Title: Disk array device for idle seek
Abstract: A disk array device that prevents the processing of the host device from being interrupted or delayed by drive idle seek, idle seek being executed by all the drives at approximately fixed intervals. The disk array device predicts the approach of the scheduled times to start idle seek of each of the disk drives 23, 23, . . . and exercises control so that, within the same parity group, idle seek is not executed by two or more drives 23, 23, . . . at the same time. Within the same parity group, one of the drives whose idle seek scheduled start time is approaching is considered in advance to be executing idle seek and hence data access is denied beforehand. When a data read request with respect to the drive 23 that is considered to be executing idle seek is received, the required data is recovered by reading data from all the other drives 23, 23, . . . belonging to the same parity group.
Patent Number: 6,985,998 Issued on 01/10/2006 to Higashijima, et al.
| Inventors:
|
Higashijima; Naoki (Machida, JP);
Morita; Seiki (Odawara, JP);
Yagisawa; Ikuya (Tokyo, JP)
|
| Assignee:
|
Hitachi, Ltd. (Tokyo, JP)
|
| Appl. No.:
|
834084 |
| Filed:
|
April 29, 2004 |
| Current U.S. Class: |
711/114; 711/100; 711/167 |
| Current Intern'l Class: |
G06F 12/00 (20060101) |
| Field of Search: |
711/100,111,114,154,167
|
References Cited [Referenced By]
U.S. Patent Documents
| 5418921 | May., 1995 | Cortney et al.
| |
| 5606470 | Feb., 1997 | Shioya et al.
| |
| 5793553 | Aug., 1998 | Smith.
| |
| 5838991 | Nov., 1998 | Shipman.
| |
| 5859738 | Jan., 1999 | Forehand et al.
| |
| 6112255 | Aug., 2000 | Dunn et al.
| |
| 6122131 | Sep., 2000 | Jeppson.
| |
| 6128762 | Oct., 2000 | Jadav et al.
| |
| 6385000 | May., 2002 | Ottesen et al.
| |
| 6502165 | Dec., 2002 | Kishi et al.
| |
| 2002/0135922 | Sep., 2002 | Smith.
| |
| Foreign Patent Documents |
| 0 978 824 | Mar., 1995 | EP.
| |
| 0 978 824 | Mar., 1995 | EP.
| |
| 0 820 059 | Jul., 1997 | EP.
| |
| 3-76078 | Aug., 1989 | JP.
| |
| 7-121306 | Oct., 1993 | JP.
| |
Primary Examiner: Thai; Tuan V.
Attorney, Agent or Firm: Reed Smith, LLP, Fisher, Esq.; Stanley P., Marquez, Esq.; Juan Carlos A.
Claims
What is claimed is:
1. A disk array device that provides a host device with required data in accordance
with a data read request from the host device, comprising:
a plurality of disk drives that belongs to at least one parity group, each disk
drive of which starts idle seek, unless data access is taking place, when a period
equal to or more than a first threshold value period has elapsed from the time
the previous idle seek ends;
a disk interface for communicating with the disk drives;
a host device interface for communicating with the host device; and
a control device that is connected to the disk interface and the host device
interface and controls the disk array device,
wherein the control device comprises:
a time monitoring unit which has ON or OFF status information for each of the
plurality of disk drives and alternately switches the status information of each
disk drive ON and OFF, such that, in cases where a period equal to or more than
a second threshold value period shorter than the first threshold value period has
elapsed after switching the status information of each of the disk drives from
ON to OFF, when the status information of all the other disk drives belonging to
the same parity group as each of the disk drives is OFF, the status information
of each of the disk drives is switched from OFF to ON, and then, once the idle
seek of each of the disk drives has ended, the status information of each of the
disk drives is switched from ON to OFF;
an idle seek start prediction unit that detects in advance a disk drive that
is predicted to start idle seek;
a multiple idle seek prevention unit, which, in cases where a period equal to
or more than the second threshold value period has elapsed after switching the
status information of each of the disk drives from ON to OFF, prevents each of
the disk drives from starting idle seek by performing dummy access with respect
to each of the disk drives when the status information of any of the other disk
drives belonging to the same parity group as the disk drives is ON; and
a data access control unit which, in cases where required data is read from the
disk drive in response to the read request from the host device, when the status
information of a target disk drive where the required data is stored is OFF, sends
the required data to the host device by reading the required data from a target
disk drive and, when the status information of the target disk drive is ON, recovers
the required data by reading data from all the other disk drives belonging to the
same parity group as the target disk drive without performing data access with
respect to the target disk drive, and then sends the required data thus recovered
to the host device,
wherein when the predicted disk drive that is predicted to start idle seek is
detected by the idle seek start prediction unit, if another drive within the same
parity group as the predicted disk drive is executing idle seek, the multiple idle
seek prevention unit performs dummy access with respect to the predicted disk drive
so that the predicted disk drive does not start idle seek.
