Patent Application
20100070706
This application relates to and claims priority from Japanese Patent Application No. 2008-237327, filed on Sep. 17, 2008, the entire disclosure of which is incorporated herein by reference.
1. Field of the Invention
The present invention relates to a storage system having capacity virtualization function, and more particularly, to a method of allocating a physical volume area to a virtualized volume.
2. Description of the Related Art
In general, in a database management system (DBMS), records are registered in sequence with a database table (DB table). In a disk volume of a storage device storing a database table, an area storing the database table is extended to meet with the storage of records. This extension of area is performed by a program contained in the DBMS or a database manager who uses management function of the DBMS. Since this extension of area is performed in sequence according to increase in capacity of the DB table, records having low access frequency and records having high access frequency are mixed together in a disk volume.
In addition, in general, as a main use of DBMS, there is a business system for Enterprise Resource Planning (ERP) In this use, whenever a transaction occurs, a record retaining contents of the transaction is added to the DB table and thus the capacity of the DB table increases with the lapse of time. In addition, the record access frequency tends to take the same characteristic every order of year when a record is added. For example, a record access mainly occurs at a year when a transaction occurs, and a record access thereafter tends to occur in an aggregating process which occurs over a span of the end of a month, the end of a term, the end of a year, or several years.
In a search from DBMS, when a record is to be accessed with limitation to a certain period of time, the record is searched with search conditions including, for example, an index column of a registered date and a term of the date. This is a typical method to prevent an unnecessary record from being searched.
Non-Patent Document 1 (described below) discloses an example of tuning performance of DBMS using a tool for reallocation of DB areas. When a plurality of disk volumes are used from the DBMS, some or all of memory areas of a DB table are reallocated so that an access from the DBMS is not concentrated on a particular disk volume.
Patent Document 1 (described below) discloses a technique for capacity virtualization of a disk volume in a storage system. In this technique, in order to improve use efficiency of memory capacity of a disk device, a virtual disk volume is provided to the outside (a host computer and so on), memory areas of a physical disk volume are allocated to the virtual disk volume in actuality whenever data are recorded, and the data are stored in the physical disk volume. At this time, the physical disk volume is allocated to one or more pool areas and the memory areas are allocated to the virtual disk volume in a predetermined order.
Patent Document 2 (described below) discloses a structure of power-saving in a RAID (Redundant Arrays of Inexpensive Disks) storage system. This spins down a parity disk from RG (RAID Group) including a plurality of disk devices on the basis of an RAID algorithm. When the RG from which the parity disk is spun down is write-accessed, write data are received in a write back mode and are temporarily held in a disk cache, and write access to the spun-down disk device is performed with a delay after spin-up. This method allows a spin-down of some disks of the plurality of disk devices constituting the RG.
[Patent Document 1] Japanese Patent Application Laid-Open No. 2007-316725
[Patent Document 2] U.S. Pat. No. 7,035,972B2
[Non-Patent Document 1] U.S. Oracle Corporation, “Oracle Database 11g Automatic Storage Management New Features Technical White Paper,” June, 2007, (http://www.oracle.com/technology/products/database/asm/pdf/11gr1%20asm%20new%20features%20wp%2005-2007.pdf)
In recent years, with a high growth of physical integration of IT systems, there is a high demand to power-saving. With development of semiconductor technologies, reduction of power consumption has been slowly achieved with the times. In compliance with the reduction of power consumption, components of an IT system accomplish more-effective power saving by consuming power only while they are in use. Such power saving can not be accomplished only by artificial means, but a power saving structure based on the nature of an IT system is required to be equipped in the IT system.
An ERP system may be representative of an IT system. Most of ERP systems are business systems using DBMS. When DBMS is used for the ERP system, transactions occur frequently with the lapse of time and records retaining contents of the transactions are added to a DB table. The contents of the transactions are inquired and updated until the transactions are terminated. When the transactions are terminated, the DB table is maintained with the contents at the termination of the transactions. As records with the transactions terminated require for result aggregation and audit, a record access occurs. Accordingly, in the ERP system, a frequency of an access to an old record with a transaction terminated of the DB table has a tendency to be lower than a frequency of an access to a new record during transaction. When records are dividedly stored in disk volumes provided by year, it is possible to spin down disk volumes storing old data. However, on this account, it is not common to design a configuration of a DB table of DBMS based on the division of records. Although data can be moved between disk volumes using a tool for DB area reallocation of Non-Patent Document 1, it takes a long time to move the enormous amount of data.
