JPS60256854A - Parallel transfer type memory - Google Patents

Abstract

PURPOSE:To accelerate the working speed of an input/output system including a cache memory by loading the prescribed unit data into a cache device in case no data of a prescribed unit on a file memory exists in the cache device. CONSTITUTION:When a CPU10 performs an operation, input/output processing is needed for a disk device group 18. A director A14 checks via a directory 16 whether or not the data serving as a subject of an input/output request given from a channel A12 exists in a disk cache device 17. When no data exists in the device 17, only the data on the group 18 is renewed and read out to be sent to a channel A12 and the device 17. In this case, other data existing on the same track are possibly referred to in near future and therefore read into the device 17. In case an access pattern is clear, the data to be accessed next is read previously to the device 17. This attains the input processing at a high speed. A director B15 also has a function exactly equal to the director A14.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to the structure and control system of a director for an input/output system having a caddy memory such as a disk cache, The present invention relates to a parallel transfer type storage device that further speeds up input/output processing for a 1'10 system having a cache and a cache.

[Background of the invention]

The present invention relates to a magnetic disk device, an optical disk device, and other file memories having a cache memory, and a typical example of the disk memory will be described below.

In the next 10 series with Denisuk Catch-device,
From the viewpoint of local referentiality of data: When a read request is received and the data is not in the disk cache, processing is performed to fetch nearby data into the disk cache. However, in the case of write processing, if the data exists on the disk, the data on the disk and disk cache are rewritten; if the data does not exist, only the data on the disk is rewritten.
In the case of read processing, data can be read in advance. Therefore, since the above-mentioned processing is newly added to the director with the introduction of the disk cache, there is a possibility that the director may become a bottleneck. In order to solve this problem, the applicant of the present invention previously applied for the invention of a parallel transfer type director device in which the director is configured with two data transfer systems to prevent the director from becoming a bottleneck. (Patent Application No. 57-2094.20) However, the current disk-based input/output interface cannot effectively utilize the parallel transfer function. In the current input/output interface, when performing a search process after the disk positioning process is completed, the disk positioning is done using the D6 director and the D9=up control.
- 1' A comparison is made between the received information and the information received from the channel. Based on the comparison result, it is possible to issue an input/output request that determines whether read processing or write processing will be performed. Therefore, when positioning is complete, all devices from the disk to the channel must be secured. Therefore, even if the director has parallel transfer capability and is capable of transferring data between the disk and disk cache during the data transfer between the disk cache and the channel, at this stage the next process is the lead. Because it was not known whether it was an operation or a write operation, the positioner received a completion notification and the data on that track could not be loaded into the disk cache.

[Purpose of the invention]

SUMMARY OF THE INVENTION An object of the present invention is to provide a parallel transfer type storage device that effectively utilizes a director having a parallel transfer function and further increases the speed of an input/output system having a cache memory.

[Summary of the invention]

The case of a disc cage will be specifically described.

In the conventional interface, if the data to be input/output is not in the disk or cache, the Id' is located in the disk device. Even if the data transfer paths between the caches were free, the data on the tracks could not be loaded into the disk cages. Therefore, in the present invention, when the data to be input/output is not in the cache, the read processing other than the sequential read is performed.

By adding an input/output command that requests data to be loaded into the cage, it is possible to increase the pressure.

[Embodiments of the invention]

An embodiment of the present invention will be described below. FIG. 1 is a schematic diagram of a computer system to which the present invention is applied.
1 main memory @411 channel A12, channel B13,
The same applies to director A14 and director B15. It is composed of a disk carriage device 17 and a disk device group 18. Channel A12 and Channel B13
The functions and configurations are exactly the same. The same applies to director A14 and director B15. Multiple accesses to the disk cache device i'zta are possible at the same time, but when the directory 16 is being used by another processing system, it is necessary to wait. However, since the time required to refer to the directory 16 is short, the waiting time is not a problem.

Next, the basic operation of each device will be described.

The CPU I O performs arithmetic calculations, but input/output processing for the disk device group 18 is required. Input/output commands are assembled according to the request contents, and channel A12 or channel B13 is activated so that the respective utilization rates are equal. Channel A12 advances input/output processing while passing input/output commands one by one to director A14.

Channel B13 performs the same processing as key ref ζB15.

