JP2001249770A - Storage device with command queuing function - Google Patents

Abstract

(57) [Summary] (with correction) [PROBLEMS] In a storage device that performs command queuing, while guaranteeing the worst value of the response of a read command,
Does not significantly reduce overall command throughput. A submersion prevention measure is applied only to a read command. The sinking prevention measure is separately applied to the read command and the write command to make the read command harder to sink. A limit (A, B) is set for the number of times that a command received after a certain command is overtaken, and when the command is overtaken by the limited number of times, the command is executed with priority. The time at which each command was received is managed, and if there is a command that has not been executed for a certain period of time after being received, that command is executed with priority.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a storage device having a tag queue. In particular, the present invention relates to a storage device that changes the execution order of commands on a tag queue according to expected processing time and the like.

[0002]

2. Description of the Related Art The SCSI-2 standard supports command queuing and reordering. By performing command queuing, a plurality of I / O requests issued from the same or a plurality of initiators can be received and queued. In the hard disk drive, the command execution order is rearranged so that the useless movement time of the read / write head is minimized. The command queuing and reordering are described on pages 114-116 of Seiichi Sugaya, CQ Publishing Company, SCSI-2 Detailed Description.

A well-known example close to the present invention is disclosed in "Execution Order Optimization Method in Disk Drive" Matsumoto et al., JP-A-6-259.
No. 198 (hereinafter referred to as known example 1). Known example 1
So, there are two command queues, a rearrangement queue and a standby queue, the command to be executed is selected from the rearrangement queue, the newly entered command is put in the standby queue,
There is disclosed a technique for preventing a command from sinking by exchanging a rearrangement queue and a standby queue when the rearrangement queue becomes empty.

[0004]

Consider a disk drive having a command queuing function connected to a disk controller with a write-back cache.
When there is no request data in the cache of the disk controller, the read command issued from the host computer does not end until the request data is read from the disk drive, so that the response is important in the read processing of the disk drive. On the other hand, the write command issued from the host computer is completed when writing the data to the disk controller cache and reporting the end, and the response of the process of writing the data written on the cache to the disk drive is not important, rather The throughput of how much data in the cache can be written to the disk drive per second is important. Despite the difference in required response throughput between read and write in this way, sinking is prevented in the same way as read commands even for write commands that do not require a response, lowering overall throughput. There was a problem that it was.

SUMMARY OF THE INVENTION An object of the present invention is to provide a storage device which ensures a response to a read command and does not significantly lower the overall throughput.

[0006]

SUMMARY OF THE INVENTION In order to achieve the above object, a measure for preventing sinking is applied only to a read command.

In order to achieve the above object, a sinking prevention measure is separately applied to a read command and a write command, so that the read command is harder to sink.

[0008] Regarding two types of concrete measures to prevent sinking,
This will be described below.

A limit is set on the number of times a command received after a certain command is overtaken, and if the command is overtaken by the limited number of times, the command is executed with priority.

[0010] The time when each command is received is managed, and if there is a command that has not been executed for a certain period of time after being received, that command is executed with priority.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS (First Embodiment) FIG. 1 shows a computer system 1200 to which the present invention is applied. The computer system 1200 includes a host computer 1010 for issuing I / O requests, and a storage system 12 for processing I / O requests.
Consists of ten.

The storage system 1210 includes a storage controller 1000 for receiving an I / O request and a storage 1100 for storing data.

The storage controller 1000 includes a host interface 1030 for exchanging with the host computer 1010, an FC interface 1070 for exchanging with the storage 1100, a CPU 1020 for managing the storage controller 1000, and write data from the host computer 1010. It comprises a cache 1050 for temporarily storing, a memory 1060 for storing programs executed by the CPU 1020 and management data of the cache 1050, and a service board 1040 for a user to input various parameters to the storage device 1100 or the storage device controller 1000.

The storage device 1100 is a storage device controller.
FC interface 1080 for communication with 1000, disk 1090 storing data, disk controller 1110 for controlling disk 1090, CPU 112 for interpreting received I / O requests and instructing disk controller 1110
0, comprising a memory 1130 for storing programs and data executed by the CPU 1120. Although the present embodiment assumes a hard disk as the disk 1090, the present invention is also applicable to an MO, a CD-RW, or the like.