2. The disk array device according to claim 1, wherein:
the time monitoring unit comprises:
a drive management table that stores, for each of the plurality of disk drives,
the status information and the elapsed period from the time the status information
was switched from ON to OFF on the previous occasion, the status information and
elapsed period being associated with discrimination information for each of the
disk drives and discrimination information for a parity group to which the disk
drives belong; and
the data access control unit judges whether to switch the status information
of each of the disk drives from OFF to ON by referencing the drive management table.
3. A disk array device that comprises a plurality of disk drives belonging to
at least one parity group, each disk drive of which starts idle seek unless data
access is taking place, in accordance with a predetermined time schedule, comprising:
an idle seek start prediction unit that detects in advance a first disk drive
that is predicted to start idle seek;
a multiple idle seek prevention unit that controls the plurality of disk drives
so that, within the same parity group, at most only one disk drive executes idle
seek and two or more disk drives do not execute idle seek at the same time; and
data read control means that, when a data read request with respect to a disk
drive executing idle seek is received, recover required data by reading data from
all the other disk drives belonging to the same parity group as the disk drive
executing idle seek,
wherein when the disk drive that is predicted to start idle seek is detected
by the idle seek start prediction unit, in cases where another drive within the
same parity group as the detected disk drive is executing idle seek, the multiple
idle seek prevention unit performs dummy access with respect to the detected disk
drive so that the detected disk drive does not start idle seek.
4. The disk array device according to claim 3, further comprising:
an idle seek start prediction unit that detects in advance a disk drive that
is predicted to start idle seek in accordance with the predetermined time schedule, wherein:
when the disk drive that is predicted to start idle seek is detected by the idle
seek start prediction unit, the data read control means consider that the detected
disk drive is executing idle seek in cases where no other disk drive within the
same parity group as the detected disk drive is executing idle seek, and, when
a data read request with respect to the detected disk drive is then received, the
data read control means recover the required data from data in all the other disk
drives belonging to the same parity group as the detected disk drive up until the
detected disk drive ends idle seek.
5. A method for controlling the idle seek of a plurality of disk drives of a
disk array device that provides a host device with required data in accordance
with a data read request from the host device, comprising: a plurality of disk
drives that belongs to at least one parity group, each disk drive of which starts
idle seek, unless data access is taking place, when a period equal to or more than
a first threshold value period has elapsed from the time the previous idle seek
ends; an idle seek start prediction unit that detects in advance said each disk
drive of which starts idle seek; a disk interface for communicating with the disk
drives; a host device interface for communicating with the host device; and a control
device that is connected to the disk interface and the host device interface and
controls the disk array device,
said method comprising the steps of:
storing ON or OFF status information for each of the plurality of disk drives;
switching the status information of each disk drive ON and OFF, such that, in
cases where a period equal to or more than a second threshold value period shorter
than the first threshold value period has elapsed after switching the status information
of each of the disk drives from ON to OFF, when the status information of all the
other disk drives belonging to the same parity group as each of the disk drives
is OFF, the status information of each of the disk drives is switched from OFF
to ON;
detecting in advance said each disk drive of which starts idle seek;
when said each disk drive of which starts idle seek is detected, if another drive
within the same parity group as the detected disk drive is executing idle seek,
performing dummy access with respect to the detected disk drive so that the detected
disk drive does not start idle seek;
once the idle seek of each of the disk drives has ended after the status information
of each of the disk drives is switched from OFF to ON, switching the status information
of each of the disk drives from ON to OFF;
in cases where a period equal to or more than the second threshold value period
has elapsed after switching the status information of each of the disk drives from
ON to OFF, preventing each of the disk drives from starting idle seek by performing
dummy access with respect to each of the disk drives when the status information
of any of the other disk drives belonging to the same parity group as the disk
drives is ON;
in cases where required data is read from the disk drive in response to the read
request from the host device, when the status information of a target disk drive
where the required data is stored is OFF, sending the required data to the host
device by reading the required data from the target disk drive and,
in cases where required data is read from the disk drive in response to the read
request from the host device, when the status information of the target disk drive
is ON, recovering the required data by reading data from all the other disk drives
belonging to the same parity group as the target disk drive without performing
data access with respect to the target disk drive, and then sending the required
data thus recovered to the host device.
6. A method of controlling the idle seek of a plurality of disk drives belonging
to at least one panty group of a disk array device that comprises the plurality
of disk drives, each disk drive of which starts idle seek unless data access is
taking place, in accordance with a predetermined time schedule, comprising the
steps of:
controlling the plurality of disk drives so that, within the same parity group,
at most only one disk drive, that is not undergoing data access, executes idle
seek such that two or more disk drives do not execute idle seek at the same time;
detecting in advance said each disk drive of which starts idle seek;
performing dummy access to another disk drive within the same parity group which
is predicted about to start idle seek so that the predicted disk drive does not
start idle seek; and
when a data read request with respect to a disk drive executing idle seek is
received, recovering required data by reading data from all the other disk drives
belonging to the same parity group as the disk drive executing idle seek.
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
This application relates to and claims priority from Japanese Patent Application
No. 2003-393643 filed on Nov. 25, 2003, the entire disclosure of which is incorporated
herein by reference.
