JPH11110138A - Storage controller - Google Patents

Abstract

(57) [Summary] [PROBLEMS] To reduce the response time at the time of multiple sequential access. A CPU of a storage control device provided between a plurality of host computers and a disk storage device.
After changing the request length of the data requested by the request issued from the host computer 90, the module 190 takes precedence over the request to the disk storage device 399 without depending on the request sent from the host computer 90 thereafter. Issue.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a storage device of a computer system, and more particularly to a technique for improving a transfer speed in a system in which multiple sequential accesses (read / write) occur.

[0002]

2. Description of the Related Art First, the configuration of a conventional general disk storage device will be described.

[0003] As shown in FIG.
9 includes a rotating disk-type storage medium 300 and a head 310 for reading and writing a recording signal moving in the radial direction on the storage medium 300. The position of the head 310 is controlled by a head control unit 315, and the rotation of the disk 300 is controlled by a motor control unit 325 and a motor 320. The stored signal read from the disk 300 is subjected to waveform shaping by the waveform shaping unit 350 and then to the ECC unit 352.
Are subjected to error detection / correction, temporarily stored in the buffer 354, and then transferred to an external host computer by the transfer means 356. The operation of each unit is controlled by the controller 390.

In such a disk storage device 399, the recording surface of the storage medium 300 is concentrically divided into a plurality of areas called tracks 301 as shown in FIG. The track 301 is further divided into units called sectors 302.
4, a position information area 303 for storing information indicating the position of the sector, a data area 304 for storing data, and an ECC for storing an ECC code for error detection / correction, as shown in FIG.
An area 305 and the like are provided. Here, the disk storage device manages data in 399 sector units (often 512 bytes).

Here, reading of a certain sector is performed as follows.

[0006] First, the head is moved to a track including the target sector, waits until the target sector comes under the head, and data is read out serially in accordance with the rotation of the disk. Then, to the serial data read from the disk,
After performing processes such as waveform shaping and error detection / correction, the data is temporarily stored in a buffer as parallel data. When the amount of data in the buffer reaches a predetermined amount, SC
Data is transferred to the host computer via an interface with the host computer, such as an SI (Small Computer System Interface), Fiber Channel, or IDE interface. Writing is performed in the same manner by the reverse procedure.

Here, the time required for reading data from the disk storage device (or writing data to the disk storage device) is as follows: (1) Time (seek time) until the head moves to a target track. 2) Waiting time until the target sector comes under the head (rotational waiting time) (3) It is determined by data reading / writing time (data transfer time). The time from when the host requests data read, after the seek time, the rotation wait time, and the data transfer time to when the end of read / write is reported is called the response time.

Thus, the disk storage device can specify one specific sector by specifying the track and the sector. In a disk storage device having a plurality of heads, a specific one sector is specified by further specifying the head. In such a case, data can be read from or written to a target sector by designating it as sector C of track B of head A.

On the other hand, as a method of specifying a sector, there is a method of assigning a unique number consecutive to the sector 0 of the track 0 of the head 0 in order from the sector and managing the sector, and specifying the sector by this number. This number is called a logical block address (LBA). For example, in the case of a disk storage device having two heads, 1024 tracks, and 128 sectors, the logical block addresses are from LBA0 to LBA2.
62143 (= 2 × 1024 × 128). In this case, the sector is called a block. When reading and writing using the logical block address, the logical block address and
By specifying the number of blocks, reading and writing of data in continuous sectors is performed. This is a common method in a disk storage device employing SCSI.

Next, an operation at the time of sequential access (read / write to a continuous logical block) performed in such a conventional disk storage device will be described.

[0011] The sequential access is, for example, L
128 blocks from BA0, 1 from block 128
28 blocks, 128 blocks from block 256,. . . .., Etc., when the request length is added to the nth requested position, the access sequence exactly matches the (n + 1) th requested position.

Here, the following sequential access (in the case of reading) is assumed. The request position is represented by a logical block address, and the request length is represented by the number of blocks.

[0013] In this case, the operation of the disk storage device is as follows.

The host computer that performs the sequential access first issues a first request. When the disk storage device receives the first request, it moves the head to the target track (seek time) and waits for the target sector to come under the head (rotation wait time) to start reading. When reading of the requested 128 blocks is completed (data transfer time), an end report is sent to the host computer. Upon receiving the end report, the host computer issues the next second request.

When the disk storage device receives the second request, it moves the head to the target track (seek time) and waits until the target sector comes under the head (rotation wait time) to start reading. . When reading of the requested 128 blocks is completed (data transfer time), an end report is sent to the host computer. The same processing is performed after the third time.

In the case of such a sequential access, tracks and sectors to be requested successively are basically continuous. Therefore, in the case of the sequential access, the seek time and the rotation time are shorter than in other cases.

By the way, the first request is completed, a completion report is sent to the host computer, and the host computer issues the second request, and by the time the second request is performed (from the first completion to 2 It usually takes several hundred microseconds to several milliseconds until the second request is made. During this time, the rotation of the disk continues to rotate at a constant speed. At the time of the start of the second request, the target sector is not under the head, and once again the target sector turns under the head. have to wait. In the worst case, one rotation waiting time occurs.