FIG. 2 shows a command management table 1500 stored in a memory 1130 on a storage device 1100. The command management table 1500 includes a plurality of command management table entries 1510. Command management table 1500 is a storage device
1100 is a table that manages I / O requests received by I / O1
One command management table entry 1510 is prepared for each.

The command management table entry 1510 includes a valid / invalid flag 1520 indicating validity / invalidity of the entry,
A read / write type 1530 indicating whether an / O request (hereinafter referred to as a command) is a read or a write, an LBA 1540 indicating a storage address of the requested data, and a size 1 indicating the length of the requested data
550, overtaking counter 156 indicating how many times the I / O request has been overtaken by a later command
0, Estimated processing time indicating the estimated time required to process the command from the current head position of the disk 1090
1570, a forward pointer 1580 and a backward pointer 1590 that constitute a tag queue 2060 described later.

FIG. 3 shows a drive management table 2000 for managing the states of the storage device 1100 and the disk 1090. The drive management table 2000 includes a limit number A2010 that specifies the limit of the number of overtakes for the read command, a limit number B2020 that specifies the limit of the number of overtakes for both the read / write command, and the current head position 2 of the disk 1090.
030, MRU pointer 20 that constitutes a tag queue 2060 described later
40 and an LRU pointer 2050. Here, the limit number A2010 is smaller than the limit number B2020 or the limit number A201.
Assume that only 0 is valid.

FIG. 4 shows a tag queue 2060 of the storage device 1100.
FIG. The MRU pointer 2040 and the LRU pointer 2050 are both ends of the tag queue 2060, and the command management table entry 1510 is bidirectionally connected by a forward pointer 1580 and a backward pointer 1590. The command received by the storage device 1100 is linked to the MRU position, and the command executed by the disk 1090 is selected from all commands on the queue in consideration of conditions. The direction approaching the MRU position is defined as front, and the direction approaching the LRU position is defined as rear. The process of selecting one of the commands on the tag queue 2060 is a next command determination process 4000 described later in this embodiment.

FIG. 5 shows a flowchart of a command reception process 3000 executed by the CPU 1120 when the storage device 1100 receives a command from the storage device controller 1000. First, in step 3010, the storage device 1100
Receive command through 80. Next, in step 3020, the CPU 1120 determines whether the received command is a read command, a write command, or any other command. If it is determined to be read or write, in step 3030, the validity / invalidity flag 1520 of the command management table entry 1510 is searched for an invalid one, and the validity / invalidity flag 1520 of the empty command management table entry 1510 found in step 3040 is validated. The type 1530, LBA 1540, and size 1550 are set according to the received command, the overtake counter 1560 is set to zero, and the expected processing time is initialized to −1, which means invalid. Then, in step 3050, the MR of the tag queue 2060
Connect the entry to the U position. Then, in a step 3060, it is determined whether or not the disk 1090 is executing any command. If the command is being executed, the command receiving process 3000 ends. If not, the execution of the command is started in step 3070, and the command receiving process 3000 ends.

If it is determined in step 3020 that the command is neither read nor write, in step 3080, it is determined whether the received command is a command for setting a limited number of times. If the command is to set the limit count, in step 3090, the limit count A2010 and the limit count B
2020 is set, and the command reception processing 3000 ends. If it is not a command to set the limit, step
Processing for other commands is performed in 3100, and the command reception processing 3000 ends. The processing for the other commands is not related to the present invention, and the description is omitted. The command for setting the limited number of times can be realized by, for example, a SCSI mode select command.

FIG. 6 shows a flowchart of a command completion process 3500 executed by the CPU 1120 when a command is completed on the disk 1090. First, in step 3510, the disk 1090 completes execution of the command, and the disk controller 1110 detects it and reports it to the CPU 1120. Then, in step 3520, the CPU 1120 reports the completion of the command to the storage device controller 1000 via the FC interface 1080. The storage controller 1000 receiving the completion report sends a completion report to the host computer 1010 if the command for which the completion report has been made is a read command. Then, in step 3530, the CPU 1020 removes the command management table entry 1510 corresponding to the command whose completion has been reported from the tag queue 2060, and invalidates the valid / invalid flag 1520 of the command management table entry 1510 in step 3540. Next, in step 3550, it is determined whether an unexecuted read command or write command remains in the tag queue 2060.
00 ends. If it still remains, the next command determination processing 4000 described later is performed in step 4000, and step 35
Execute the command determined in step 70
An instruction is given to the disk 1090 via 1110, and the command completion processing 3500 ends.

