JP3419475B2 - Magnetic disk drive - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to emulation of a CKD disk device by an FBA disk device, and more particularly to a magnetic disk device suitable for random access, which is a characteristic of CKD.

[0002]

2. Description of the Related Art CKD tracks and FBA tracks will be described. The CKD track is shown in Fig. 3. There is one index marker 11 on the track,
The purpose is to recognize the beginning of the track. The home address (hereinafter abbreviated as "HA") 12 is the first field of each track and indicates the state and track number of that track. Records 0 to n exist after HA12. Each record includes a count part (hereinafter abbreviated as “C part”), a key part (hereinafter abbreviated as “K part”), and a data part (hereinafter abbreviated as “D part”).
The C parts 13 and 15 are the first field of each record and indicate the state, position, and length of the record. The K portion 16 is a field for recording information that serves as a guide for the subsequent D portion and used for searching the D portion and the like. Key length = 1 to C
It exists when 255 is specified, and when the key length = 0, the K part does not exist. The D sections 14 and 17 are fields for recording pure data. The value indicated by the data length of part C is D
It is the length of the part, and when the data length = 0, it represents the end of file. There is a gap (hereinafter abbreviated as “G”) separating the HA, C portion, K portion, and D portion. Record 0 (hereinafter abbreviated as "R0") is a record immediately after HA, and has a key length = 0 and a data length = 8 in the standard format used for system control. Records 1 to n are areas that the user can actually use.

Next, the FBA track will be described.
The FBA track is shown in Figure 2. The track is composed of an index marker 21 and a block of a fixed length. Each block has an ID section 22, 24 and a data section 2
3, 25, and a gap. The ID parts (position identification information parts) 22 and 24 record the block numbers and the like. The data sections 23 and 25 record pure data. The lengths of the data parts 23 and 25 are common to all blocks and are set to 512 bytes, for example.

A conventional technique for writing CKD data to FBA is described in Japanese Patent Laid-Open No. 60-45857. This simply writes the CKD data to the FBA drive in order. The above is for holding the data for backup, and is not intended to transfer the data with the upper channel device using this data. Further, performing the processing as CKD with the data stored by the above method causes a large overhead as compared with the CKD method, resulting in deterioration in performance. That is, the data for one track is read into the cache or the buffer in the control device, then converted into the CKD format, and the data is transferred to the host device, and the data transfer time for one track is always an overhead.

[0005]

In the above conventional technique, C
Since the data of KD is simply written to the FBA drive in order, it is difficult to search the target record having the data in a short time, and the process of CKD is performed by the drive in the random access operation. ,
Especially, the performance is significantly degraded.

The present invention relates to an FB that stores CKD data.
The purpose is to perform the search process in a short time by using the A drive.

[0007]

According to the present invention, in order to solve the above problems, in a fixed length format (FBA) magnetic disk drive in which each track is divided into blocks of a fixed length, each track is arbitrary. When emulating a variable-length format (CKD) magnetic disk device that is divided into records each having the same length, each block has information indicating whether or not there is a CKD record after the block. It is a thing.

[0008]

In a fixed length format (FBA) magnetic disk drive in which each track is divided into blocks of a fixed length, a variable length format (CKD) magnetic disk in which each track is divided into records of an arbitrary length. When emulating the device, each block has information indicating whether or not there is a CKD record after the block. Therefore, the CKD record can be searched at high speed.

[0009]

Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 8 shows a system using this magnetic disk device. The host device comprises a central processing unit 181 and a channel device 182. A cache memory (hereinafter abbreviated as “cache”) 91 is installed in the control device 90. The disk device 92 exists at the lowest level. The disk device uses FBA to store data as shown in FIG.
Further, the cache 91 is equipped with a battery 93, and the cache 91 continues to store the previous data even after the power is turned off. The control device 90 temporarily holds the transfer data in the cache 91 and sends it to the channel device 182 or the disk device 92.