Therefore, in order to reduce the useless rotation waiting time, a recent disk storage device performs a process called a prefetch (prefetch) process.

When performing prefetch (prefetch) processing,
When the first request is completed, the disk storage device continues to read continuous sectors (and tracks) and holds the read data in a buffer while performing processing of returning an end report to the host computer. As a result of this processing, when the second request is made from the host computer, the target sector has already passed under the head, but since the data of the target sector is read and stored in the buffer, the data is It can be transferred to the host computer. Thereby, the sequential access can be continued without waiting for the target sector to come under the head.

[0020]

Here, a case of two multiplex sequential accesses in which two host computers execute sequential accesses respectively will be considered.

Now, assume the following sequential access (in the case of reading). The request position is represented by a logical block address, and the request length is represented by the number of blocks.

The first host computer A is: The second host computer B performs sequential access at Performs sequential access.

In this case, the host computer A that performs the sequential access first issues a first request. The disk storage device is one of the host computers A
When the second request is received, the head is moved to the target track (seek time), and reading is started after the target sector comes under the head (rotation wait time). When reading of the requested 128 blocks is completed (data transfer time), a completion report is sent to the host computer, and the above-described prefetch is performed. Upon receiving the end report, the host computer issues the next second request.

Here, before the host computer A issues the second request, the host computer B issues the first request. Upon receiving the first request from the host computer B, the disk storage device stops prefetching based on the first request from the host computer A, and moves the head to the target track of the first request from the host computer B (seeking). Time), the reading is started after the target sector comes under the head (rotation waiting time). When reading of the requested 128 blocks is completed (data transfer time), an end report is sent to the host computer. After the second time, the same processing is performed. That is, the sequential access requests of the host computer A and the host computer B are mixed. The same applies to the case where the number of host computers becomes n and the number of host computers becomes n multiplex sequential.

The difference between a single sequential access by a single host computer and a multiple sequential access by a plurality of host computers is that the requests from each host computer are mixed because the positions of the sequential accesses of the plurality of host computers are different. Then, a seek time until the head is moved to the target track and a rotation waiting time until the target sector comes under the head take a long time, resulting in a poor response time.

In other words, in the case of single sequential access, since requests are continuous, the continuity of sectors (or tracks) could be expected. Since the continuity cannot be expected, the sequential access speed (transfer speed) to each host computer is greatly reduced. In addition, when there is no continuity of the sector (or track continuity) for two subsequent requests,
The prefetch based on the previous instruction is also stopped by the next request, and does not function sufficiently effectively. As described above, there is also known a system in which, when the prefetch is stopped, data read by the prefetch is discarded before the stop.

Therefore, an object of the present invention is to reduce the response time to each host computer in multiple sequential access.

[0028]

In order to achieve the above object,
The present invention is a storage control device that is connected between a plurality of host devices and a storage device and issues a read request to the storage device based on a read request issued from the host device for data stored in the storage device. A data holding unit that functions as a cache of the storage device for the plurality of host devices; and a data length change unit that changes a data length of data targeted by a read request issued from the host device to a longer length. And at least a part of the data that is not stored when all of the data targeted by the read request whose data length has been changed by the data length changing means is not held in the data holding means. A read request is issued from the host device after the read request in which the data length changing means changes the data length. Request issuing means for issuing data to the storage device prior to a read request and reading data from the storage device, wherein the data holding means responds to the read request issued by the request issuing means and stores the data in the storage device. Provided is a storage control device that holds data transferred from a device.

According to the storage control device of the present invention, the prefetch processing based on the read request from the host device is performed regardless of the block to which the subsequent read request is made from the host device. Prior to reading based on Then, the prefetched data is cached in advance.

Therefore, it is expected that the average response time at the time of multiple sequential access can be reduced.

[0031]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below.

First, a first embodiment of the present invention will be described.

FIG. 1 shows the configuration of a computer system according to the first embodiment.

As shown, the computer system includes a plurality of host computers 90, a storage controller 1,
It comprises a disk storage device 399.

The storage control device 1 includes a host computer 9
SCSI interface controller 100 for connecting to a disk storage device, and a SCSI interface controller 100 for connecting to a disk storage device.
An I interface controller 110, a memory 150 for holding data sent from a host computer and data read from a disk storage device, and a CPU module 190 are provided.

The CPU module includes a control processor (CPU) having a primary cache, a memory for storing programs and various tables to be described later, a secondary cache memory for accelerating processing, and a PCI bus. It is composed of a bus converter and the like.