FIG. 7 shows a flowchart of the next command determination process 4000. First, in step 4010, the LR of the tag queue 2060
The command at the U position is selected, set as the next execution candidate in step 4020, and the expected processing time of the command is calculated in step 4030 and set to the estimated processing time 1570. The expected processing time is preferably calculated from the seek distance, the rotation waiting time, and the medium access time, but in this embodiment, for simplicity, the absolute value of the difference between the LBA 1540 and the head position 2030 is approximated to the expected processing time. I will use it as a value.
Next, in step 4040, it is determined whether or not the limit number B2020 is valid and the overtake counter 1560 of the command is larger than the limit number B2020. If not, in step 4050, the limit count is A2010
Is valid and the read / write type 1530 of the command is read and the command is overtaken, and it is determined whether the counter 1560 is larger than the limit number A2010.If so, the process jumps to step 4110;
In step 4060, it is determined whether there is any more command by looking at the forward pointer 1580 of the command. If it is determined that there is not, jump to step 4120, and if it is determined that there is, step
At 4070, the command indicated by the forward pointer 1580 of the command is selected, and thereafter, the command is set as the command. Then, in step 4080, the estimated processing time of the command is calculated and set to the estimated processing time 1570.
At 4090, the expected processing time 1570 of the command is compared with the expected processing time 1570 of the next execution candidate command, and if the expected processing time 1570 of the command is shorter, step 4100
In step 40 as the next execution candidate command
The process jumps to step 40, and if not smaller, jumps to step 4040 without performing step 4100.

In step 4110, the command is set as the next execution candidate command. In step 4120, the next execution candidate command is determined as the next execution command. In step 4130, the value obtained by adding the size 1550 to the head position 2030, the LBA 1540 of the next execution command, , And in step 4140, trace the backward pointer 1590 of the next execution command, add 1 to all overtake counters 1560 of commands closer to the LRU position of the tag queue 2060 than the next execution command, and end the next command determination processing 4000. .

Here, the limit number A2010 is the limit number B20.
Considering the case where the value is less than 20 or only the limit A2010 is valid, the read command is overtaken by less than the write command from step 4050 and becomes the next execution command with the value of the counter 1560. The worst value is reduced. However, the limit count
If only A2010 is valid, the execution time of the write command may be extremely slow.Therefore, take a very large timeout value in the storage controller 1000, do not apply the timeout to the write command,
It is necessary to take measures such as extending the timeout time in response to the termination of another command. In addition, depending on the value of the limit number of times A2010 and the limit number of times B2020, not only the write command but also the read command may not be timed out, or the time-out time may be extended according to the end of another command. Need to do.

FIG. 8 shows the storage controller 1000 when the user enters a limited number of times from the service board 1040.
37 is a flowchart of a limit number setting process 4500 executed by the CPU 1020 in the server. First, in step 4510, the user inputs two types of limit times from the service board 1040. Then, in step 4520, a limit count setting command is created on the memory 1060. Limit count set command entered 2
This is a command for passing the type limit count to the storage device 1100, and can be realized using, for example, a SCSI mode select command. Then in step 4530 FC interface 1070
A limit count setting command is issued to the storage device 1100 through the process, and the limit count setting process 4500 ends.

FIG. 9 shows that when the host computer 1010 issues a read command or a write command to the storage controller 1000,
47 is a flowchart of a synchronization command processing 5000 executed by the CPU 1020 on the 00. First, in step 5010, a command is received via the host interface 1030. Then, it is determined whether or not the received command is a read in step 5020, and if not, the process jumps to step 5070. If it is a read, it is determined in step 5030 whether or not the requested data is in the cache 1050. If it is determined that the data is present, the process jumps to step 5060. If not, a read request is issued to the storage device 1100 in step 5040. And step
Cache read data from storage device 1100 in 5050 10
In step 5060, the requested data in the cache 1050 is transferred to the host computer 1010 via the host interface 1030, and in step 5080, the completion of the command is reported to the host computer 1010, and the synchronous command processing 5000 ends. I do.