Next, the emulation method will be described. The case where CKD is emulated on FBA is shown in FIG. Figure 1 CKD data in the upper row is the FBA in the lower row
At the time of recording, the data indicating the position of the next C portion (hereinafter abbreviated as “CP”) is written at the beginning of each block, and then the record is recorded. Records are part C, K
Section and D section are recorded continuously, and then G between fields is recorded.
Write. As a result, C section A33 goes to CA42, K section A34 goes to KA43, D section A35 goes to DA44, and GA3 goes to
8~GC40 total to the sum of the GA '45 and GB' 46 of, to respond like that. At this time, CPA49 is CA4
2 shows the position, and CPB50 shows the position of CB47. Since G is also reflected on the FBA, the length of the record on the CKD and the length of the portion corresponding to the record on the FBA are the same.

If record 0 is a standard format, record 1 is recorded from the beginning of the block group corresponding to the CKD1 track. The normal record 0 has a standard format, and several tracks or several cylinders are recorded in one block together with HA. In terms of capacity, emulation is performed by allocating an FBAn block (30 Kbytes for one track of 60 Kbytes and one block of 2 Kbytes) to one CKD track.
The n block becomes a logical CKD1 track.

A concrete example is shown in FIG. As an example of a CKD drive, the number of heads per cylinder is 15
When 30 blocks of tracks are applied, there are 450 blocks (0 to 449) per cylinder of CKD format. The bit map indicates that record 0 is a standard format. One bit corresponds to one track, and it is written at the rear end of cylinder 0 and head 14 of the FBA drive, and from the beginning of the following cylinder 1 and head 0. Write HA and record 0 for several tracks or for several cylinders. Although the area up to the rear end of the cylinder 2 is prepared as an area for storing HA and record 0, it is not always used up to the rear end of the cylinder depending on the number of tracks of the CKD drive. User data (record 1 and later of the CKD track) is stored from the cylinder 3. The reason for selecting the storage location is as follows. The head of the track storing HA and record 0 and the head of the track storing user data are easier to locate at the head of the cylinder (head 0). This is because the correction is easy in the SEEK process. Further, as will be described later, since the bitmap is made resident in the cache and the HA and the record 0 are stored at a position different from the user data, if the above two can be read collectively, the processing becomes easy. Therefore, the bit map is recorded at the rear end of the cylinder 0 so as to have continuity with the track for storing HA and record 0. The unused portion of the cylinder 0 can be used for multiple purposes by providing a spare for control.

Next, an instruction (hereinafter, abbreviated as "command") from a host device in CKD will be described.
Normally, the command shown in FIG. 4 is issued. The SEEK command 60 (hereinafter abbreviated as "SK") is the cylinder number,
It is used to send the head number and select one track. The SET SECTOR command 61 (hereinafter abbreviated as "SS") is used to position the sector number to the target sector. The tracks in each cylinder of the disk are divided into fan shapes with equal spaces, and each sector is a sector. SEARCH ID command 62 (hereinafter "SC
It is abbreviated as "H". ) Is used for positioning the head to the target record, and compares the sent data with the data of the C section read from the disk. TIC command 63
SCH6 until the data match on the previous SCH62
It is an instruction to repeat 2. A READ or WRITE command 64 is issued after positioning to the target record.

Next, the operation when the above command is received after emulation will be described. At the time of SK60, it is possible to easily obtain the block corresponding to the track. At SS61, it is necessary to convert the sector number into a block number. The number of sectors on the CKD track is α, FB
If the number of blocks used in the A track is β, the head block of the track is 0, and the sector number of the data sent from the higher-level device is γ, the target block number can be obtained from equation 1.

[0016]

[Equation 1]

Next, the SCH 62 and the TIC 63 locate the target record. C part n of the block as shown in FIG.
When searching 71, the CP 70 sent from the block to the control device indicates the position of the C part n71 (for example, how many bytes after itself), so the control device can easily recognize and compare the C part n71. Further, even when searching the C part n + 1 (72), the C length is determined by the key length and the data length of the C part n71.
The position of the part n + 1 (72) can be easily calculated. If there is no CP, the controller needs to recognize part C by some means. For example, the tracks are sequentially recognized from the beginning. As described above, the SCH operation can be performed in almost the same time as CKD.