In the storage control device 1, a two-level PCI
We adopt bus. PCI bus 120 and PCI bus 1
30. SCSI interface controller 10
0 and the CPU module 190 to the PCI bus 120, and connect the SCSI interface controller 110 to the PC
It is connected to the I bus 130. PCI bus 120 and PC
A bridge 160 and a memory 150 are connected between the I buses 130. The CPU module 190 can control the SCSI interface controller 110 connected to the PCI bus 130 from the PCI bus 120 via the bridge 160. In the present storage control device 1, the memory 150 is divided into two PCI buses 120 and a PCI bus 130.
Connected to both. Since the PCI bus 120 and the PCI bus 130 form a hierarchy by a bridge,
SCSI interface controller 100 and memory 1
Data transfer between the SCSI interface controller 110 and the memory 150 can be performed at the same time.

Here, two host computers 90 are connected to the SCSI interface controller 100 in a daisy chain system (daisy chain system). Also, SC
The disk storage device 399 is connected to the SI interface controller 110.

Next, the configuration of the disk storage device 399 is shown in FIG.

The disk storage device 399 includes a rotating recording medium 300, a motor 320 for rotating the recording medium,
Motor control means 325 for keeping the number of rotations of the motor constant;
Head 310 movable on recording medium 300 in radial direction
A head control unit 315 for positioning the head 310 on the recording medium 300; a waveform shaping unit 350 for shaping a signal read from the head 310; an ECC unit 352 for checking the validity of the read data; , A transfer unit 356 for transferring data in the buffer, and a controller 390 for controlling each unit.

Next, FIG. 3 shows a data storage method on the storage medium 300 in the disk storage device 399.

Data is held by a concentric track 301 formed on the recording medium 300 and the track 301.
Are performed in units of sectors 302 created by dividing Generally, a disk storage device that holds 512 bytes of data per sector 302 is common.

For example, 512 bytes of data can be stored per sector, and the number of sectors per track is 128.
When the number of tracks and the number of tracks are 1024, 64 MB (= 512 bytes × 128 × 1024) of data can be held on one side of the recording medium. Therefore, 1 on both sides of the recording medium
28MB. If five recording media (ten heads) are used, 640 MB of data can be held. The total number of sectors is 1310720 (= 128 × 1024 × 1)
0). In the case of this disk storage device, head numbers 0 to 9, track numbers 0 to 1023, and sector numbers 0
If 127 is given, only one sector can be specified.

When reading data from the disk storage device 399 or writing data to the disk storage device 399, a target sector is designated by a head number, a track number, and a sector number. In the case of a SCSI-compatible disk storage device, a method is used in which each sector is assigned a unique number called a logical block address and managed. For example, in the case of the 640 MB disk storage device shown in the above example, a logical block address from 0 to 1310719 is assigned to each sector. In this case, when reading data from the disk storage device 399 or writing data to the disk storage device 399, a target sector is specified by a logical block address. Here, as shown in FIG. Area 303, data area 304, and ECC area 3
05. In the position information area 303, a track number and a sector number are recorded. Thus, the information in the position information area 303 read by the head can be checked, and it can be confirmed that the head has correctly moved to the target track and that the target sector has reached the position of the head. The data area 304 holds data. E
The CC area 305 holds ECC information calculated from the data held in the data area. At the time of writing, when data is held in the data area 304, an ECC (Error Co
rrecting Code) is calculated and stored in the ECC area 305. At the time of reading, data is read from the data area 304, ECC is read from the ECC area 305, and it is checked whether the read data is correct. If the ECC for the data does not match, the error of the data can be corrected to some extent by the ability of the ECC. EC
The correction capability of C differs depending on the type of ECC to be added.
If the data cannot be corrected by ECC, an error occurs and the sector cannot be read. This is called "sector failure".

The cause of the sector failure is a scratch on the recording medium, deterioration of the magnetic material, unevenness of the magnetic material, and the like.

Next, FIG. 5 shows the configuration of a control program stored in a memory in the CPU module 190 and executed by the CPU module 190 in FIG.

The control program 200 includes a request receiving module 210 for receiving a request from the host computer.
(And a request conversion module 220 for converting a request from the host computer and a request issuing module 240 for issuing the converted request to the disk storage device).

Next, various tables provided in the memory in the CPU module 190 will be described.

As shown in FIG. 6, the CPU module 19
The request management table 2 that holds the request
50 and a held data management table 261 for managing data held in the data holding area.

In the memory 150, a data holding area 260 for holding data read from the disk storage device and data sent from the host computer is provided.

The data holding area 260 is a disk cache memory of the disk storage device 299. A predetermined amount of data selected in accordance with a predetermined algorithm and previously accessed by the host computer is held.

FIG. 7 shows the structure of the request management table 250.

The request management table 250 is a table for holding request positions and request lengths. The request position is held by the logical block address, and the request length is held by the number of blocks.

The request held in the request management table is a request from the host computer received by the request receiving module 210. Each time a request is received, it is stored from the first in the request management table. In the case of FIG. 7, if the request is retained up to 100, the oldest request is discarded,
Hold the 101st request. Thereafter, the request is similarly held.

FIG. 8 shows the configuration of the held data management table 261.

The held data management table 261 is a table for holding a logical block address (LBA) of data held in the data holding area 260 and the number of blocks, and is used as a disk cache management table. Each table element is composed of three elements: a logical block address, the number of blocks, and a pointer. The pointer holds a pointer to an address at which data in the data holding area 260 is stored.