If it is determined in step 5030 that the data is not a read, in step 5070, the write data from the host computer 1010 is received on the cache 1050 via the host interface 1030, and immediately in step 5090, the write data is sent to the host computer 1010. Report completion of write command. Then, in step 5100, the storage device 1100
Issue a write request to the storage device 1100 to write unreflected data in the cache 1050 to the
At 10, the write data is transferred to the storage device 1100, and the synchronous command processing 5000 ends.

In this embodiment, for simplicity, it is assumed that the commands issued from the host computer 1010 are only read / write.

Here, a multitasking operating system runs on the CPU 1020 so that the controller controller 1000 can simultaneously process a plurality of read / write commands, and the synchronous command processing 5000 and the limit number setting processing 4500 are executed as tasks. Shall be performed.

As described above, for the host computer 1010, it is necessary to wait for the read command to read the requested data from the storage device 1100 when there is no requested data in the cache 1050. Since the completion report is received at the time of writing, there is no need to wait for the storage device 1100 to actually write data to the disk 1090. Therefore, the read command executed in the storage device 1100 has the worst value of the response time as well as the throughput as an important performance index.
In the write command executed in the above, only the throughput needs to be considered, and the worst value of the response time does not need to be particularly considered.

On the other hand, the presence of the step 4050 of the next command determination processing 4000 allows the read command to be written if the value of the limit number B2020 is invalid or the limit number A2010 is smaller than the limit number B2020. Since the command is executed preferentially if it is overtaken by fewer times than the command, while guaranteeing the response time of the read command, the write does not limit the worst value of the response time as much as the read,
Overall throughput can be ensured.

(Second Embodiment) In the first embodiment, the number of overtaking commands is limited to the limited number of times A2010 or B.
As soon as 2020 is exceeded, priority is given without considering the current head position, so the overtaking limit has a large effect on throughput.

In the second embodiment, the overtaking counter
By making the command easy to execute according to the value of 1560, the overtake counter 1560 is set to the limit number of A201
It is difficult to exceed 0 or the limit number B2020, and the influence of the overtaken limit on the throughput is reduced.

Hereinafter, only portions different from the first embodiment will be described.

FIG. 10 is a flowchart of the next command determination processing A4001. In the present embodiment, the next command determination processing A400 is used instead of the next command determination processing 4000 in the first embodiment.
Execute 1.

First, at step 4010, the LRU of the tag queue 2060
The command at the position is selected, and is set as a next execution candidate in step 4020. In step 4160, the expected processing time of the command is calculated and corrected in the same manner as in the first embodiment, and set to the estimated processing time 1570. If the command is a read, the command is overtaken by the counter 15
The value obtained by dividing the value of 60 by the smaller value of the limit number A2010 or the limit number B2020 is applied to the expected processing time, and if it is a write, the value of the overtake counter 1560 of the command is changed to the value of the limit number B2020. Divide the value by the expected processing time. If the value of the limit number of times B2020 is invalid, no correction is made for the light. As a result, as the value of the passing counter 1560 increases, the value of the estimated processing time 1570 becomes smaller than the expected value of the actual processing time, and the possibility that the command is executed increases. Next, in step 4040, the limit number B2020 is valid, and the command is overtaken by the counter 1560.
It judges whether it is larger than 020, and if it is larger, jumps to 4110. If not, limit the number of times in step 4050
A2010 is valid and the read / write type of the command is 1530
Is read and the command is overtaken by counter 1560
Is larger than the limit number A2010, and if so, the process jumps to step 4110. Otherwise, in step 4060, it is determined whether there is any more command by looking at the forward pointer 1580 of the command. If it is determined that there is no command, the process jumps to step 4120. If it is determined that there is, the command indicated by the forward pointer 1580 of the command is selected in step 4070, and the command is set as the command thereafter. Then, in step 4150, the expected processing time of the command is calculated, corrected in the same manner as in step 4160, and set to the estimated processing time 1570. If the command is a read, the value of the overtake counter 1560 of the command is divided by the smaller of the limit count A2010 and the limit count B2020, whichever is smaller, and the result is applied to the estimated processing time obtained in step 4150. If it is a write, the value of the overtake counter 1560 of the command is
The result of dividing by the value of B2020 is applied to the estimated processing time obtained in step 4150. As a result, as the value of the counter 1560 is overtaken, the expected processing time 15
A value of 70 will be less than the expected actual processing time,
The possibility that the command is executed increases. And
In step 4090, the expected processing time 1570 of the command is compared with the expected processing time 1570 of the next execution candidate command. If the expected processing time 1570 of the command is smaller, in step 4100 the command is set as the next execution candidate command in step 4040. Jump to step 4100 if not small
Jump to step 4040 without performing.