The director AJ4 checks the directory 16 to see if the data that is the subject of the input/output request from the channel AI2 exists in the disk cache device 17. Directory-6 indicates which portion of data of the disk device group 18 is stored on the disk cache device 17. Disk cache device 17
is usually configured with volatile IC memory, so in the case of output processing, whether or not the data to be processed exists on the disk carriage device 17, the data is stored on the disk device group 18. The process is completed when the writing is completed. For this reason, in the case of output processing, the director A14 makes sure that the requested data is not stored on the disk.
When present on the cache device 17, the disk cache device 17. The data on both the input/output target devices of the disk device group 18 is updated. On the other hand, disk
If there is no data on the carriage device 17, then
Only the data on the disk device group 18 is updated. In the case of an input request, the director A 14 transfers the data directly from the disk cache device 17 when the data exists on one device 17. When the track is not on the same track, the data is read from the disk drive group 18 and sent to the channel A12 and the disk carriage unit 17. At this time, assuming that other data existing on the same track will likely be referenced in the near future, disk·
Loaded into the carriage @17. In addition, when the access pattern is consistent, such as sequential read, the data to be accessed next can be read in advance onto the disk carriage device 17, thereby speeding up the input processing. It is possible to The above are the functions that director A14 has, but director B also has the same functions.

The above are the basic functions of the director A14 and director B15. This is where the words become established. The presence of data to be accessed on the disk cache device 17 is called a hit, and the absence thereof is called a miss. In the invention of Japanese Patent Application No. 57-209420, this function is distributed to two data transfer systems in each director, and each director has a parallel transfer function. In the present invention, in order to efficiently utilize this parallel transfer function, in the case of a read process, as shown in FIG. A new instruction 20 (LOAD TRK) that loads the data on the track onto the disk cache device flV if there is a miss.
It is distinctive in that it includes the following. (In the case of sequential read, it is known to add an instruction requesting pre-reading.) FIG. 3 shows the structure of the directories (directory A14, directory B15). The directory is
Data transfer system a30, data transfer system b31゜interface register a32, interface register b3
3. It is composed of a common register 34, a completion notification register 35, an empty block pointer 36, and a block 37.

When it is necessary to issue an input/output request to the disk device group 18, the data transfer system a30 takes out the block 37 linked to the free block pointer 36 and stores the necessary information therein. The block 37 in which information is stored is
When the data transfer system a30 receives the completion notification, the disk positioning is linked to the block 37 linked from the interface register a32. If the data transfer system b31 receives it, it is linked to the interface register b33, and if either system can receive it, it is linked to the linked block 37 from the common register 34. The data transfer system b3■ receives the positioning completion notification, and when the processing is completed, links the corresponding block 37 to the end of the blocks linked to the completion notification register 35. When all processing is completed, the block 37 is transferred to the free block pointer 36 by the data transfer system a30.

1) FIG. 4 is a detailed diagram of block 37. Pointer 40 to other blocks is a pointer for connecting with other blocks. Device ID 41 is disk device group 18
Enter the disk device number to which data will be transferred. Cylinder, track, and sector numbers 42 indicate addresses on the disk device to be input and output. The access type 43 distinguishes read miss, read hit, write miss, write hit, and prefetch of sequential read. The number of read tracks is 44.
It is meaningful only when the type 43 is sequential read, and the number of tracks to be pre-read is stored. The disk cache address 45 stores the address of the disk carriage device 17 to which data is transferred.

In this case, when a plurality of tracks are subject to data transfer, such as in prefetching of sequential read, the number of stored addresses also becomes plural. Positioning is request flag 46
The positioning for the disk device group 18 is turned ON when a request is actually issued, and turned OFF when no request is issued. Wait flag 47! Indicates that something is occurring. ;1' Next, the processing contents of the data transfer system a30 and data transfer system b31 will be described. Among the functions possessed by the director, the data transfer thread a30 is free when the data transfer thread a30 completes the positioning process of the disk device group 18, when write processing, read/hit processing, read misses other than sequential read occur. If so, accept it and send the input request gater to the channel and disk cache device 1.
7. Furthermore, when the positioning of the disk device group 18 is completed in the event of a read miss, if the data transfer system a30 is busy and the data transfer system b31 is free, the data transfer system b31 accepts the data for one track including the input request data. data into the disk cache. After this processing, the data transfer system a30 transfers the input request data from the disk cache device 17 to the channel.