By referring to the held data management table 261, it can be easily confirmed whether or not the target data is in the data holding area 260.

Hereinafter, the data read operation of such a computer system will be described.

When the host computer 90 reads data from the disk storage device 399, the storage control device 1 is connected between the host computer 90 and the storage device 399 as shown in FIG. The storage control device 1 receives a data read request from the storage device 399 issued by the storage device 90.

A read request from the host computer is issued by a command block called CDB based on the SCSI standard.

SCSI interface controller 10
0 receives the CDB from the host computer,
An interrupt signal is sent to the PU module 190. Upon receiving the interrupt signal, the CPU module activates the request receiving module.

The request receiving module extracts the CDB from the SCSI interface controller 100, generates a packet holding the request position and the request length, and passes the packet to the request conversion module.

The request conversion module refers to the request management table 250 based on the received packet, and addresses (block numbers) of the block targeted by the request of the packet and the block requested by the packet received in the past. Is determined. For example, if a request (request position 140 and request length 10) as shown in FIG. 9B is passed when the request management table has the contents shown in FIG. Referring to the table 250, the block requested by the current packet is continuous with the target block in the request of the past packet in the order of the entry numbers 1 → 3 → 5 → 6. Can be determined. Also, the number of hits in the request management table 250 for past requests (the number of entries in the request management table in which past requests for blocks that are continuous with the block requested by the current packet are registered, in this case, 4) And the total number of blocks in the past (the total number of blocks targeted by past requests for blocks that are continuous with the block requested by the current packet, in this case, 40).

The request conversion module changes the current request length based on the two pieces of information obtained in this way, the number of hits and the total number of blocks.

Here, the change of the request length is changed according to the following rule. By dividing the total number of blocks by the number of hits, it is possible to determine the average required length per time (this is called the average required length). Here, a value obtained by multiplying the average required length by a constant K is set as a new required length. This constant K
Is changed, the transfer speed at the time of multiple sequential access changes. For example, if the constant K is 5, the requested length is 50 in this case.

Next, referring to the held data management table 261, the requested position requested by the host computer,
Data based on the changed request length is stored in the data holding area 26.
Check if it is at zero. As a result of the check, if all the data exists in the data holding area 260, the request data of the host computer is transferred from the data holding area 260 without reading the data from the disk storage device 399. In addition, when there is a partial presence (when there is no 50% or more) or when there is no such portion, the request position and the request length are changed so as to be a request for reading the missing portion. , Generates a new packet, and passes the packet to the request issuing module 240.

The request issuing module 240 creates a CDB according to the SCSI standard based on the received packet, and issues the CDB to the disk storage device 399. When the disk storage device 399 receives the CDB, the
B is interpreted, data is read, and transfer to the storage controller 1 is started.

The storage control device 1 transmits data transferred from the disk storage device to the SCSI controller 110
The data is held in the data holding area 260 in the memory 150 via the CI bus 130. The request issuing module 240
When the reading process from the disk storage device 399 is completed, the held data management table 261 is updated so as to match the data held in the data holding area 260.

The request receiving module, based on a request from the host computer (a request that is converted by the request conversion module), stores the data in the data holding area 2 in the memory 150.
The SCSI interface controller 100 transfers the 60 data to the host computer 90 via the PCI bus 120. When the data transfer is completed, the SCSI interface controller 100 reports a read completion report to the host computer 90. with this,
The read processing ends.

The request receiving module rejects the reception of the next read request from the host computer until the process of reading one read request sent from the host computer is completed. Alternatively, the next read request is accepted, but the read processing is always performed after the read processing of the previous read request. The request is read out and processed when it becomes possible to start the request reading process. That is, the read processing of the read request received earlier from the host computer is preferentially performed.

When a write request is issued from the host computer, the requested writing is performed on the data holding area, and the held data management table is updated so as to be consistent therewith. When the data holding area becomes full, the data to be written back to the disk storage device is determined by a predetermined writing back algorithm, the data is written back to the disk storage device, and the held data management table is updated to match the data. .

By the above processing, when multiple sequential accesses occur, the seek time and rotation wait time generated in the disk storage device can be reduced, and as a result, the speed of the sequential access processing can be increased. This effect will be described with reference to FIG.

Assume that two host computers A and B are connected as shown in FIG. 1 and two sequential accesses occur. 30m seek time between each data
s, the average rotation waiting time is 4 ms, and the transfer time per block is 0.05 ms (10 MB / s). When each host computer performs sequential access of 128 blocks, the data transfer time is 6.4.
ms.

In the case of the single sequential access, as shown in FIG. 10A, only the data transfer can be performed with almost no seek time and no rotation waiting time. In this case, almost 100% efficiency can be achieved.