In step 4110, the command is set as the next execution candidate command. In step 4120, the next execution candidate command is determined as the next execution command. In step 4130, the value obtained by adding the size 1550 to the head position 2030, the LBA 1540 of the next execution command, , And in step 4140, trace the backward pointer 1590 of the next execution command, add 1 to all of the overtake counters 1560 of commands closer to the LRU position of the tag queue 2060 than the next execution command, and end the next command determination processing A4001. .

(Third Embodiment) In the first embodiment and the second embodiment, the worst value of the response is guaranteed by limiting the number of times that a command that has come later is overtaken. In the example, the worst value of the response is assured by specifying the maximum value of the time that the command is connected to the tag queue without being executed. Hereinafter, only the portions different from the first embodiment will be described.

FIG. 11 shows a computer system A1220 to which the present invention is applied. Computer system A1220
It comprises a host computer 1010 that issues I / O requests, and a storage system A1230 that processes I / O requests.

The storage system A1230 comprises a storage controller 1000 for receiving an I / O request and a storage A1101 for storing data.

The storage controller 1000 includes a host interface 1030 for exchanging with the host computer 1010, an FC interface 1070 for exchanging with the storage 1100, a CPU 1020 for managing the storage controller 1000, and write data from the host computer 1010. It comprises a cache 1050 for temporarily storing, a memory 1060 for storing programs executed by the CPU 1020 and management data of the cache 1050, and a service board 1040 for a user to input various parameters to the storage device 1100 or the storage device controller 1000.

The storage device A1101 includes an FC interface 1080 for communicating with the storage device controller 1000, a disk 1090 storing data, a disk controller 1110 for controlling the disk 1090, and a disk controller for interpreting received I / O requests. CPU 112 to instruct 1110
0, a memory 1130 for storing programs and data executed by the CPU 1120, and a timer 1140. Disc 1090
This example assumes a hard disk, but MO,
The present invention can be applied to a CD-RW or the like.

FIG. 12 shows a command management A table 1501 stored in the memory 1130 on the storage device A1101. The command management table A1501 includes a plurality of command management table entries A1511. Command management table A1501
Is a table for managing I / O requests received by the storage device A1101, and one command management table entry A1511 is prepared for each I / O.

The command management table entry A1511 is
Valid / invalid flag 1520 indicating valid / invalid of the entry,
A read / write type 1530 indicating whether an I / O request (hereinafter referred to as a command) is a read or a write, an LBA 1540 indicating a storage address of the requested data, and a size 1 indicating the length of the requested data
550, a reception time 1600 of the command, an expected processing time 1570 indicating an expected time required to process the command from the current head position of the disk 1090, a tag queue 2060.
And a forward pointer 1580 and a backward pointer 1590.

FIG. 13 shows the storage device A 1101 and the disk 109.
It is a drive management table A2100 that manages the state of 0.
The drive management table A2100 includes a time limit A2060 that defines a time limit on the tag queue 2060 for a read command, and a time limit that specifies a time limit on the tag queue 2060 for both a read and write command.
B2070, current head position 2030 of disk 1090, MRU pointer 2040 and LRU pointer 2050 that make up tag queue 2060
Consists of Here, the time limit A2060 is the time limit B20
It is assumed that it is smaller than 70 or only the time limit A2060 is valid.

FIG. 14 is a flowchart of a command receiving process A3200 executed by the CPU 1120 when the storage device 1101 receives a command from the storage device controller 1000. First, in step 3010, the storage device 1100 receives a command through the FC interface 1080. Then step 30
In 20, the CPU 1120 determines whether the received command is a read command, a write command, or any other command. If it is determined to be read or write, in step 3030, a valid / invalid flag 1520 of the command management table entry A1511 is searched for an invalid one, and the valid / invalid flag 1520 of the empty command management A table entry 1511 found in step 3040 is validated. The write type 1530, LBA 1540, and size 1550 are set in accordance with the received command, the value of the timer 1140 is set to the reception time 1600, and the expected processing time is initialized with −1 meaning invalid. Then, in step 3050, the entry is connected to the MRU position of the tag queue 2060. Then, in a step 3060, it is determined whether or not the disk 1090 is executing any command. If the command is being executed, the command receiving process A3200 ends. Step 3070 if not running
To start execution of the command, and execute command reception processing A320
0 ends.