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a technology for controlling the idle seek of
a plurality of disk drives in a disk array device.
2. Description of the Related Art
In order to prevent minute amounts of dust from being deposited on a particular
track, disk drives generally have a function such as idle seek that blows away
minute amounts of dust by moving the head at fixed intervals. Normally, when a
period of a fixed period or more has elapsed since the end of the previous idle
seek operation, a disk drive attempts to start idle seek, and starts idle seek
if data access is not being executed at this time. However, if data access is being
executed precisely at that point in time, an attempt to start idle seek is made
again after the data access has ended.
Japanese Patent Application No. H7-121306 discloses a technology that reduces
the overhead of the CPU and extends the life of the disk drive as far as possible
by not executing idle seek even in the event of an idle seek request when a power
conservation processing state is assumed.
However, when data access with respect to a disk drive is frequently not
performed in a disk array device, there are cases where a state in which data access
is performed even when the disk drive has attempted idle seek prevails and hence
idle seek is not executed for a long period. Further, when a read request with
respect to the disk drive is received from the host device during disk-drive idle
seek, the disk drive cannot be accessed until the idle seek ends. Hence, there
are also problems such as that of an interruption of or delay in the processing
of the host device. These problems are particularly prominent in the case of a
high-capacity disk drive and a high-load environment.
SUMMARY OF THE INVENTION
Accordingly, an object of the present invention is to prevent a delay
in the processing of the host device caused by disk-drive idle seek.
Another object of the present invention is to uniformly execute disk-drive
idle seek as far as possible approaching fixed intervals.
A disk array device that provides a host device with required data in accordance
with a data read request from the host device according to one aspect of the present
invention comprises: a plurality of disk drives that belongs to at least one parity
group, each disk drive of which starts idle seek, unless data access is taking
place, when a period equal to or more than a first threshold value period has elapsed
from the time the previous idle seek ends; a disk interface for communicating with
the disk drives; a host device interface for communicating with the host device;
and a control device that is connected to the disk interface and the host device
interface and controls the disk array device. Further, the control device comprises:
(1) a time monitoring unit which has ON or OFF status information for each of
the plurality of disk drives and alternately switches the status information of
each disk drive ON and OFF, such that, in cases where a period equal to or more
than a second threshold value period shorter than the first threshold value period
has elapsed after switching the status information of each of the disk drives from
ON to OFF, when the status information of all the other disk drives belonging to
the same parity group as each of the disk drives is OFF, the status information
of each of the disk drives is switched from OFF to ON, and then, once the idle
seek of each of the disk drives has ended, the status information of each of the
disk drives is switched from ON to OFF;
(2) a multiple idle seek prevention unit, which, in cases where a period equal
to or more than the second threshold value period has elapsed after switching the
status information of each of the disk drives from ON to OFF, prevents each of
the disk drives from starting idle seek by performing dummy access with respect
to each of the disk drives when the status information of any of the other disk
drives belonging to the same parity group as the disk drives is ON; and
(3) a data access control unit which, in cases where required data is read from
the disk drive in response to the read request from the host device, when the status
information of a target disk drive where the required data is stored is OFF, sends
the required data to the host device by reading the required data from the target
disk drive and, when the status information of the target disk drive is ON, recovers
the required data by reading data from all the other disk drives belonging to the
same parity group as the target disk drive without performing data access with
respect to the target disk drive, and then sends the required data thus recovered
to the host device.
According to one embodiment, the time monitoring unit comprises: a drive
management table that stores, for each of the plurality of disk drives, the status
information and the elapsed period from the time the status information was switched
from ON to OFF on the previous occasion, the status information and elapsed period
being associated with discrimination information for each of the disk drives and
discrimination information for a parity group to which the disk drives belong;
and the data access control unit judges whether to switch the status information
of each of the disk drives from OFF to ON by referencing the drive management table.
A disk array device comprising a plurality of disk drives belonging to at least
one parity group, each disk drive of which starts idle seek unless data access
is taking place, in accordance with a predetermined time schedule according to
another aspect of the present invention comprises: a multiple idle seek prevention
unit that controls the plurality of disk drives so that, within the same parity
group, at most only one disk drive executes idle seek and two or more disk drives
do not execute idle seek at the same time; and data read control means that, when
a data read request with respect to a disk drive executing idle seek is received,
recover required data by reading data from all the other disk drives belonging
to the same parity group as the disk drive executing idle seek.
According to an embodiment, the disk array device further comprises an
idle seek start prediction unit that detects in advance a disk drive that is predicted
to start idle seek in accordance with the predetermined time schedule. Further,
when the disk drive that is predicted to start idle seek is detected by the idle
seek start prediction unit, in cases where another drive within the same parity
group as the detected disk drive is executing idle seek, the multiple idle seek
prevention unit performs dummy access with respect to the detected disk drive so
that the detected disk drive does not start idle seek.