In addition, virtual disk volumes provided to a host computer to which a volume capacity virtualization technique disclosed in Patent Document 1 is applied has no one-to-one correspondence with physical disk volumes (or RAID Groups). In addition, since a repository of a memory area is determined by a storage device, a correct memory area repository can not be known. Accordingly, even of the DB area reallocation tool of Non-Patent Document 1 is used, a record having the same access frequency can not be moved to the same physical disk volume. In addition, when a plurality of disk volumes is used to disperse an access to a particular disk volume, if the disk volumes are virtual volumes, even through the virtual volumes are disk volumes recognized by a host separately, sine their memory areas may be actually allocated to one physical disk volume, dispersion of access may not be achieved.
As apparent from the above description, in the conventional techniques, in a storage device having a capacity virtualization function for the purpose of efficient use of disk capacity in an ERP system, even when capacity virtualization disk volumes are provided by year, a physical repository of a record can not be stored in physical disk volumes different by year. That is, old and new records are stored in disk volumes of the physical unit which can be spun-down. Accordingly, there arises a problem of impossible creation of physical disk volumes which can be spun-down for power saving.
To overcome the above problem, according to an aspect of the invention, there is provided a storage device including: a memory unit including a plurality of physical RAID groups including a first physical RAID group; and a controller, the controller sequentially allocating a memory area of the first physical RAID group or a pool memory area including a plurality of first physical RAID groups to a first virtual volume based on a use order according to an access to the first virtual volume, receiving a break request, and allocating the memory area of the first physical RAID group of the next use order or the pool memory area including the plurality of first physical RAID groups of the next use order to the first virtual volume according to the access to the first virtual volume after receiving the break request.
According to another aspect of the invention, there is provided a method of allocating a memory area to a virtual volume of a storage device including a memory unit including a plurality of physical RAID groups including a first physical RAID group, and a controller, the method including the steps of: sequentially allocating a memory area of the first physical RAID group or a pool memory area including a plurality of first physical RAID groups to a first virtual volume based on a use order according to an access to the first virtual volume; receiving a break request; and allocating the memory area of the first physical RAID group of the next use order or the pool memory area including the plurality of first physical RAID groups of the next use order to the first virtual volume according to the access to the first virtual volume after receiving the break request.
10000: Database Management System (DBMS) server
20000: Storage device
30000: Storage management sever
40000: Storage Area Network (SAN)
50000: Management network
41001˜41002: Storage connection path
51001˜51003: Network connection path
Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings.
The RAID controller 21000 retains a configuration management processing program 21100, a break request processing program 21200 and an input/output request processing program 21300. The RAID controller 21000 is connected to the SAN 40000 via the data path 41001 and is connected to the management network 50000 via the data path 51001. In addition, the RAID controller 21000 is connected to the memory 22000 and the physical RG(a) 23100 to physical RG(d) 23400 via one or more data paths.
The memory 22000 stores management information 22100. The management information 22100 includes a table “configuration management” 22110, a table “user area management” 22120 and a table “free area management” 22130.
In the following description, allocation of a memory area in a physical RG is provided only by way of an example for the purpose of simplification of description. Also, it is assumed that a DB manager first creates DB basic areas in three virtual volumes, respectively. Configurations of a physical RG and a virtual volume are not limited to those shown and described in the drawings and the description.
The physical RG(a) 23100 includes a user area(a) 23110. This example corresponds to a state where the physical RG(a) 23100 has no free area. The used area(a) 23110 includes a basic DB area(A1) 23111, a basic DB area(B1) 23112, a basic DB area(C1) 23113 and an extended DB area(A2) 23114.
The physical RG(b) 23200 includes a use area(b) 23210 and an free area(b) 23220. The used area(b) 23210 includes an extended DB area(B2) 23211 and an extended DB area (B3) 23212. Here, an free area refers to a memory area which has been not yet allocated to a virtual volume. Accordingly, the physical RG(b) corresponds to an example of a physical RG having still a unused area (non-allocated area). Typically, when memory areas are needed according to an access to a virtual VOL, the memory areas are sequentially allocated from a unused area of the physical RG(b). In this embodiment, an access includes a write access and a read access.