If it is determined in step 3020 that the command is neither read nor write, in step 3110, it is determined whether the received command is a command for setting a time limit. If the command is to set the time limit, in step 3120, the time limit A2060 and the time limit B
2070 is set, and the command reception processing A3200 ends.
If it is not the command to set the time limit, the process for other commands is performed in step 3100, and the command receiving process A3200 ends. The processing for the other commands is not related to the present invention, and the description is omitted. The command for setting the time limit can be realized by, for example, a SCSI mode select command.

FIG. 15 shows a flowchart of the next command determination processing B4002. First, in step 4010, the tag queue 2060
Is selected as the next execution candidate in step 4020, the expected processing time of the command is calculated in step 4030, and is set to the expected processing time 1570. The expected processing time is preferably calculated from the seek distance, the rotation waiting time, and the medium access time, but in this embodiment, for simplicity, the absolute value of the difference between the LBA 1540 and the head position 2030 is approximated to the expected processing time. I will use it as a value. Next, in step 4170, the time limit B2070 is valid and the reception time 1600 of the command is determined based on the current value of the timer 1140.
It is determined whether or not the value obtained by subtracting is larger than the time limit B2070. If not, in step 4180, the time limit A2060 is valid, the read / write type 1530 of the command is read, and the current timer 11
It is determined whether the value obtained by subtracting the reception time 1600 of the command from the value of 40 is greater than the time limit A2060, and if so, the process jumps to step 4110; otherwise, the forward pointer 1580 of the command in step 4060 To see if there are any more commands. If it is determined that there is no command, the process jumps to step 4120. If it is determined that there is, the command indicated by the forward pointer 1580 of the command is selected in step 4070, and the command is set as the command thereafter. Then, in step 4080, the estimated processing time of the command is calculated and set to the estimated processing time 1570.
At 4090, the expected processing time 1570 of the command is compared with the expected processing time 1570 of the next execution candidate command, and if the expected processing time 1570 of the command is shorter, step 4100
In step 41 as the next execution candidate command
The process jumps to 70, and if not small, jumps to step 4170 without performing step 4100.

In step 4110, the command is set as the next execution candidate command. In step 4120, the next execution candidate command is determined as the next execution command. In step 4130, the value obtained by adding the size 1550 to the head position 2030, the LBA 1540 of the next execution command, And the next command determination processing B4002 ends.

Here, the time limit A2060 is shorter than the time limit B20.
Considering the case where the time is smaller than 70 or only the time limit A2060 is valid, the read command becomes the next execution command with a time limit shorter than the write command from step 4180, so the worst value of the response time of the read command is reduced Is done. However, if only the time limit A2060 is valid, the execution time of the write command may be extremely slow. It is necessary to take measures such as not applying or extending the timeout time according to the end of another command. In addition, depending on the value of the time limit A2060 or the time limit B2070, not only the write command but also the read command may not be timed out, or the time-out time may be extended according to the end of another command. Need to do.

FIG. 16 shows a case where the user inputs a time limit from the service board 1040 and the storage controller 10
40 is a flowchart of a time limit setting process 4600 executed by the CPU 1020 in 00. First, in step 4610, the user inputs two types of time limits from the service board 1040. Then, at step 4620, a time limit set command is created on the memory 1060. Time limit set command entered
This is a command for passing two types of time limits to the storage device A1101, and can be realized using, for example, a SCSI mode select command. And in step 4630 FC interface 1070
Then, a time limit set command is issued to the storage device A 1101 through the command line, and the time limit setting process 4600 ends.

In this embodiment, similarly to the first embodiment,
Due to the presence of step 4180 of the next command determination processing B4002, the value of the time limit B2070 is invalid or the time limit A2060
Is smaller than the limit number B2070, the read command is executed preferentially with the value that the time in the command queue is shorter than the write command, that is,
The worst value of the response time of the read command can be expected to be smaller than that of the write command.