According to an embodiment, when the disk drive that is predicted to start
idle seek is detected by the idle seek start prediction unit, the data read control
means consider that the detected disk drive is executing idle seek in cases where
no other disk drive within the same parity group as the detected disk drive is
executing idle seek, and, when a data read request with respect to the detected
disk drive is then received up until the detected disk drive ends idle seek, the
data read control means recover the required data from data in all the other disk
drives belonging to the same parity group as the detected disk drive.
A disk array device that comprises a plurality of disk drives belonging to at
least
one parity group according to yet another aspect of the present invention comprises
an idle seek control unit that controls the plurality of disk drives so that, within
the same parity group, at most one disk drive that is not undergoing data access
executes idle seek in accordance with a predetermined time schedule; and data read
control means which, when a data read request with respect to a disk drive executing
idle seek is received, recover the required data by reading data from all the other
disk drives belonging to the same parity group as the disk drive executing the
idle seek.
According to an embodiment, the idle seek control unit controls the disk
drives so that, within the same parity group, at most one disk drive for which
a period equal to or more than a predetermined period from the end of the previous
idle seek has elapsed executes idle seek unless data access is taking place.
A disk drive that is constituted to measure time, and, unless data access is
taking
place, start idle seek in accordance with a predetermined time schedule based on
the measured time according to yet another aspect of the present invention comprises:
a status information communication unit that communicates status information on
whether idle seek is being executed to an external device; and a time communication
unit that communicates the measured time to the external device.
According to an embodiment, the disk drive further comprises an idle seek
external control unit that, upon receiving a command from the external device,
is forced to start or not start idle seek irrespective of the predetermined time schedule.
A method for controlling the idle seek of a disk array device according to yet
another aspect of the present invention is applied to a disk array device that
provides a host device with required data in accordance with a data read request
from the host device, comprising: a plurality of disk drives that belongs to at
least one parity group, each disk drive of which starts idle seek, unless data
access is taking place, when a period equal to or more than a first threshold value
period has elapsed from the time the previous idle seek ends; a disk interface
for communicating with the disk drives; a host device interface for communicating
with the host device; and a control device that is connected to the disk interface
and the host device interface and controls the disk array device. The control method
comprises the steps of:
(1) storing ON or OFF status information for each of the plurality of disk drives;
(2) switching the status information of each disk drive ON and OFF, such that,
in cases where a period equal to or more than a second threshold value period shorter
than the first threshold value period has elapsed after switching the status information
of each of the disk drives from ON to OFF, when the status information of all the
other disk drives belonging to the same parity group as each of the disk drives
is OFF, the status information of each of the disk drives is switched from OFF
to ON;
(3) once the idle seek of each of the disk drives has ended after the status
information of each of the disk drives is switched from OFF to ON, switching the
status information of each of the disk drives from ON to OFF;
(4) in cases where a period equal to or more than the second threshold value
period has elapsed after switching the status information of each of the disk drives
from ON to OFF, preventing each of the disk drives from starting idle seek by performing
dummy access with respect to each of the disk drives when the status information
of any of the other disk drives belonging to the same parity group as the disk
drives is ON;
(5) in cases where required data is read from the disk drive in response to the
read request from the host device, when the status information of a target disk
drive where the required data is stored is OFF, sending the required data to the
host device by reading the required data from the target disk drive and,
(6) in cases where required data is read from the disk drive in response to the
read request from the host device, when the status information of the target disk
drive is ON, recovering the required data by reading data from all the other disk
drives belonging to the same parity group as the target disk drive without performing
data access with respect to the target disk drive, and then sending the required
data thus recovered to the host device.
A method of controlling the idle seek of a disk array device according to yet
another
aspect of the present invention is applied to a disk array device that comprises
a plurality of disk drives belonging to at least one parity group, each disk drive
of which starts idle seek unless data access is taking place, in accordance with
a predetermined time schedule. The control method comprises the steps of: controlling
the plurality of disk drives so that, within the same parity group, at most only
one disk drive executes idle seek and two or more disk drives do not execute idle
seek at the same time; and when a data read request with respect to a disk drive
executing idle seek is received, recovering required data by reading data from
all the other disk drives belonging to the same parity group as the disk drive
executing idle seek.
A method of controlling the idle seek of a plurality of disk drives belonging
to
at least one parity group of a disk array device that comprises the plurality of
disk drives according to yet another aspect of the present invention, comprising
the steps of: controlling the plurality of disk drives so that, within the same
parity group, at most one disk drive that is not undergoing data access executes
idle seek in accordance with a predetermined time schedule; and, when a data read
request with respect to a disk drive executing idle seek is received, recovering
the required data by reading data from all the other disk drives belonging to the
same parity group as the disk drive executing the idle seek.
According to the present invention, within the same parity group, at most
only one disk drive executes idle seek, while the other disk drives do not execute
idle seek at the same time. Further, in the event of a data access request with
respect to the disk drive executing idle seek, the required data is recovered by
means of data read from other disk drives belonging to the same parity group, whereby
it is possible to prevent the processing of the host device from being delayed
by disk-drive idle seek.