When the requests from the two host computers are issued to the disk storage device as they are, the requests from the two host computers alternately arrive on average, as shown in FIG. Seek occurs. For this reason, the ratio at which the substantial transfer is performed is obtained by 6.4３０ (30 + 4 + 6.4)-(Equation 1), and is 15.84%. Furthermore, because of the two-multiplex sequential access, the transfer rate per host computer is halved to 7.92%. this is,
This shows that even when a disk storage device having a single sequential access performance of 10 MB / s is used, only a transfer rate of 0.792 MB / s can be realized in the case of two-multiplex sequential access.

To improve the transfer speed at the time of multiple sequential access, it can be solved by increasing the ratio of the data transfer time to the seek time and the rotation wait time in the above (Equation 1). For example, when reading is performed for every 640 blocks, which is 5 times, the result is as shown in FIG. Assuming two-multiple sequential access, the transfer rate per host computer is halved to 24.24%. in this way,
The data transfer speed can be improved by 3.06 times.

Of course, this is about the transfer from the disk storage device to the storage controller 1, and not from the disk storage device to the host computer.

Not all data transferred from the disk storage device to the storage controller 1 is necessarily used by the host computer. By changing the request length, the host computer requests the storage controller 1 for data read ahead. May not be issued after all. However, in the present embodiment, by changing the request length in accordance with the strength of continuity of a target block between a request and a past request, it is possible that the request is one of requests during sequential access. The corresponding block number is read ahead. Therefore, it can be expected that the host computer issues a subsequent request for the pre-read block during subsequent sequential access including the request, thereby reducing the possibility of wasting the pre-read block.

FIG. 11 shows assumptions (30 ms, 4 ms, 0.
The transfer efficiency when the number of blocks to be read with respect to one seek time and the rotation wait time is changed under the condition of (05 ms, 2 multiplexes) is shown in a graph. As the number of blocks increases, the transfer efficiency increases.

As described above, according to the present embodiment, the sequential access is detected by comparing the request from the host computer with the past request, and the request length is changed in accordance with the sequential access. By increasing the ratio of the data transfer time of the data requested or expected by the host computer to the time and the rotation waiting time, it is possible to reduce the average response time at the time of the multiple sequential access. .

Hereinafter, a second embodiment of the present invention will be described.

The second embodiment is different from the first embodiment in that
The request management table 250 is omitted, and the rule for changing the request length performed by the request conversion module 220 is changed.

That is, in the present embodiment, the request change module 220 uses the packet received from the request
Referring to 61, it is checked whether or not the data requested by the host computer exists in the data holding area 260.

When the target data exists, the request length is not changed, and the request receiving module 220
The SCSI interface controller 100 transfers the data in the data holding area 260 in the memory 150 to the host computer 90 via the PCI bus 120 based on a request from the host computer (a request that is converted by the request conversion module). .

On the other hand, if the target data does not exist, the request length is changed. The required length is replaced with a constant value J. Here, J is assumed to be 2048. The request conversion module changes the request length to 2048, generates a packet, and passes the packet to the request issuing module 240.

The request issuing module 240 creates a SCSI CDB in accordance with the received packet and issues the CDB to the disk storage device 399. Upon receiving the CDB, the disk storage device 399 interprets the CDB, reads data, and starts transfer. The storage control device 1
Data from the disk storage device is held in the data holding area 260 in the memory 150 via the SCSI controller 110 and the PCI bus 130. When the reading process from the disk storage device 399 ends, the request issuing module updates the held data management table 261.

The request receiving module, based on a request from the host computer (a request that is converted by the request conversion module), stores the data in the data holding area 2 in the memory 150.
The SCSI interface controller 100 transfers the 60 data to the host computer 90 via the PCI bus 120. When the data transfer is completed, the SCSI interface controller 100 reports a read completion report to the host computer 90. with this,
The read processing ends.

As in the first embodiment, the request receiving module receives the next read request from the host computer until the read processing of one read request sent from the host computer is completed. Reject. Alternatively, the next read request is accepted, but the read processing is always performed after the read processing of the previous read request. The request is read out and processed when it becomes possible to start the request reading process. That is, the read processing of the read request received earlier from the host computer is preferentially performed.

According to the second embodiment, the prefetch processing based on the read request from the host computer is performed based on the subsequent read request irrespective of the block to which the subsequent read request is made from the host computer. This is performed prior to reading. Then, the pre-read data is stored in the data holding area, and the data is transferred from the data holding area to the host computer in response to a subsequent read request for the data.

Therefore, it is expected that the average response time at the time of multiple sequential access can be shortened.

Hereinafter, a third embodiment of the present invention will be described.

The third embodiment differs from the first embodiment in that the request management table 250 is omitted, and the held data management table 261 shown in FIG.
62, and the request conversion module 220 converts the request length in accordance with the held data management table 262.

As shown in FIG. 14, the held data management table 262 is a table for holding a logical block address (LBA) of data held in the data holding area 260 and the number of blocks. Elements of each table are logical block address, number of blocks, expected value, and pointer.
It consists of one. The pointer holds a pointer to an address at which data in the data holding area 260 corresponding to the LBA is stored. The expected value will be described later.