If no record exists after this block, information indicating that (eg, a zero pattern) is recorded in the CP, so that the path with the host device can be released until the index marker, which is a waste. It is possible to avoid waiting for a proper rotation.

The same applies to the SCH processing and the READ processing, but when the block does not have a CP, it is necessary to read the data corresponding to the C portion from the beginning of the block corresponding to the track. When the target data exists in the block n, if there is no CP, the data is read from the block 0. Therefore, the time required to transfer the data of n blocks becomes an overhead. The previous SCT is due to the existence of CP
Processing becomes possible, and the SCH / READ processing can be performed from the completion of reading in the target block. Therefore, depending on the presence or absence of CP, a difference in overhead of about 1 track at maximum occurs, and the amount of data read to the cache is reduced to only the target block.

Usually, the HA and the record 0 are hardly updated, and the record 0 is in the standard format. Since record 1 and the subsequent areas are usable by the user, there is a high probability that the data will be updated. Therefore, HA, record 0 and record 1
The processing speed is faster if the subsequent blocks are not stored in the same block. Assuming that the block is composed of only HA, record 0, and record 1, even if only record 1 is updated, the block must be read into the cache once, and record 1 must be updated and written to the block. Therefore, HA, record 0, record 1 and subsequent records are stored in separate blocks. Then, although the recording method as shown in FIG. 6 is conceivable, in most cases, the record 0 (82) has a standard format, and therefore, an unused area 83 is generated in the rear of the block 0 (80). Therefore, the HA 81 and the record 0 (82) are collectively stored in different blocks for several tracks or several cylinders. As a result, the track shown in FIG. 6 can be a pure data track as shown in FIG. 7, and the unused area 83 generated in FIG. 6 can be reduced. If the record 0 is not in the standard format, data exceeding the standard format (9th byte and later including K and D sections) is stored from the block 0 of the pure data track.

Further, when tracks (block groups) as shown in FIG. 6 exist continuously, writing is performed up to the end of a certain track, and when writing is also made continuously from record 1 of the next track, the next track is written. The writing process must be temporarily stopped at the place where HA and record 0 exist. As a characteristic of the FBA drive of the SCSI interface, it is impossible to resume the write process one block after the process is once interrupted. Therefore, waiting for rotation occurs. If the track is as shown in FIG. 7, the writing process can be continuously performed on the next track. W
At the time of RITE processing, the completion report is immediately issued to the upper device immediately after the data is input to the cache, and then written to the disk device. Therefore, when the disk device is used or the WRITE process is continuously performed, a state in which continuous tracks are collectively written occurs. For this reason, the track format of FIG. 7 is more advantageous.

Further, the access performance can be improved by storing the format state of the track in a specific block on the assumption that the cache memory is used.

Two examples will be described below.

It has information indicating whether or not the record 1 and subsequent records of the track exist. As a result, the erasing operation of the track (erasing the records after 1) is only set or reset of the information, and the processing time can be shortened.

Since HA and record 0 are collectively held in another block, frequent reading will reduce the system throughput. In normal processing, HA is rarely read, and record 0 is mostly read because record 1 is updated. At this time, if the record 0 is in the standard format, the record 1 exists at the head of the block group corresponding to the track, so if the record 0 is in the standard format, information indicating this (for example, bitmap)
If it is stored in the cache, the position of the record 1 can be found only by referring to the information, and the position of the record 1 can be directly located, so that the processing time can be shortened.

FIG. 9 shows the time relationship between the data on the drive, the data on the cache, and the data transferred to the host device during the READ processing. The disk device converts the sector number into a block number by SS processing and performs positioning operation. After reconnecting with the control device in the target block (block N100), the final block (block N + n101) for data transfer is reached. Transfer data. When the control device refers to the CP of the block and recognizes the target field 102, it reconnects to the host device and starts data transfer.
Data equivalent to G is not transferred to the host device. When the final field 103 is transferred, the data transfer ends.