The physical RG(c) 23300 includes a free area(c) 23320. The physical RG(c) corresponds to an example of a new physical RG which has been not yet allocated to a virtual volume. In this example, in case where a free area of the physical RG(b) disappears, when memory areas are needed according to an access to a following virtual VOL, the memory areas are allocated from this physical RG(c).
The physical RG(d) 23400 includes a DB area management area (A0) 23411, a DB area management area (B0) 23412 and a DB area management area (C0) 23413. In this embodiment, in the physical RG(d) 23400, all memory areas are always allocated to a virtual VOL(d) without being sequentially allocated to a virtual volume according to an access or the like to use it. In this example, a DB area management area refers to an area storing information used for repository management of a DB area by DBMS. A DB manager configures DBMS so that information which is accessed uniformly or is expected to be accessed uniformly is stored in advance in a DB area management area, irrespective of whether information is new or old.
An arrow of a correspondence relation 22401 indicates a head of a free area held by the table “free area management” 22130.
The virtual VOL (D) 24400 includes the DB area management area(A0) 23411, the DB area management area(B0) 23412 and the DB area management area (C0) 23413. In the virtual VOL (D), all memory areas of the physical RG(d) are allocated as memory areas of the virtual VOL(D) instead of allocating memory areas of the physical RG(b) from the physical RG(a) according to an access.
As described above, it can be seen from
The memory 12000 stores a DB area reallocation program 12100 and a DBMS program 12200. The DB area reallocation program 12100 is software of a conventional general database management system. The DBMS program 12200 also is software of a conventional general database management system.
The communication network adapter 13000 may be a general Nic supporting an IP protocol and is connected to the management network 50000 via the data path 51002. The storage adapter 14000 may be a general HostBus Adapter such as SCSI, Fiber Channel and the like and is connected to the SAN 40000 via the data path 41002.
The storage management server 30000 includes a CPU 31000, a memory 32000, a communication network adapter 33000 and a user IF adapter 34000, which are interconnected via data paths, respectively. The memory 32000 stores a storage management program 32100. In this embodiment, the storage management program 32100 may be a program to cause an instruction from a user IF to be sent to the storage device 20000 via the management network 50000.
The communication network adapter 33000 may be a general Nic supporting an IP protocol and is connected to the management network 50000 via the data path 51002.
The user IF adapter 34000 is an interface adapter connecting a display 34001, as a device for showing information to an operator, to a keyboard 34002 and a pointing device 34003 as devices for inputting information. Examples of the pointing device 34003 may include a mouse, a track ball and so on.
When the RAID controller 21000 receives an access 21401 to the extended DB area(C2) via the data path 41001, the input/output request processing program 21300 contained in the RAID controller 21000 is executed to make an access to the extended DB area (C2). This process will be described in more detail later with reference to
Upon receiving the break request 21402, the RAID controller 21000 executes the internal break request processing program 21200. The break request processing program 21200 updates to point a head memory area of the free area(c) 23320 of the next physical RG(c) 23300 from a head memory area of the free area(b) 23220 of the physical RG(b) 23200 as a memory area of a physical RG which is being pointed by the table “free area management” and is allocated to a virtual volume next.
Upon receiving the access 22000 to the extended DB area (C2) via the SAN 40000, the RAID controller 21000 executes the internal input/output request processing program 21300 to acquire a new memory area from a head memory area of the physical RG(C) 23300 represented by the table “free area management” 22130. An extended DB area(C2) 23311 of a used area 23310 is held as a memory area allocated to a virtual volume. The area represented by the table “empty area management” 22130 is updated as a head memory area of the free area 23320 except for an allocated area. In comparison with
The processes from
In this embodiment, physical RGs registered in an ascending order are allocated to virtual volumes with values held in the column “use order” 22110. That is, according to an instructed order of allocation of a physical areas to virtual volumes, physical areas of the physical RG having a lower use order are first allocated to virtual volumes, and after completion of allocation of the physical areas of the physical RG to the virtual volumes, physical areas of the physical RG having a next lower use order are next allocated to the virtual volumes.
