drive mechanism (7a) at the device address 'A' is powered. A new command from route (4a) is not accepted with a device visit response, and data transfer is executed even though the target record exists in cache memory (6) and route (4a) is available. I can't.

Normally, if the central processing unit (1) actually issues commands and processes them by receiving device bus responses from the disk cache unit (2), this will only increase unnecessary processing, so the central processing The control program on the device (1) waits without issuing any input/output requests to the same device address until the read/write command previously issued from path (4a) or (4b) is completed. Therefore, even if the target record exists in the cache memory (6) and the path (4a) or (4b) is empty, input/output requests to the same device address will continue to be queued in the control program. This results in an increase in waiting time.

Moreover, this situation is due to the increase in the number of routes.
This problem not only becomes more noticeable, but also tends to become more serious as the storage capacity of each head drive mechanism increases due to the increase in the capacity of disk devices. This invention was made to solve the above problems, and even if multiple records exist on a magnetic disk that are accessed by the same head drive mechanism,
If they exist on the cache memory, we obtain a disk cache memory access method that allows data transfer between the central processing unit (1) and the disk cache device (2) simultaneously using separate paths. The purpose is to

[Means for Solving the Problems] A disk cache memory access method according to a first aspect of the present invention is a method for accessing predetermined data from among data stored on a magnetic disk whose data is read and written by one head drive mechanism. A memory area of the magnetic disk, in which the data is transferred from either the magnetic disk or the cache memory via the data transfer path by the magnetic disk control unit when a data transfer request is made from the central processing unit. is divided into a plurality of areas for each predetermined data, and an independent device address is assigned to each area, and the above-mentioned Linux disk control unit transfers each corresponding data to the cache memory and memory according to the device address when multiple data transfer requests are received from the central processing unit. The data is transferred from both magnetic disks or from the cache memory via a plurality of data transfer paths.

Further, in addition to the first invention, a disk cache memory access method according to a second aspect of the present invention includes an operation mode that specifies a cache memory access method in divided areas of a magnetic disk to which independent device addresses are assigned. The operating mode is stored and read out by device address designation to set a predetermined operating mode.

[Effect]

According to this invention, the data area of a magnetic disk in which data is read and written by a single head drive mechanism is divided into a plurality of parts for each piece of data, an independent device address is assigned to each divided area, and each device address is Since data access can be performed in units, a single head drive mechanism with a physically large storage capacity can be controlled as a plurality of independent head drive mechanisms with a small capacity. Therefore, multiple different data stored on the same magnetic disk,
Through multiple data transfer paths, the magnetic disk and cache memory, or each data can be read from the cache memory by specifying each device address and simultaneously transferred to the central processing unit. Therefore, when transferring multiple pieces of data by accessing the same magnetic disk, there is no need to wait for the previous data transfer to complete before transferring the next data.
Device waiting time is reduced.

(Example) Hereinafter, an example of the present invention will be explained with reference to the drawings.
In the figure, all the components (1) to (8) are physically the same as the conventional device shown in FIG. However, the major difference in FIG. 1 is that the magnetic disks (8a) to (8d) are logically divided and given different device addresses. For example, in FIG. 1, the magnetic disk (8a) is divided into three areas, each of which is assigned device addresses 'A', 'B', and 'C'.

Such division may be fixedly determined by hardware settings, or dynamically determined by instructions from the central processing unit.

By logically dividing a single magnetic disk in this way, the physical cylinder address and head address on the magnetic disk and the logical cylinder address and head address for each device address used by the program can be determined. However, the conversion during this time is performed within the disk cache device.

In such a device, for example, the device address 'A'
A read/write command for the record above is issued from path (4a), a read/write command for the record at device address 'B' is issued from path (4b), and if both records exist on cache memory (6), both records are data transfer can be performed simultaneously using two routes.

Also, if one record exists in the cache memory (6) but the other record does not exist, the data transfer of the record existing in the cache memory (6) will be executed immediately, and the data will be transferred immediately. For records that are missing, the head drive mechanism is used to search for the desired record on the magnetic disk. Therefore, unlike conventional devices, there is no need for one to wait until the other finishes its operation.

In addition, if both records do not exist on the cache memory (6), both records are searched sequentially on the magnetic disk using the head drive mechanism.
Data transfer can only be performed sequentially. In this case, it is the same as in the conventional device, but the probability that neither record exists on the cache memory (6) is, for example, the probability that any record exists on the cache memory (6).
%, P = (1 -0.7)x (1 -0.7)=
At 0.09, it is 9%, and it is thought that the number of occurrences of such cases will decrease.

Note that in the past, there has been a method called virtual disks that treats one disk as multiple disks in software, but this method is only used to improve usability for software users by using software to treat one disk as a logical disk. Since only a physical disk is converted, there is no effect of improving performance by reducing the waiting time of the device as in the present invention.