Therefore, the records can be positioned with a delay of the data transfer time of one block at the maximum, and the access time close to that of the CKD disk can be obtained.

Even when the CKD1 track is distributed to a plurality of FBA drives (array disk), this method can be performed and the above-mentioned characteristics can be obtained.

As described above, according to this embodiment, F
Since the head of each block of BA has data indicating the head position of the next CKD record, it is easy to execute the CKD command (particularly, the record search process).

Further, since there is information indicating that there is no record after this block, it is possible to release the path with the host device from after this block to the beginning of the next track, improving the system throughput. be able to.

Further, since the CKD record 1 is stored from the beginning of the FBA block, positioning and retrieval are easy and the speed is high.

Further, since the track information has the information indicating whether or not the record 1 or later of the track exists, the data erasing of the track (erasing the record 1 or later) only updates the information, and the processing time is shortened. Can be made.

[0033]

According to the present invention, the retrieval process can be performed in a short time by using the FBA drive storing the CKD data.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of a method for emulating CKD with FBA.

FIG. 2 is an explanatory diagram of a track format of FBA.

FIG. 3 is an explanatory diagram of a CKD track format.

FIG. 4 is an explanatory diagram of commands.

FIG. 5 is an explanatory diagram of an FBA track format after emulation.

FIG. 6 is an explanatory diagram of a track format of FBA after emulation.

FIG. 7 is an explanatory diagram of an FBA track format after emulation.

FIG. 8 is a system configuration diagram.

FIG. 9 is an explanatory diagram showing data transfer between devices.

FIG. 10 is an explanatory diagram of a track format when emulated by an FBA drive.

[Explanation of symbols]

11,21 Index marker 12 HA 13,15,33,36 C section 16,34,37 K section 14,17,35 D section 22,24 ID section 23,25 Data section 38-41 G 49,50,70 CP 60-64 Command 42-48, 71, 72, 82-84 Data in block 90 Disk control device 91 Cache memory 92 Disk device 93 Battery 100, 101 Block 102, 103 Start point and end point of data transferred to upper device

─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Kuji Takeuchi 2880, Kozu, Odawara, Kanagawa, Ltd., Odawara Plant, Hitachi, Ltd. (72) Yoshihiro Uchiyama, 2880, Kozu, Odawara, Kanagawa, Ltd., Hitachi, Ltd., Odawara Plant (72) 72) Inventor, Yoshiko Matsumoto, 2880, Kozu, Odawara, Kanagawa, Ltd., inside the Odawara Plant, Hitachi, Ltd. (72) Inventor, Hisayoshi Kawabata, 2880, Kozu, Odawara, Kanagawa, Ltd. (56) Reference, Odawara, Hitachi, Ltd. -306917 (JP, A) JP-A-3-5965 (JP, A) JP-A-60-74018 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) G11B 20/10- 20/16 G06F 3/06-3/08

Claims (2)

(57) [Claims] 1. Each has a home address and
Record of arbitrary length including record 0 and record 1
Multiple variable length format (CKD) files divided into
Each of the data recorded in the rack has a fixed length.
Fixed-length format (FBA) divided into blocks
A magnetic disk device for recording in a rack , wherein when writing the data , data of the plurality of CKD tracks is written.
The home address and the record 0 of
It is different from the block that records the records after the record 1.
Recording in at least one of the blocks.
Magnetic disk device to be used . 2. Connected to a host device having a channel device
Yes and a control device that is connected to the control device, each of the home address
Any record containing record 0 and record 1
Multiple variable-length formats divided into length records
Each of the data recorded on the (CKD) track
Fixed-length format divided into blocks of fixed length
(FBA) A disk device for recording on a track and a channel device or a front device provided in the control device.
A cache that temporarily holds the transfer data to the disk device.
The disk device, when writing the data,
The home address and the record for several CKD tracks
The record 0 and the records after the record 1 are recorded.
At least one block different from the block
A disk system characterized by recording on a disk .