The column “physical RG identification information” 22120 may be a number or a symbol for identifying physical RGs handled by the storage device 20000.
The column “memory area number” 22130 may be a volume of memory blocks, such as the number of logical blocks or its integral multiples, which is a value corresponding to memory capacity of a physical RG.
The column “free area number” 22140 may be a volume, which is not allocated to a virtual VOL, of a volume of the memory blocks.
The column “operation mode” 22150 may be a value representing an operation mode of a physical RG. In this embodiment, as values of this column, “energy-saving” corresponds to a power-saving state and is intended to represent a spin down state with stop of the rotation of a disk, and “normal” corresponds to a state, which is responsible to an access from a host, and is intended to represent a spin up state. “Energy-saving” may include not only the spin down state but also a state of reduction of the number of rotations of a disk and a state of stop of supply of power to a disk.
A physical RG used to be allocated to a virtual volume before a break request is received may be set to be “energy-saving.” After the break request is received, if an access to a physical RG which is being used to be allocated to a virtual volume before the break request is received is reduced or completely vanished, power consumption can be reduced by setting the physical RG to be a power-saving state.
In addition, the virtual volume (d) (physical VOL (d)) is preferably is set to be a “normal” state since an access occurs uniformly or it is expected that an access will occur uniformly, irrespective of whether information is new or old.
In this embodiment, the column “VOL attribute” 22320 stores information for distinguishing virtualization methods of its virtual volume. For example, as values of this column, “capacity virtualization” represents a virtualization method of sequentially allocating memory areas from a physical RG if the memory areas are needed according to an access to a virtual volume, and “normal” represents a virtualization method without allocating memory areas of a physical RG according to a virtualization access by all memory areas are always allocated from the physical RG in a one-to-one correspondence between a virtual volume and a physical RG without depending on an access or the like.
Specifically, in this embodiment, the virtual VOL (a) to the virtual VOL(c) become “capacity virtualization” and the virtual VOL(d) becomes “normal.”
The column “physical RG identification information for acquisition of memory area” 22330 represents a physical RG acquiring a newly required memory area. Also, this column represents a corresponding physical RG in a virtual volume to which capacity virtualization is not applied.
The column “VOL attribute” 22320 is not necessarily incorporated into the table 22300, but may be contained in a different table managing a volume.
Hereinafter, an operation of the RAID controller 20000 will be described with reference to relevant figures.
If the received management request is a configuration management request at operation block 90102, the configuration management processing program 21100 is executed at operation block 90104. Specifically, at operation block 90106, a value of the column “physical RG identification information” 22120 of a record having a value of the column “operation mode” 22150 of the table “configuration management” 22100 as “power-saving” is acquired, and then a spin down request is sent to a physical RG corresponding to the acquired value. At operation block 90107, a value of the column “physical RG identification information” 22120 of a record having a value of the column “operation mode” 22150 of the table “configuration management” 22100 as “normal” is acquired, and then a spin up request is sent to a physical RG corresponding to the acquired value. Here, the spin down request and the spin up request correspond to commands “Start/Stop Unit” (1Bh) in SCSI standards, for example and are requests for control of rotation of a motor of a hard disk drive. When the rotation of the motor is stopped according to the spin down request, power consumption of the hard disk is generally reduced accordingly. The spin down request may not have to be immediately sent to the physical RG set to be “energy-saving,” and if there occurs no access to the physical RG within a limited period of time, the spin down request may be sent to the physical RG.
If the received management request is a break request at operation block 90103, the break request processing program 21200 is executed at operation block 90105. If the received management request is not a break request, the flow is ended.
If the configuration management request is not a configuration display request at operation block 90302, at operation block 90303, it is determined based on the received parameter whether or the configuration management request is a configuration update request. If the configuration management request is a configuration update request, at operation block 90305, specified information is received from the storage management server 30000, and the table “configuration management” 22110, the table “used area management” 22120 and the table “free area management” 22130, which are management information tables stored in a memory, are updated.