Hereinafter, a process when the host computer 90 reads data from the disk storage device 399 will be described.

As shown in FIG. 1, since the storage control device 1 is connected between the host computer 90 and the storage device 399, a data read request issued by the host computer 90 to the storage device 399 is not stored in the main storage device. The control device 1 receives it. The read request from the host computer is
According to the SCSI standard, it is issued by a command block called CDB.

SCSI interface controller 10
0 receives the CDB from the host computer,
An interrupt signal is sent to the PU module 190. Upon receiving the interrupt signal, the CPU module activates the request receiving module 219.

The request receiving module 210 extracts the CDB from the SCSI interface controller 100, generates a packet holding the request position and the request length, and passes the packet to the request conversion 220 module.

Upon receiving the packet, the request conversion module 220 refers to the held data management table 262 based on the received packet and checks whether the requested data is in the data holding area 260. If the target data exists, an expected value is obtained from the held data management table 262.

If the target data does not exist, the data immediately before the requested data is stored in the data holding area 26.
Check if it is within 0. If the previous data exists,
An expected value is obtained from the held data management table 262.
If the immediately preceding data does not exist, the expected value is set to zero.

Next, the request conversion module 220 obtains the expected value of the data to be read this time based on the obtained expected value. The expected value of this data is a value obtained by adding 1 to the obtained expected value. The request conversion module 220 changes the request length based on the expected value of the data to be read repeatedly. Here, a value obtained by multiplying the expected value by a constant M is set as a new required length.
If the new request length is smaller than the old request length, the old request length is used as it is. If the new request length exceeds a predetermined maximum value (for example, 2048), the maximum value is set as the request length.

Next, by referring to the held data management table 262 again, it is checked whether or not data based on the requested position requested by the host computer and the changed request length exists in the data holding area 260. As a result of the check, if all the data exists in the data holding area 260, the requested data is transferred to the host computer via the request receiving module 210 without reading the data from the disk storage device 399, and the process is terminated. In addition, when there is a partial presence (when there is no 50% or more) or when there is no such portion, a change is made so that a request for reading a portion that lacks the required position and required length is made. A new packet is generated, and the packet is passed to the request issuing module 240.

The request issuing module 240 creates a CDB according to the SCSI standard based on the received packet, and issues the CDB to the disk storage device 399. When the disk storage device 399 receives the CDB, the
B is interpreted, data is read, and transfer to the storage controller 1 is started.

The storage control device 1 transmits the data transferred from the disk storage device to the SCSI controller 110 and the P
The data is held in the data holding area 260 in the memory 150 via the CI bus 130. The request issuing module 240
When the reading process from the disk storage device 399 is completed, the held data management table 261 is updated so as to match the data held in the data holding area 260.
In addition, as an expected value of the data transferred from the current disk storage device and held in the data holding area 260, the expected value of the previously read data to be read this time is written in the entry of the held data management table 262 corresponding to the data. , The request receiving module 210, based on a request from the host computer (a request that is converted by the request conversion module), stores the data holding area 26 in the memory 150.
0 data to the SCSI interface controller 1
00 is transferred to the host computer 90 via the PCI bus 120. When the data transfer is completed, the SCSI interface controller 100 reports a read completion report to the host computer 90. This completes the reading process.

As in the first embodiment, the request receiving module receives the next read request from the host computer until the read processing of one read request sent from the host computer is completed. Reject. Alternatively, the next read request is accepted, but the read processing is always performed after the read processing of the previous read request. The request is read out and processed when it becomes possible to start the request reading process. That is, the read processing of the read request received earlier from the host computer is preferentially performed.

In the third embodiment, as in the first embodiment, the required length is changed according to the degree of continuity of the target block with the previously accessed block. The number of blocks is read ahead according to the likelihood that the request is one of requests during sequential access. Therefore, as in the first embodiment, the average response time during multiple sequential access can be further reduced.

Hereinafter, a fourth embodiment of the present invention will be described.

FIG. 14 shows the configuration of a computer system according to the fourth embodiment.

The difference from the computer system according to the first embodiment shown in FIG.
30 SCSI interface controllers 110
And each SCSI interface controller has one disk storage device (391 to 3).
94). In the fourth embodiment, as in the second embodiment, the request management table 2
50 is omitted, and the rule for changing the request length performed by the request conversion module 220 is changed from that of the first embodiment.

Here, each of the four disk storage devices (391 to 394) connected to the storage control device 1 has a logical block address. In order to make the four disk storage devices appear as one to the host computer 90, the request conversion module 220 performs logical block address conversion according to the following rules.

For example, as shown in FIG. 15, when each disk storage device (391 to 394) stores data in units of 128 blocks, the request conversion module 220 of the storage control device 1 sends the logical block address indicated by the host computer. 0 to the logical block address 0 of the drive 391, and the logical block address 128 indicated by the host computer to the logical block address 0 of the drive 392.
The logical block address 1 indicated by the host computer
280 is the logical block address 25 of the drive 393
6 makes it possible to use four disk storage devices (391 to 394) in the same manner as when the host computer accesses one disk storage device. Such a logical block address conversion process in the request conversion module 220 is referred to as an “address conversion process”.