Further, a disk drive whose idle seek is started in accordance with a predetermined
time schedule is pre-assessed, and, unless any other disk drive within the same
parity group is also executing idle seek, the detected disk drive is considered
to have commenced idle seek beforehand, meaning that data access with respect to
the detected disk drive is stopped. When data on the detected disk drive is to
be recovered from data that is read from another drive belonging to the same parity
group, disk-drive idle seek is executed as far as possible approaching fixed intervals.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a block diagram showing an outline constitution of the disk array
device according to an embodiment of the present invention;
FIG. 2 is a block diagram illustrating a parity group;
FIG. 3 shows a constitutional example of a drive management table;
FIG. 4 is a block diagram illustrating the correlation between the processing
of a RAID control program, a command issue control program and a time monitoring program;
FIG. 5 is a flowchart of processing for reading data from a disk; and
FIG. 6 is a flowchart of processing for preventing multiple idle seeks.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
An embodiment of the disk array device according to the present invention will
be described below.
FIG. 1 is an outline constitutional view of the disk array device according
to an embodiment of the present invention.
The disk array device
1 is connected to one or a plurality of host devices
5 via a host interface (hereinafter called a 'host I/F')
3. The disk
array device
1 and host device
5 can be connected via an interface
such as a SCSI (Small Computer System Interface), a fiber channel, or the like,
for example.
The disk array device
1 comprises a cache memory
9, a data control
memory
11, one or a plurality of CPU
7, a drive interface (hereinafter
called a 'drive I/F')
21, and a plurality of disk drives (hereinafter referred
to simply as 'drive')
23,
23 . . . connected to the drive I/F
21.
A host I/F
3 mainly communicates with the host device
5, sending
and receiving read/write target data together with a command such as a data access
request with respect to the host device
5 and the corresponding response,
for example. The drive I/F
21 communicates with drives
23,
23,
. . . , sending and receiving read/write target data and commands such as data
access commands with respect to the drives
23,
23, . . . , for example.
The CPU
7 parses and processes commands such as data access requests from
the host device
5, and performs overall control of the disk array device
1. The CPU
7 also controls the idle seek of the drives
23,
23, . . . according to the present invention.
The cache memory
9 temporarily stores data received from the host device
5 and data read from the drives
23,
23, . . . . For example,
upon receiving a read request for data in a certain drive
23 from the host
device
5, the disk array device
1 checks whether or not this data
is in the cache memory
9, and, when this data is indeed present in the cache
memory
9, this data is read from the drive
23 to the cache memory
9 and then sent from the cache memory
9 to the host device
5.
The data control memory
11 stores a variety of management and control
tables that are referenced and updated by the CPU
11, and various programs
run by the CPU
7. Of these tables and programs, the programs and tables
for controlling the idle seek according to the present invention in particular
include a RAID control program
13, a command issue control program
15,
a time monitoring program
17, and a drive management table
19. The
functions and operation of the RAID control program
13, the command issue
control program
15, and the time monitoring program
17 will be described
in detail subsequently.
Period measurements are made for the plurality of drives
23,
23,
. . . , and, unless data access is taking place, idle seek is started autonomously
according to a predetermined time schedule based on the periods thus measured.
More specifically, each drive
23 autonomously measures the period elapsed
since the previous idle seek end time for the drive, and, when the elapsed period
is equal to or more than a predetermined threshold value period, an attempt is
made autonomously to start idle seek, and as long as data access is not being performed
at this time, idle seek is started. If, on the other hand, data access is being
performed at the moment the attempt to start idle seek is made, each drive
23
reattempts the idle seek start after a short interval. If data access is not being
performed at this time, idle seek is started. Further, there is a plurality of
types of idle seek whose methods differ, each type of idle seek being performed
as mentioned above according to an intrinsic time schedule. Control according to
the principles of the present invention (described in detail subsequently) can
be performed in parallel relative to this plurality of different types of idle seek.
Each drive
23 does not normally have an interface for communicating,
to an external device, the progress of the time schedule of an autonomously performed
idle seek (such as the elapsed period from the end of the idle seek, for example)
and the status such as whether the idle seek is actually being performed. Therefore,
as described subsequently, the CPU
11 estimates, from outside, the progress
of the time schedule of the idle seek of each drive
23 (such as the elapsed
period from the end of the idle seek, for example) and the status such as whether
the idle seek is actually being executed (that is, simulates a period measurement
for scheduling the idle seek performed by each drive
23), and thus controls
each drive
23 based on the results of this estimation, that is, simulation.
Further, the plurality of drives
23,
23, . . . constitutes
a parity group that follows the principles of RAID for a predetermined number of
drives. FIG. 2 shows a method of storing data in the parity group in the format
3D+
1P as an example. As shown in FIG. 2A, a parity group PG in the
format
3D+
1P is constituted for each of four disk drives. As shown
in FIG. 2B, storage regions of four disk drives belonging to the same parity group
are divided into a multiplicity of small regions known as 'data stripes' extending
over all four drives
23,
23, . . . . One data stripe is constituted
by four small regions (referred to hereinbelow as 'data blocks') that are contained
in the four drives
23,
23, . . . . Of four data blocks that constitute
one data stripe, target data that is an access target of the host device
5
are stored in three data blocks, while parity data created from the target data
is stored in the one remaining data block. For example, the data stripe shown at
the very top of FIG. 2B is such that B
1, B
2, and B
3 are target
data and P(1 to 3) are parity data created from B
1, B
2, and B
3.