(Fourth Embodiment) In the third embodiment, the time during which a command is connected to the command queue is limited to the time limit A2.
As soon as the time exceeds 060 or the time limit B2070, the priority is executed without considering the current head position, so that the effect of the time limit on the throughput is large.

In the fourth embodiment, the command is easily executed in accordance with the value obtained by subtracting the reception time 1600 from the current value of the timer 1140, so that the current timer 11
The value obtained by subtracting the reception time 1600 from the value of 40 is the time limit A 2060
Alternatively, it is difficult to exceed the time limit B2070, and the influence of the time limit on the throughput is reduced.

Hereinafter, only portions different from the third embodiment will be described.

FIG. 17 is a flowchart of the next command determination processing C4003. In the present embodiment, the next command determination processing C40 is used instead of the next command determination processing B4002 in the third embodiment.
Execute 03.

First, at step 4010, the LRU of the tag queue 2060
The command at the position is selected, and is set as the next execution candidate in step 4020. In step 4200, the expected processing time of the command is calculated, corrected, and set to the estimated processing time 1570 in the same manner as in the third embodiment. If the command is a read, the value obtained by subtracting the reception time 1600 of the command from the current value of the timer 1140 and the value obtained by subtracting the time limit A2060 or the time limit B2070, whichever is smaller, is used in step 4190. In the case of a write, the value obtained by subtracting the reception time 1600 of the command from the current value of the timer 1140 is divided by the value of the time limit B2070. Processing time will be spent. Thus, as the value obtained by subtracting the reception time 1600 of the command from the current value of the timer 1140 increases, the value of the expected processing time 1570 becomes smaller than the expected value of the actual processing time, and the command may be executed. Will be higher.

Next, at step 4170, it is determined whether the time limit B2070 is valid and the value obtained by subtracting the reception time 1600 of the command from the current value of the timer 1140 is larger than the time limit B2070. . If not, in step 4180, it is determined whether the time limit A2060 is valid, the read / write type 1530 of the command is read, and the value obtained by subtracting the reception time 1600 of the command from the current value of the timer 1140 is larger than the time limit A2060. Judge,
If so, the process jumps to step 4110; otherwise, in step 4060, it is determined whether there is any more command by looking at the forward pointer 1580 of the command. If it is determined that there is no command, the process jumps to step 4120. If it is determined that there is, the command indicated by the forward pointer 1580 of the command is selected in step 4070, and the command is set as the command thereafter. Then, in step 4190, the expected processing time of the command is calculated, and in step 4200, the obtained estimated processing time is corrected in the same manner as in step 4200, and set to the estimated processing time 1570. Then, in Step 4090, the expected processing time 1570 of the command is compared with the expected processing time 1570 of the next execution candidate command. If the expected processing time 1570 of the command is smaller, the command is set as the next execution candidate command in Step 4100. The process jumps to step 4170, and if not smaller, jumps to step 4170 without performing step 4100.

In step 4110, the command is set as the next execution candidate command. In step 4120, the next execution candidate command is determined as the next execution command. In step 4130, the value obtained by adding the size 1550 to the head position 2030, the LBA 1540 of the next execution command, And the next command determination process C4003 ends.

[0060]

According to the present invention, it is possible to minimize the decrease in the throughput while guaranteeing the worst value of the response time of the read command.

[Brief description of the drawings]

FIG. 1 is a diagram showing a computer system to which the present invention is applied.

FIG. 2 is a diagram showing a command management table and its entries.

FIG. 3 is a diagram showing a drive management table.

FIG. 4 is an image diagram of a tag queue.

FIG. 5 is a flowchart of a command receiving process.

FIG. 6 is a flowchart of a command completion process.

FIG. 7 is a flowchart of a next command determination process.

FIG. 8 is a flowchart of a limit number setting process.

FIG. 9 is a flowchart of a synchronous command process.

FIG. 10 shows a next command determination process corresponding to the second embodiment.
6 is a flowchart of A.

FIG. 11 is a diagram illustrating a computer system A corresponding to the third embodiment.

FIG. 12 is a diagram showing a command management table A and its entries corresponding to the third embodiment.

FIG. 13 is a diagram showing a drive management table A corresponding to the third embodiment.