In such a configuration, in the present embodiment, the request conversion module 220 refers to the held data management table 261 based on the packet received from the request reception module 210, and stores the data requested by the host computer. It is checked whether or not the data is in the data holding area 261.

When the target data exists, the request length is not changed. The request receiving module sends the data in the data holding area 260 in the memory 150 to the SCSI interface controller 100 via the PCI bus 120 based on the request from the host computer (the request from which the request conversion module 220 converts the request). Transfer to the host computer 90.

On the other hand, if the target data does not exist, the request length is changed. The required length is replaced with a constant value J. Here, J is assumed to be 2048. Next, the request conversion module performs the above-described “address conversion processing”.
If the request spans multiple disk storage devices, it generates multiple packets and passes the packets to request issuing module 240.

The request issuing module 240 creates a SCSI CDB based on the received packet and issues the CDB to the disk storage device. When a plurality of packets are received, a CDB is issued to each of a plurality of disk storage devices.

Upon receiving the CDB, the disk storage device interprets the CDB, reads out the data, and starts transfer. The storage control device 1 transfers the data transferred from the disk storage device to the data holding area 26 in the memory 150 via the SCSI controller 110 and the PCI bus 130.
Hold at 0.

When the reading process from the disk storage device is completed, the request issuing module 220 updates the held data management table 261. When a request has been issued to a plurality of disk storage devices, it is of course necessary to wait for the reading process of all issued disk storage devices to be completed.

The request receiving module, based on a request from the host computer (a request that is converted by the request conversion module), stores the data in the data holding area 2 in the memory 150.
The SCSI interface controller 100 transfers the 60 data to the host computer 90 via the PCI bus 120. When the data transfer is completed, the SCSI interface controller 100 reports a read completion report to the host computer 90. with this,
The read processing ends.

Note that, as in the first embodiment, the request receiving module receives the next read request from the host computer until the read processing of one read request sent from the host computer is completed. Reject. Alternatively, the next read request is accepted, but the read processing is always performed after the read processing of the previous read request. The request is read out and processed when it becomes possible to start the request reading process. That is, the read processing of the read request received earlier from the host computer is preferentially performed.

According to the fourth embodiment, similarly to the second embodiment, the average response time at the time of multiple sequential access can be further reduced.

[0120]

As described above, according to the present invention, the response time to each host computer in multiple sequential access can be reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a computer system.

FIG. 2 is a block diagram illustrating a configuration of a disk storage device.

FIG. 3 is a diagram showing a recording format on a storage medium.

FIG. 4 is a diagram showing a configuration of a sector.

FIG. 5 is a block diagram illustrating a configuration of a control program.

FIG. 6 is a diagram showing a table on a memory.

FIG. 7 is a diagram showing a configuration of a request management table.

FIG. 8 is a diagram showing a configuration of a held data management table.

FIG. 9 is a diagram showing an example of the contents of a request management table.

FIG. 10 is a diagram showing a relationship among a seek time, a rotation waiting time, and a data transfer time.

FIG. 11 is a diagram illustrating a relationship between the number of blocks and data transfer efficiency.

FIG. 12 is a diagram showing a table on a memory.

FIG. 13 is a diagram showing a configuration of a held data management table.

FIG. 14 is a block diagram illustrating a configuration of a computer system.

FIG. 15 is a diagram showing how logical blocks are stored in a plurality of disk storage devices.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Storage control device, 10 ... Request reception means, 20 ... Request conversion means, 30 ... Information storage means, 40 ... Request issuing means, 9
0: Host computer, 99: Storage device 100: SCSI interface controller, 110
... SCSI interface controller, 120 ... PC
I bus, 130: PCI bus, 150: memory, 160
.. Bridge, 190 CPU module 200 control program 210 request receiving module 220 request conversion module 240 request issuing module 250 request management table 260 data holding area
261, held data management table, 262, held data management table 300, storage medium, 301, track, 302, sector, 303, position information area, 304, data area, 30
5 ECC area, 310 head, 315 head control means, 320 motor, 325 motor control means, 35
0: waveform shaping means, 352: ECC means, 354: buffer, 356: transfer means, 390: controller

Claims (10)