As a result, by assigning multiple device addresses to a magnetic disk that is accessed by a single head drive mechanism,
The waiting time of the device can be shortened and the performance of the computer system can be improved without increasing the hardware amount of the device or changing the basic structure of the control program.

The main purpose of the above embodiment is to shorten the waiting time of the device, but as in other embodiments described below, the purpose of the embodiment is to increase the usage efficiency of the disk cache device and improve the usability for the user. There is a disk cache memory access method. Generally, a disk cache device has several operating modes depending on the nature of the data on the disk in order to increase the probability that a record requested by the central processing unit exists in the cache memory and improve performance. Examples of typical operating modes include a mode that bypasses the cache memory, a mode that makes data resident on the cache memory, and an operating mode that is suitable for sequential access.

Conventionally, these operating modes are set for each head drive mechanism, or the magnetic disk accessed by one head drive mechanism is divided into several areas, and an operation mode is set for each divided area. things were being done. However, with the first method, the operation mode can be managed in units of device addresses, making it easy to manage, but as the storage capacity per device address becomes large, waste tends to occur. In this case, the operation mode can be set in small units and waste is less likely to occur, but on the other hand, both the user and the program must always be aware of the boundaries between the cylinder address and head address within the device address, making it extremely complicated. Something happened.

Therefore, in order to eliminate such problems, there is a method of setting an operating mode by specifying a given device address for each divided area of a magnetic disk that is accessed by one head drive mechanism. For example, if you want to set the area at device address ゛A'' on the magnetic disk (8a) in Figure 1 to cache bypass mode, the program on the central processing unit (1) will specify the device address ``A''. Then, a command to set the cache bypass mode can be issued.

By doing this, unlike the first conventional method, the entire magnetic disk (8a) is not set to cache bypass mode, and like the second method, when setting the mode, the magnetic disk (8a) is not set to the cache bypass mode. It is no longer necessary to specify the cylinder address and head address within the disk (8a), and the user can also recognize the area in cache bypass mode as the device address 'A' rather than the cylinder address or head address. It will be a good thing.

As described above, by setting the mode, you can set the optimal disk cache operation mode for several areas on a single magnetic disk, so you can further improve the performance of the computer system and also change the operation mode. Since settings can be made in units of device addresses, it is possible to provide a disk cache operation mode setting method that is extremely easy to manage, is more user-friendly, and is less likely to cause waste.

〔Effect of the invention〕

As described above, according to the first invention, the data area of a magnetic disk equipped with a disk cache memory is divided for each piece of data, an independent device address is assigned to each divided area, and a plurality of Since the corresponding data can be transferred simultaneously from the magnetic disk and disk cache memory, or each data from each data area of the disk cache memory, via multiple data transfer paths,
The waiting time for data transfer is reduced, and a large amount of data can be transferred in a short period of time.

Further, according to the second invention, since the optimal disk cache operation mode for several divided areas in the magnetic disk is set for each device address, the storage capacity of the magnetic disk is not wasted. Furthermore, it is possible to provide a method for setting the disk cache operation mode that is easy to use.

[Brief explanation of drawings]

FIG. 1 is a diagram of a computer system using a disk cache device for implementing a disk cache memory access method according to an embodiment of the present invention, and FIG. 2 is a computer system using a conventional disk cache device. (1) is the central processing unit, (2) is the disk cache device, (3) is the disk device, (4a) and (4b) are the paths for data transfer, (5a)
) and (5b) are magnetic disk control units, (6) is a cache memory, (7a) to (7d) are head drive mechanisms, and (8a) to (8
d) is a magnetic disk. In addition, in the figures, the same reference numerals indicate the same or corresponding parts.

Claims (2)

[Claims] (1) Data is read and written by a single head drive mechanism.It is equipped with a cache memory that copies predetermined data from among the data stored on the magnetic disk, and when a data transfer request is made from the central processing unit, the magnetic disk control unit responds to the request. In a device that transfers data from either a magnetic disk or a cache memory via a data transfer path, the memory area of the magnetic disk is divided into a plurality of areas for each predetermined data, and an independent device address is assigned to each area. , the magnetic disk control unit transfers each applicable data from both the cache memory and the magnetic disk according to the device address when multiple data transfer requests are made from the central processing unit.
Alternatively, a disk cache memory access method is characterized in that each corresponding data is transferred from a cache memory via a plurality of data transfer paths. (2) In the disk cache memory access method according to claim 1, an operation mode that specifies a cache memory access method is stored in divided areas of the magnetic disk, each of which is assigned an independent device address, and the operation mode is assigned a device address. A disk cache memory access method that reads according to specifications and sets a predetermined operating mode.