In this figure, the break request processing program 21200 starts when a break request is received from the storage management server 30000. This break request is automatically issued from the DBMS server or the management server. However, without being limited to this, the break request may be manually issued by manipulation of an operator. A timing at which the break request is issued is date and time after the deadline, such as the end of a month, the end of a term, the end of a year or the like, which is predetermined by a timer on the storage management server or the DBMS server. By issuing the break request for date and time after the deadline, such as the end of a month, the end of a term, the end of a year or the like, it is possible to store data having a low access frequency due to termination of a transaction and data having a high access frequency during a transaction in different physical RGs.
However, without being limited to this, a timing at which the break request is issued may be when a backup of volume data is started or ended. By issuing the break request at this timing, data can be dividedly stored in a physical RG storing backup data and a physical RG storing other data, thereby reducing performance interference without issuing a normal access and a backup access to the same physical RG.
In the operation of the break request processing program 21200, first, at operation block 90401, virtual VOL identification information, which is an object of the break request, is received from the storage management server 30000.
Thereafter, at operation block 90402, records having the received virtual VOL identification information are searched from the column “virtual VOL identification information” 22310 of the table “free area management” 22300 and values of the column “VOL attribute” 22320 and the column “physical RG identification information for acquisition of memory area” 22330 are acquired.
Next, at operation block 90403, it is determined whether or the acquired value of the column “VOL attribute” 22320 represents capacity virtualization. If this value represents capacity virtualization, the program proceeds to operation block 90404, and if this value does not represent capacity virtualization, the program is ended.
At operation block 90404, a record having the acquired physical RG identification information of the column “physical RG identification information for acquisition of memory area” is searched from the column “physical RG identification information” 22120 of the table “configuration management” 22100, and a value of the column “use order” 22110 of the record is acquired.
At operation block 90405, a record having a value obtained by adding 1 to a use order of a search result is searched from the column “use order” 22110 of the table “configuration management” 22100, and a value of the column “physical RG identification information” 22120 of the record is acquired.
At operation block 90406, the column “physical RG identification information for acquisition of memory area” 22330 of the record searched from the table “free area management” 22300 is updated as the physical RG identification information acquired in the previous operation block, and the program is ended.
According to the above-described break request process, when a physical volume is newly allocated to a virtual volume by the next input/output process, not a physical volume being currently used to be allocated to a virtual volume but a new physical volume having the lowest use order in a physical RG not being used (all memory areas are free areas) is allocated.
Accordingly, the latest newly allocated memory area can be allocated from a different physical RG after a certain time zone. If records stored in different virtual volumes are records created within a certain period of time (a period of time ranging from a previous division instruction to a current division instruction), the records can be collectively stored in at least one physical RG.
In this embodiment, in the operation of the break request processing program 21200, an operator may send a configuration management request from the storage management server 30000 and update the column “physical RG identification information for acquisition of memory areas” 22330 of the table “free area management” 22130 in the configuration management processing program 21100 shown in
Now, the operation of the input/output request processing program 21300 will be described with reference to
First, referring to
Next, at operation block 90502, records in which the column “virtual VOL identification information” 22210 and the column “virtual VOL memory area identification number” 22220 of the table “used area management” 22200 match the received “logical VOL identification information” and “logical VOL memory area identification number,” respectively, are searched.
At operation block 90503, if it is determined based on a result of the search that there is a matching record, operation block 90505 is performed. If it is determined that there is no matching record, operation block 90504 is performed.
At operation block 90505, memory areas corresponding to values of the column “physical RG identification information” 22230 and the column “physical RG memory area identification number” 22240 of the matching record of the result of the search are accessed to perform a data input/output operation, and then the process is terminated.
If it is determined at operation block 90503 that there is no record to match the received identification number in the table “used area management” 22200, at operation block 90504, a record having the same logical VOL identification information of an object as a value of the column “virtual VOL identification information” 22310 of the table “free area management” 22300 is searched to acquire a value of the column “physical RG identification information for acquisition of memory areas” 22330 of the searched record.
Next, at operation block 90506, records in which the value of the RG identification information acquired in operation block 90504 matches a value of the column “physical RG identification information” 22120 of the table “configuration management” 22100 are searched to acquire values of the column “memory area number” 22130 and the column “free area number” 22140 of the searched records.