[Claims] 1. A storage controller connected between a plurality of host devices and a storage device, and issuing a read request to the storage device based on a read request issued from the host device for data stored in the storage device. A data holding unit that functions as a cache of the storage device for the plurality of host devices, and a data length that changes a data length of data targeted by a read request issued from the host device to a longer length. Changing means, at least, when not all of the data targeted by the read request whose data length changing means has changed the data length is held in the data holding means, at least a part of the data that is not stored is targeted. Read request to
Request issuing means for issuing data to the storage device in preference to a read request issued from the host device after the read request in which the data length changing device has changed the data length, and reading data from the storage device, The storage control device, wherein the data holding unit holds data transferred from the storage device in response to a read request issued by the request issuing unit. 2. A storage control device connected between a plurality of host devices and a storage device, and issuing a read request to the storage device based on a read request issued from the host device for data stored in the storage device. A data holding unit that functions as a cache of the storage device for the plurality of host devices, and a data length that changes a data length of data targeted by a read request issued from the host device to a longer length. Changing means, at least, when all of the data targeted by the read request whose data length changing means has changed the data length is not held in the data holding means, at least a part of the data that is not stored is set as a target. Read request to
Regardless of the address area on the storage device of the data targeted by the read request issued from the host device after the read request in which the data length changing means has changed the data length, the data is issued to the storage device, Request issuing means for reading data from the storage device, wherein the data holding means holds data transferred from the storage device in response to the read request issued by the request issuing means. Storage controller. 3. A storage controller connected between a plurality of host devices and a storage device, and issuing a read request to the storage device based on a read request issued from the host device for data stored in the storage device. And a data holding unit that functions as a cache of the storage device for the plurality of host devices, and an address area on the storage device of data targeted by a read request issued from the host device, the A continuity determining means for determining the degree of continuity of the data targeted by the read request issued from the device with respect to the address area on the storage device; and a read request issued from the host device. Data length changing means for changing the data length of the data to a length corresponding to the strength of the continuity determined by the continuity determining means And at least, when not all of the data targeted by the read request whose data length has been changed by the data length changing means is held in the data holding means, at least the data portion not stored is read targeted at Request
Request issuing means for issuing data to the storage device and reading data from the storage device, wherein the data holding means transmits data transferred from the storage device in response to the read request issued by the request issuing means. A storage control device characterized by holding. 4. The storage control device according to claim 3, further comprising history management means for storing an address area on the storage device of data targeted by a read request issued from the host device in the past, The continuity determination unit refers to the history management unit, and in a past address stored in the history management unit, in an address area on the storage device of data targeted by a read request issued from a host device. The strength of the continuity of the data targeted by the read request issued from the host device on the address area on the storage device is determined, and the request issuing unit changes the data length by the data length changing unit. If not all of the data targeted by the read request is stored in the data holding unit, the data is read out of the unstored data portion. Issue a request to the storage device, among the data transferred from the storage device in response to the issued read request, transfer to the host device the data targeted by the read request issued by the host device, When all of the data targeted by the read request whose data length has been changed by the data length changing means is stored in the data holding means, the data stored in the data holding means may include the host device. A storage control device for transferring data targeted by an issued read request to the host device. 5. The storage control device according to claim 3, wherein a temporary expected value of a read request issued in the past from said host device is replaced with an expected value set for an address area immediately before said address area. Is larger, the expected value setting means is set to be larger, the continuity determining means is issued from the host device in accordance with the provisional expected value of the read request issued from the host device set by the expected value setting means. The strength of the continuity of the address area on the storage device of the data targeted by the read request with respect to the address area on the storage device of the data targeted by the read request issued from the host device in the past. The request issuing unit determines that all of the data targeted by the read request whose data length has been changed by the data length changing unit is stored in the data holding unit. If it is not stored, a read request is issued to the storage device for the portion of the data that is not stored, and the address area of the data transferred from the storage device in response to the issued read request is issued. Setting the provisional expected value of the issued read request as an expected value, and among the data transferred from the storage device in response to the issued read request, the data targeted by the read request issued by the host device is When all of the data targeted by the read request whose data length is changed by the data length changing unit is stored in the data holding unit, the data stored in the data holding unit is transferred to the host device. Wherein the data targeted by the read request issued by the host device is transferred to the host device via the transfer means. Storage control device. 6. A storage system, comprising: the storage control device according to claim 1, 2, 3, 4, or 5, and a storage device connected to the storage control device. 7. The storage system according to claim 6, wherein said storage device is a disk-type storage device. 8. The storage system according to claim 7, wherein said storage device is a disk array storage device having a plurality of storage units, each of which manages a data storage area with a local address. The request issuing means of the control device, as a read request issued to the storage device, a plurality of data for each of a plurality of storage units storing the data targeted by the read request, the target data at the local address A storage system for issuing a designated read command. 9. A storage control device according to claim 1, 2, 3, 4, or 5, a storage device connected to said storage control device, and a storage control device connected to said storage control device and read out to said storage control device. A data processing system comprising: a plurality of host devices connected to read data from the storage device by issuing a request. 10. A method connected between a plurality of host devices and a storage device and prefetching data from the storage device based on a read request issued from the host device for data stored in the storage device. A data holding unit that functions as a cache of the storage device with respect to the plurality of host devices, and changes a data length of data targeted by a read request issued from the host device to a longer length; When all of the data targeted by the read request whose data length has been changed by the data length changing unit is not held in the data holding unit, a read request that targets at least a part of the data that is not stored is read,
The data length changing unit issues a request to the storage device in priority to a read request issued from the host device after the read request having changed the data length, reads the data from the storage device, and reads the data from the storage device. A read-ahead method, comprising: storing data transferred from the storage device in response to a read request issued by the request issuing unit.