The data transfer system b31 performs read-ahead processing of the sequential IJ-card processing, when a read miss occurs, when disk positioning is completed, and when the data transfer system a30 is busy.
One track worth of data, including the requested gator, is given to free Hatosuke as small data '6! - Load the scanned data for one track into the disk carriage device 17.

Below, data transmission system a30. The detailed functions of each data transfer system b31 will be explained according to the processing flow diagram. Fifth
The figure is a processing flow diagram of the data transfer system a30. The data transfer system a30 checks whether there is a completion notification from the disk device, whether it is when receiving a processing request from the channel (step 500) or when receiving a completion notification from the data transfer system b31 (step 501). Step 50
2) Check whether processing is required and start processing if necessary (step 503).

First, the processing performed when a processing request is received from a channel will be explained. In step 504, processing is classified. In the case of a sequential read hit or a read hit, the data on the disk cache device is transferred to the channel in step 505. In the case of a read hit, the processing can be ended at this point, but in the case of a sequential read hit, it is necessary to check whether there is a need to read ahead after this.

Therefore, in step 506, a check is made to see if it is a sequential read hit, and if so, in step 507, a check is made to see if there is a need for pre-reading. Step 5 if you need to read ahead
This process is performed in 0B. Specifically, empty blocks
The block 37 connected to the pointer 36 is taken out, necessary information is stored therein, and the block 37 linked from the interface register b33 is connected. (In this case, the data transfer system b31 receives the positioning completion notification.

) At this time, if a disk device in the corresponding disk device group 18 is vacant, the locator issues a request to the disk device, and the locator turns on the request flag 46.

If this is not the case, the request flag for positioning should be turned off. Also, the wait flag 47 is turned off.

Next, the case of write processing will be described. In the case of write processing, since it is necessary to access the disk device group 18, this preparation processing is performed in step 509. The specific content is almost the same as that described in step 508, and the only difference is that block 37 is connected to the block linked from interface register a32. (In this case, the data transfer system a30 receives the positioning completion notification.) Next, the case of read miss will be described. In this case as well, since access to the disk device group 18 is required, this process is performed in step 510. The processing content is almost the same as that described in step 508, except that the processor 37 is connected to the block linked to the common register 35. (In this case, the positioning will be notified by the data transfer system a30 when the positioning is completed.
If it is free, the data transfer system a30 accepts it,
If not, the data transfer system b31 accepts it.

)-1. In this case, the weight :j flag 47 is turned on.

Finally, let's talk about sequential read errors. In this case, since the block 37 for issuing a positioning request has already been used, only the wait flag 47 is turned on in step 511.

When you receive a completion notification from data transfer system b31,
If the read-ahead processing in the case of sequential read has been completed and there is a problem with this read-ahead, or
This is a case where loading of one track of data including the requested data into the disk carriage device 17 is completed at the time of a read or miss.

Therefore, in either case, step 512 forwards the requested data to the channel, and then step 5'l
3, block 37 is set to free block pointer 3
Post-processing is performed to connect to block 37 linked to block 6.

The data transfer system a30 receives a disk positioning completion notification in the case of write processing or read miss. At step 514, appropriate data transfer processing is performed. In the case of a write hit, the data on the disk device group 18 is updated, and in the case of a write miss, only the data on the disk device group 18 is updated. In the case of a read miss, the input request data is transferred to the channel and simultaneously transferred onto the disk carriage 17. Next, in step 515, the processing is classified, and in the case of write processing,
After that, in step 516, the block 37 corresponding to the completed write process is connected to the empty block pointer 36, and further, in step 517, the block 3
7, if the request flag 46 is off and there is a disk in the disk device group 18 to be accessed, the locater issues a request,
For positioning, post-processing of turning on the request flag 46 is performed in step 517. In case of read miss, step 5
] In step 8, a processing request is issued to the data transfer system b31 in order to load the data other than the input request data into the booteratisf cache.

Next, regarding the processing contents of data transfer system b31, ψ^9
1 phrase 〒 C hill 711ff ghee, niヱmutually q
1 θ) h1 does not appear 7 low diagram is shown. Data transfer system b
31 is when the processing for positioning the disk device group 18 is completed (step 600) and the processing is to be executed by the data transfer system b31 (step 601), or when a processing request is received from the data transfer system a30. The process is performed when the process is executed (step 602).