Then, at operation block 90507, it is determined whether or not the number of free areas acquired at operation block 90506 is more than a predetermined constant value. If so, operation block 90601 of
Referring to
At operation block 90602, a head memory area identification number of an free area is calculated from values of the column “memory area number” and the column “free area number” acquired at operation block 90506. Here, for example, a value obtained by subtracting the value of the column “free area number” from the value of the column “memory area number” may be used for this calculation.
At operation block 60603, virtual VOL identification information of an input/output object, a virtual VOL memory area identification number, the physical RG identification information of the searched free area and the head memory area identification number of the calculated free area are configured and added as a record of the table “used area management” 22200. Thereafter, operation block 90502 of
Referring to
At operation block 90702, it is determined whether or not there is a matching record in the search of operation block 90701 (that is, whether or not a physical RG of the next use order exists). If so, operation block 90703 is performed, and otherwise, the program is ended.
At operation block 90703, the value of the column “physical RG identification information for acquisition of memory areas” of the searched record of the table “free area management” is updated to a value of the column “physical RG identification information” of the searched record of operation block 90701, and then operation block 90504 of
Hitherto, the operations of the configuration management processing program 21100, the break request processing program 21200 and the input/output request processing program 21300 contained in the RAID controller 20000 have been described.
At operation block 90801, a break request is sent from the storage management server 30000 to a storage device.
At operation block 90802, an extended area of a DB area is additionally manipulated using the DBMS program 12200 from the DB server 10000.
At operation block 90803, the DBMS server 10000 reallocates data of a memory area having a high access frequency to a memory area of the virtual VOL (D) 24400 using the DB area reallocation program 12100.
Now, an operation of the DB area reallocation program 12100 will be described with reference to
As can be seen from the above description, particularly, for example, in an ERP system having greatly different access frequencies before and after an account deadline such as the end of a term or the end of a year, a break request is issued after termination of an account closing process. In addition, by changing an operation mode of a physical RG used before break request performance to “energy-saving,” it is possible to reduce power consumption of the physical RG.
When there occurs an access to the physical RG, the physical RG changed to “energy-saving” may be controlled to be changed to a “normal” operation mode. In addition, this physical RG may be again changed to the operation mode of “energy-saving” after a certain period of time elapses from the last access.
In addition, an access schedule such as an access to the physical RG changed to “energy-saving,” which is concentrated within a certain period of time is set, and then the physical RG may be controlled to be changed to the “normal” operation mode according to the set access schedule.
In addition, in case of no application of the present invention, for example, in case where no division instruction is issued to the storage device 20000, since new and old records are mixed in the same physical RG, an enormous amount of data is required to reallocate data at operation block 90803. That is, the present invention has an effect of greatly reducing the amount of data required for reallocation.
In addition, when a record is backed up before a division instruction, since a stored physical RG is different from a record after the division instruction, a physical RG accessed for backup is different from a physical RG accessed for transaction to a new record after the division instruction. Accordingly, even when the backup and the transaction coincide, it is possible to prevent process performance from being deteriorated.
The DB table “transaction history” 60000 includes, for example, a column “date” 60010, a column “lot number” 60020, a column “sales volume” 60030 and a column “sales” 60040. Here, the column “date” 60010 is defined as an INDEX column.
Embodiment 2 of the present invention is an IT system in which
Accordingly, Embodiment 2 has a configuration where the DBMS server 10000 of
A configuration of the host device is shown in
Similarly, other components of the present invention can be practiced with no limitation.
With the above configuration, in file storage for archive written-once with lapse of time, a storing site of a physical RG of data of an old file before a division instruction is different from a storing site of a physical RG of data of a new file after the division instruction. Accordingly, the physical RG in which the old file before the division instruction is stored can be power-saved such as spin down or the like.
In addition, when a data reallocation program is applied by performance of a break request, the amount of reallocated data can be reduced.
In addition, when a file is backed up before a division instruction, since a stored physical RG is different from a file after the division instruction, a physical RG accessed for backup is different from a physical RG accessed for transaction to a new record after the division instruction. Accordingly, even when the backup and the transaction coincide, it is possible to prevent process performance from being deteriorated.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2008-237327 | Sep 2008 | JP | national |