The three target data and single parity data item are related in that any data
item can be recovered from the three other data items. Therefore, within the same
parity group, the data of each drive
23 of the four drives
23,
23,
. . . can be recovered from data of the other three drives
23,
23,
. . . . Further, the parity group in the format constituted by the four drives
above is merely a simple example. A parity group in another format that is constituted
by a separate plurality of drives can also be adopted. Whichever format is adopted,
within the same parity group, any drive data can be recovered from the data of
all the other drives, and is one advantage that the parity drive possesses. As
described below, this advantage is utilized in the control of the idle seek according
to the principles of the present invention.
FIG. 3 shows a constitutional example of the drive management table
19.
Stored in the drive management table
19 for all the drives
23,
23,
. . . that the disk array device
1 comprises are, respectively, discrimination
information for discriminating the parity group to which the drive belongs, such
as the parity group number, for example, discrimination information for discriminating
the drive, such as a drive number, for example, the elapsed period from the time
the status information was switched from ON to OFF on the previous occasion, and
the status information. Here, status information has either of the values 'ON'
and 'OFF'. The 'ON' of the status information signifies that the CPU
11
considers (estimates) that idle seek is being executed by the corresponding drive
23, or that the CPU
11 permits the execution of idle seek by the
corresponding drive
23. Conversely, the 'OFF' of the status information
signifies that the CPU
11 considers (estimates) that the idle seek is not
being executed by the corresponding drive
23, or that the CPU
11
denies the execution of idle seek by the corresponding drive
23. The drive
management table
19 is mainly updated and referenced by the time monitoring
program
17 (described later). As described subsequently, the elapsed period
is measured by the time monitoring program
17, an operation to switch the
status information from ON to OFF and vice versa with timing that is based on the
elapsed period is performed, and data access with respect to the drive
23
is controlled by the RAID control program
13 and command issue control program
15 on the basis of the status information of the drives
23,
23,
. . . . Accordingly, the schedule for commencing idle seek that is autonomously
controlled by the drives
23,
23, . . . is suitably adjusted.
A detailed description is provided below for the control operation performed
by
the CPU
11 by executing the RAID control program
13, command issue
control program
15, and time monitoring program
17.
FIG. 4 shows the correlation between the RAID control program
13, command
issue control program
15, and time monitoring program
17.
The RAID control program
13 receives a host command
31 requesting
data access from the host device
5 (data read and data write), parses this
command, and sends an internal command
33 such as a command to read or write
data of a given storage region of a given disk to the command issue control program
15. The command issue control program
15 then sends a drive command
35 for reading or writing data to the drive
23 in accordance with
the internal command
33. The access target drive
23 does not perform
an operation according to the drive command
35 and, as a result, sends back
a drive response
37 to the drive command
35 to the command issue
control program
15. More particularly, when the drive command
35
requests data reading, data that is read (read data) is included in the drive response
37 corresponding with this drive command
35.
The command issue control program
15 sends back an internal response
39
to the internal command
33 to the RAID control program
13. More particularly,
when the internal command
33 requests data reading, the read data is included
in the corresponding internal response
39. In this case, although omitted
from the drawings, the read data is temporarily written to the cache memory
9
by the command issue control program
15 and then the cache memory
9
is read by the RAID control program
13. The RAID control program
13
sends back a host response
41 to the host command
31 to the host
device
5. More particularly, when the host command
31 requests data
reading, the read data is included in the corresponding host response
41.
The basic operation of the RAID control program
13 and command issue control
program
15 was described above. In addition to this basic operation, the
following operations are performed in order to control the idle seek of the drives
23,
23,
23 . . . .
The time monitoring program
17 continuously measures the elapsed period
from the time the respective status information items of all the drives
23,
23,
23 . . . were switched from ON to OFF on the previous occasion,
and records the elapsed period
43 in the drive management table
19.
The measurement of the elapsed period is started in sync with the time the operation
is started when the drives
23,
23, . . . are turned ON, for example,
and therefore an elapsed period that substantially matches the elapsed period from
the end of the previous idle seek measured within each of the drives
23,
23, . . . can be measured.
When the RAID control program
13 receives the data access request
31
from the host device
5 and this is a data 'read' request, the RAID control
program
13 sends a status check instruction
45 for checking the status
of the drive (hereinafter called the 'target drive')
23 carrying the read
target data (hereinafter called 'target data') to the time monitoring program
17.
In response to the status check instruction
45, the time monitoring program
17 reads status information
49 for the target drive
23 from
the drive management table
19 and transfers the read status information
47 to the command issue control program
15. The command issue control
program
15 receives the internal command (in this case, the data read request
for the target drive
23)
33 from the RAID control program
13
and receives the status information
47 on the target drive
23 from
the time monitoring program
17.