First, the positioning of the disk device group 18 will be described when a request is received. The data transfer system b31 receives a positioning request in the case of a sequential read prefetch process or a read miss. At step 603,
In the case of 0 sequential read for which this determination is made, data is transferred, so step 605 is performed.

In the case of a read miss, the step 604 checks whether the data transfer system a30 is busy. In the case of free
The data transfer system a30 performs processing first, so the process returns to step 602.Next, data is transferred in step 605.In the case of a read miss, one track's worth of data is transferred in the case of sequential read. loads only the predetermined amount of data into the disk cache device 17. Then, in step 606,
Processing similar to that performed in step 517 is performed, and if there is a disk device in the disk device group 18 that can issue a locating request, the locating request is issued. Next, in step 607, if the wait flag 46 is on, in step 608, the block 37 corresponding to this process is linked from the completion notification register to the linked block 37, and the block 37 is linked to the data transfer system a30. Issue a completion notice. If the wait flag is off, then in step 609 this block 37 is passed to the free block pointer 36.

Next, a case will be described in which a processing request is received from the data transfer system a30. In this case, in the event of a read miss, the request is to read 222 minutes of data remaining on the input request data to the disk carriage device 17, so this process is performed in step 610. After that, in step 611, the block 37 corresponding to this process is connected to the block 37 linked to the free block pointer 36, and in step 612, the same process as that in step 517 is performed, and the disk device group 18, if there is a disk device that can issue a request, the positioner will issue the request.

I explained the case of disk memory in detail above, but
As already stated, the idea of the present invention is applicable to all memories having cache memories.

〔Effect of the invention〕

According to the present invention, when recombining with a file device in the event of a read miss, there is no need to secure the device from the file to the channel. If a data transfer path between the caches could be secured, recombination became possible. Therefore, in the embodiment, even when the data transfer system a30 is performing read hit processing between the disk carriage and the channel, recombination is possible. 0, which is generally applicable to file devices with a cache memory and a J+A book.

[Brief explanation of drawings]

Fig. 1 is a block diagram showing the environment in which the director, which is the object of the present invention, is left alone; Fig. 2 is an explanatory diagram of new input/output commands provided in the present invention; Fig. 3 is a configuration diagram of the director device; Fig. 4 is a detailed explanatory diagram of the block 37 of the present invention, and FIG. 5 is a diagram showing the data transfer system a of the present invention.
30 processing flow diagram, FIG. 6 is the data transfer system b3 of the present invention.
'l processing flow diagram. Explanation of symbols 14 (15): Parallel transfer type director device, 20: New input/output command provided to effectively utilize the functions of the parallel transfer type director device. Kuzu 1 Figure 4 Figure 4 Graphic J7 Continuation of page 1 0 Inventor Akira Kurano Elementary School

Claims (1)

[Claims] 1. One or more director devices that manage to place data that is frequently referenced by the CPU on a main storage device, a Chiassy device, a file memory, a processing device, and a cache device. In an information system consisting of
In the input/output command passed from the processing device to the director device, when a predetermined unit of data in the file memory to be input/output does not exist in the cache, the predetermined unit of data is transferred to the cache. A parallel transfer storage device characterized by having means for loading into the device. 2. In the above director device, when data is transferred between the main storage device and the cache device, the positioning of the device receives a completion notification and transfers between the file device and the cache device.
2. The parallel transfer storage device according to claim 1, comprising means for transferring data between devices. 3. Configure the director device with two data transfer systems,
The first system has a function to transfer the data from the cache to the channel when input target data exists in the cache in input processing other than write processing and sequential read processing; If the data does not exist in the file device, when the positioning of the data on the file device is completed, if the first system is not transferring the data, it receives a notification of the completion and transfers the data to the carrier, the server, and the channel. After this, a function is assigned to the second system to load the remaining data of the predetermined unit in which the data exists into the cache, and the second system performs sequential read prefetch processing and other input processing. When the positioning of the input target data file is completed,
When the first system is ``G'', the second system receives the completion notification and is assigned a function of loading a predetermined unit of data to which the data belongs to the cache, and the first system is ``G''. The second system is characterized by having means for assigning to the second system a function of loading a predetermined unit of data other than the input target data into the cache when the second system receives the completion notification. 2. The parallel transfer storage device according to claim 1.