FIG. 14 shows a command receiving process A corresponding to the third embodiment.
It is a flowchart of FIG.

FIG. 15 is a next command determination process corresponding to the third embodiment;
6 is a flowchart of B.

FIG. 16 is a flowchart of a time limit setting process corresponding to the third embodiment.

FIG. 17 is a next command determination process corresponding to the fourth embodiment;
It is a flowchart of C.

[Explanation of symbols]

1200 computer system, 1210 storage system, 1010 host computer, 1000 storage controller, 1040 service board, 1100 storage,
1090: Disk, 1120: CPU, 1130: Memory, 1110: Disk controller, 1500: Command management table,
1510: Command management table entry, 1560: Overtake counter, 2000: Drive management table, 2010: Limited number A, 2020: Limited number B, 2030: Head position.

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor ▲ Taka ▼ Yutaka Tadashi 2880 Kozu, Odawara-shi, Kanagawa Prefecture, Hitachi, Ltd.Storage Systems Division (72) Inventor Yusuke Hirakawa 1099 Ozenji, Aso-ku, Kawasaki-shi, Kanagawa, Japan F-term in Hitachi, Ltd. System Development Laboratory (reference) 5B065 BA01 BA03 CA15 CC08 CH15 5D044 BC01 CC04 DE91 HL02

Claims (16)

[Claims] Claims 1. A storage device having a command queuing function, which may not execute commands in the order in which they are received, and a function of limiting the number of times that commands are passed, Characterized in that the storage device has a function of preparing two or more of the above-mentioned limits and applying different values for the read command and the write command. 2. A storage system comprising: the storage device according to claim 1; and a storage device control device that controls the storage device, wherein the number of overtaking times from the storage device controller to the storage device is two or two. A storage system, wherein the number of the storage devices is set to a value larger than that. 3. The storage device system according to claim 2, wherein the storage device has a function of inputting a limit number of two or more of the number of times of overtaking from the storage device control device. system. 4. A storage device having a command queuing function and a function of not executing commands in the order in which they are received, and having a function of limiting the number of times of overtaking by a command received later. A storage device characterized in that the limit number is not applied to a write command. 5. A storage device system comprising the storage device according to claim 4 and a storage device control device for controlling the storage device, wherein a limit number of the number of times that the storage device is overtaken from the storage device controller to the storage device is set. A storage device system, characterized in that: 6. The storage device system according to claim 5, wherein the storage device system has a function of allowing a user to input a limited number of times of overtaking from the storage device control device. 7. A storage device having a command queuing function, a function of not executing commands in the order in which they are received, and a function of limiting the number of times of overtaking by a command received later. In determining the command to be executed, the estimated time required for executing each command on the command queue is calculated, and the estimated time is corrected based on the number of times that a command received later than the command overtakes the command. A storage device for comparing values. 8. A storage device system comprising the storage device according to claim 7, and a storage device control device for controlling the storage device. 9. A storage device having a function of command queuing, which may not execute commands in the order in which they are received, and a function of limiting the processing time of commands. A storage device having a function of preparing two or more upper limits and applying different values for a read command and a write command. 10. A storage device system comprising the storage device according to claim 9 and a storage device control device for controlling the storage device, wherein two or more upper limit times for processing commands from the storage device controller to the storage device. A storage device system comprising the above settings. 11. The storage device system according to claim 10, wherein a user has a function of inputting two or more upper limit times of command processing from the storage device control device. . 12. It has a command queuing function,
A storage device having a function of limiting the processing time of a command, which may not be executed in the order in which the commands are received, wherein the upper limit of the time connected on the command queue is not applied to a write command. Storage device. 13. A storage system comprising the storage device according to claim 12 and a storage device control device for controlling the storage device, wherein an upper limit of a command processing time is set from the storage device controller to the storage device. A storage device system characterized by the above-mentioned. 14. The storage device system according to claim 13, further comprising a function of allowing a user to input an upper limit of a command processing time from said storage device control device. 15. It has a command queuing function,
A storage device having a function to limit the processing time of a command, which may not be executed in the order in which the commands are received. When determining the next command to be executed, an estimated time required to execute each command on a command queue. And calculating the estimated time based on the time during which the command is connected to the command queue, and comparing the obtained values. 16. A storage system comprising: the storage device according to claim 15; and a storage device control device that controls the storage device.