Then, if the status information for the target drive
23 is OFF (in short,
if the target drive
23 is not executing idle seek), the command issue control
program
15 reads the target data from the target drive
23 and sends
this target data to the RAID control program
13 via the cache memory
9
as the internal response
39 as mentioned earlier. When, on the other hand,
the status of the target drive
23 is ON, (in short, if the target drive
23 is executing idle seek), the command issue control program
15
reads three data items from three data blocks belonging to the same data stripe
as the data block of the target data in the three other drives
23,
23,
. . . within the same parity group as the target drive
23, recovers the
target data from these three data items, and then sends the recovered target data
to the RAID control program
13 via the cache memory
9 as the internal
response
39. The RAID control program
13 then sends the target data
received from the command issue control program
15 via the cache memory
9 to the host device
5 as the host response
41. Therefore,
even when the target drive
23 is executing idle seek, the host device
5
is able to obtain the target data without waiting until the end of idle seek.
Further, the time monitoring program
17 detects, at an early stage
some time before the actual idle seek start time, a drive among the drives
23,
23, . . . that has been estimated to be approaching the scheduled start
time for idle seek according to the time schedule that is autonomously controlled
by monitoring the elapsed period
51 of all the drives
23,
23,
. . . in the drive management table
19. Described more specifically, this
method of detection is as follows. That is, as described earlier, when, according
to the time schedule in each drive
23, a period that is equal to or more
than a predetermined threshold value period from the previous idle seek end time
(hereinafter called the 'reference threshold value period') has elapsed, idle seek
is started unless data access is being executed at this time. Therefore, the time
monitoring program
17 checks whether the elapsed period of each drive
23
in the drive management table
19 is equal to or more than a second threshold
value period (hereinafter called the 'advance threshold value period') that is
set shorter than the reference threshold value period of each drive
23 by
a predetermined period, and thus detects a drive
23 whose elapsed period
is equal to or more than this second threshold value period as a drive whose idle
seek scheduled start time is approaching.
Upon detecting such a drive, the time monitoring program
17 references
status information
52 in the drive management table
19 for other
drives within the same parity group as the detected disk drive and then checks
to see whether this status information
52 is ON (in short, whether the other
drives are executing idle seek). When, as a result, there is no drive with an ON
status among the other drives within the same parity group (that is, when none
of the other drives is executing idle seek), the time monitoring program
17
changes the status information of the detected disk drive from OFF to ON (that
is, the idle seek is considered to be ON before the idle seek is actually started)
(
53). As a result, data access to the detected disk drive that has been
considered to be executing idle seek is avoided through the data access control
by the above command issue control program
15, and therefore the autonomous
start of idle seek by the detected disk drive is assured. Further, after the time
monitoring program
17 has changed the status information of the detected
disk drive from OFF to ON, the detected disk drive actually starts idle seek, and
at the time this idle seek is estimated to have ended (when the measured elapsed
period reaches a value rendered by adding a predetermined idle seek required period
to the above-mentioned reference threshold value period, for example), the status
information of the detected disk drive is restored once again to OFF (
53).
On the other hand, when the status information of any of the other drives within
the same parity group as the detected disk drive is ON (in short, when any of the
other drives is executing idle seek), the time monitoring program
17 sends
discrimination information for the detected disk drive, that is, the drive number
55, to the command issue control program
15. The command issue control
program
15 sends a predetermined command (a dummy access command for allowing
dummy access, for example)
57 to the drive specified by the dummy number
55 (the detected disk drive) to prevent the detected disk drive from autonomously
starting idle seek. As a result, the concurrent execution of idle seek by two or
more drives within the same parity group (hereinafter called 'multiple idle seek')
does not take place.
As a result of the above control, at most only one drive
23 in a single
parity group is able to execute idle seek. Further, while the status information
of a certain single drive
23 is ON (during idle seek execution), read processing
is completed without interrupting processing by creating data from the other drives
23,
23, . . . within the same parity group as mentioned above, whereby
a delay in the processing caused by the drive
23 performing idle seek can
be prevented.
Further, when there is no other drive whose status information is ON within
the same parity group, the status information of a drive
23 whose idle seek
scheduled start time is approaching is changed to ON, and, when a data read request
for the drive
23 whose status information is ON arrives as mentioned earlier,
target data from the other drives
23,
23, . . . within the same parity
group is recovered and data access to this drive
23 (drive whose idle seek
scheduled start time is approaching) is prevented. Therefore, this drive
23
is not hindered by data access and is able to start idle seek as per the schedule.
In consequence, all the drives
23,
23, . . . are able to execute
idle seek substantially equally and to a certain extent at periods that approach
fixed intervals.
Next, the above control will be described with reference to a flowchart.
First, the flow of the processing when a data read request from the host device
5 is received will be described by using FIG. 5.
When, in the processing of the data read request, because, when target data
is stored in the cache memory
9, target data can be obtained from the cache
memory
9 and sent to the host, there is no need to perform data access with
respect to the drive
23. The processing shown in FIG. 5 is processing for
a case where there is no target data in the cache memory
9 and the drive
23 must therefore be accessed.
Upon receiving (S
1) the data access request
31 from the host device
5, the RAID control program
13 parses the data access request
31
and specifies (S
2) the drive to be accessed (hereinafter called the 'target
drive'). The RAID control program
13 instructs the time monitoring program
17 to check the status information of the target drive. The time monitoring
program
17 then reads (S
3) the status information of the target drive
in the drive management table
19 and communicates this status to the command
issue control program
15. The command issue control program
15 receives
the internal command (a data read request in this case)
33 shown in FIG.
3 from the RAID control program
13 and receives status information for the
target drive from the time monitoring program
17, and thus checks whether
the status information is ON or OFF (S
4). If the status information of the
target drive is ON, the command issue control program
15 reads the target
data from the other drives within the same parity group and data belonging to the
same data stripe and recovers the target data on the basis of these data items
before returning the recovered target data to the RAID control program
13
(S
5). If the status information of the target drive is OFF, the command
issue control program
15 reads the target data from the target drive and
returns this target data to the RAID control program
13 (S
6). Once
the RAID control program
13 receives the target data, same sends the target
data to the host device
5 (S
7).
Further, when a data write request is received from the host device
5,
the write target data is temporarily written to the cache memory
9 and a
response regarding write completion for the host device
5 is sent back at
this time, and then the write target data can be written from the cache memory
9 to the drives
23,
23, . . . by selecting a convenient period.
For this reason, there is no particular problem even when data is not written to
a drive during idle seek, and hence the idle seek control as shown in FIG. 5 need
not be performed.
Next, the flow of the control preventing multiple idle seek will be described
by using FIG. 6.
The time monitoring program
17 cyclically repeats the processing shown
in FIG. 6 at a shorter interval than the interval at which each drive
23
repeatedly attempts idle seek start after the scheduled execution time (when the
above-mentioned reference threshold value period has elapsed after the end of the
previous idle seek). In short, the processing shown in FIG. 6 is repeated substantially continuously.
When starting the processing in FIG. 6, the time monitoring program
17
references the elapsed period from the time the status information of the drives
23,
23,
23, . . . is switched from ON to OFF, and checks (S
21)
whether or not a drive whose idle seek scheduled start time is approaching by checking
whether the elapsed period is equal to or more than the advance threshold value
period above. When such a drive is detected, the time monitoring program
17
checks (S
22), by referencing the drive management table
19, whether
there is a drive
23 whose status information is ON within the same parity
group as the detected disk drive. When, as a result, a drive whose status information
is ON is present within the same parity group, the command issue control program
15 performs dummy access with respect to the detected disk drive
23
(S
23). By making this dummy access, multiple idle seek within the same parity
group is prevented. On the other hand, when the result of step S
22 is that
there is not a single drive whose status information is ON within the same parity
group, the status information of the detected disk drive
23 is changed from
OFF to ON (S
24). Although not illustrated in FIG. 6, the time monitoring
program
17 returns the status information from ON to OFF when a period (or
greater period) has elapsed from the moment the status information of the detected
disk drive
23 is changed from OFF to ON, this period being found by adding
together the differential period until the idle seek is started by the detected
disk drive
23 and the period required to execute the idle seek.
The above control of the idle seek is performed in parallel with the parity group.
The control in FIG. 6 is repeated cyclically at short intervals as mentioned
earlier. Therefore, when, after a predetermined interval, the control in FIG. 6
is executed again with respect to a drive the idle seek start of which has been
placed on standby by the processing of step S
23 (dummy access), if a certain
drive whose status information is ON is not present within the same parity group,
the status information is changed to ON, and, as a result, the start of idle seek
is granted. Because the period required to execute for idle seek is sufficiently
short compared with the advance threshold value period mentioned earlier, the period
over which the start of idle seek is placed on standby is short and hence all the
drives in the parity group are able to perform idle seek reliably at certain intervals.
Further, when the elapsed period of two or more drives within the same parity
group exceeds the advance threshold value period, control over which drive to grant
the start of idle seek to first can be implemented by means of a method that adheres
to an order of priority employing drive numbers or similar.
The embodiment of the present invention was described above but merely as an
example serving to explain the present invention, there being no intention to limit
the scope of the present invention to this embodiment alone. Therefore, the present
invention can be implemented in a variety of forms different from the embodiment
above without departing from the spirit of the present invention.
For example, in the above embodiment, the progress of the schedule for the autonomous
idle seek within the drives
23,
23, . . . and the idle seek status
cannot be communicated by the CPU
11, which estimates such information.
However, if the drives
23,
23, . . . comprise an interface that communicates,
to an external device, such internal information, such as status information on
whether idle seek is being executed, and the elapsed period measured from the time
the previous idle seek ends, for example, the CPU
11 obtains this information
from the drives
23,
23, . . . via the interface and includes this
information in the drive management table
19, whereby the above-mentioned
control can be performed more accurately. Further, the drives
23,
23,
. . . receive commands from the CPU
11 via the interface and may comprise
a function to start or not start the idle seek forcedly irrespective of the autonomously
controlled time schedule therein.
*
