This application is based upon and claims the benefit of priority of the prior Japanese Patent Application No. 2013-227724, filed on Oct. 31, 2013, the entire contents of which are incorporated herein by reference.
The embodiments discussed herein are related to an information processing apparatus, a storage device control circuit, and a storage device control method.
In recent information processing apparatuses, systems that use memory devices, such as dynamic random access memories (DRAM) or NAND flash memories, as external storage devices have been widespread. For example, there is an information processing apparatus or the like, in which a dual inline memory module (DIMM) formed of multiple DRAMS is used as a memory and a solid state drive (SSD) formed of multiple NAND devices is used as a storage.
A memory controller that controls the memory devices as described above has multiple channels, to each of which several memory devices are connected in series. The information processing apparatus improves performances of read and write processes on the memory devices by accessing the channels in parallel.
Further, if a non-volatile memory is used as a storage, a data protection technique called a redundant arrays of inexpensive disks (RAID) may be used in order to improve fault tolerance.
The memory controller, upon receiving a write request from a central processing unit (CPU) or the like, performs processes, such as an address translation or a parity calculation in RAID. Subsequently, the memory controller transfers the write request and data to a channel having a memory device to which the data and a parity are to be written.
On the contrary, the memory controller, upon receiving a read request from the CPU or the like, performs a process, such as address translation. Subsequently, the memory controller transfers the read request to a channel having a memory device in which the data is stored, reads the data from this memory device, and transmits the data to a host. Channel control units can operate in parallel, so that by performing write and read on the respective channels in a distributed manner, it is possible to improve throughput to access the memory devices.
Meanwhile, in recent years, to improve the cost effectiveness, a storage device may be configured by integrating multiple types of memory controllers and storage devices. For example, a storage system is configured by combining an SSD, which enables high-speed read and write but which is expensive, and a hard disk drive, which performs read and write at a low speed but which is cheap. In this case, to cancel out a performance difference between the memory devices, there is a proposed technique that optimizes data allocation to each of the memory devices based on information, such as an access frequency, for each data.
Patent Document 1: Japanese Laid-open Patent Publication No. 2004-21370
However, in recent years, with the advent of high-capacity memory devices, a storage capacity managed by a single memory controller is increasing. Therefore, if the technique is used that optimizes data allocation to each of the memory devices based on information, such as an access frequency for each data, because the granularity of distribution control per controller is low, it is difficult to realize flexible optimization of use efficiencies.
To cope with this, it may be possible to change a processing performance in units of channels of the memory controller in order to take a balance of performance requirements. However, if the performance is biased for each channel in a single memory controller, requests to the channels with low processing speeds are accumulated in the memory controller and may compress resources shared by all of the channels. In this case, the entire processing speed may be reduced.
According to an aspect of an embodiment, an information processing apparatus includes: an arithmetic processing unit; a plurality of channels each having a storage device; a plurality of data processing units that perform data processing on the storage devices by using the channels according to a data processing command from the arithmetic processing unit; a group managing unit that stores therein a configuration of groups into which the channels are classified based on processing performances of the respective channels, and stores therein assignments of the data processing units that perform data processing by using the channels contained in each of the groups; and a control unit that, based on the data processing command from the arithmetic processing unit, selects a group for performing data processing from among the groups stored in the group managing unit, and causes the data processing unit assigned to the selected group to perform the data processing.
The object and advantages of the invention will be realized and attained by means of the elements and combinations particularly pointed out in the claims.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are not restrictive of the invention, as claimed.
Preferred embodiments of the present invention will be explained with reference to accompanying drawings. The information processing apparatus, the storage device control circuit, and the storage device control method disclosed in this specification will not be limited by the embodiments below.
The memories 2a and 2b include multiple DIMMs 3. The CPU 4a includes, although not illustrated, a memory controller that controls the memory 2a. The CPU 4b includes a memory controller that controls the memory 2b.
The SSD 6a includes a NAND controller 10a and multiple NAND flash memories 7 that are NAND devices.
The SSD 6b includes a NAND controller 10b and multiple NAND flash memories 7. The SSD 6a and the SSD 6b have the same functions. Further, the NAND controller 10a and the NAND controller 10b have the same functions. In the explanation below, when the SSD 6b and the NAND controller 10b are not discriminated from the SSD 6a and the NAND controller 10a, they will be described as “the SSD 6” and “the NAND controller 10”.
The memories 2a and 2b are storage devices that store therein data used for arithmetic processing by the CPUs 4a and 4b, respectively. The CPUs 4a and 4b are arithmetic processing devices that perform various types of arithmetic processing by using data stored in the DIMMs 3 provided in the memories 2a and 2b, respectively. For example, by using a non-uniform memory access (NUMA) technique, the CPUs 4a and 4b acquire data stored in the DIMMs 3 in the memories 2a and 2b and perform arithmetic processing by using the acquired data.
Further, the CPUs 4a and 4b acquire data stored in the SSDs 6a and 6b, respectively, via the I/O hub 5, and perform arithmetic processing by using the acquired data. Specifically, the CPU 4a issues a read request or write request for data to the SSD 6a, and reads or writes data from or to each of the NAND flash memories 7. For example, the CPU 4a issues, to the SSD 6a, a read request containing a logical address designating data to be read. Further, the CPU 4a issues a write request containing a logical address designating a destination to which data is to be written and containing the data to be written. The CPUs 4a and 4b are examples of an “arithmetic processing unit”.
The NAND flash memories 7 are non-volatile memories that store therein various types of data. Specifically, the NAND flash memories 7 include multiple physical pages serving as data storage areas, and write data in units of a physical page. Further, the NAND flash memories 7 include multiple physical blocks with multiple physical pages, and delete data in units of a block. Meanwhile, a single physical page has a storage capacity of 8 kilobytes for example, and a single physical block has 128 physical pages for example. The NAND flash memories 7 are examples of “storage devices”.
The NAND controller 10 accesses each of the NAND flash memories 7 and reads or writes data. For example, the NAND controller 10 includes an address translation table that stores therein a logical address, which is used when each of the CPUs 4a and 4b designates a storage area storing data, and a physical address, which indicates a storage area of the NAND flash memories 7 storing data, in an associated manner.
Upon receiving a logical address together with a read request, the NAND controller 10 identifies a physical address associated with the logical address by using the address translation table, and reads data from a storage area indicated by the identified physical address. Thereafter, the NAND controller 10 sends the read data to the CPU 4a or 4b via the I/O hub 5.
With reference to
The NAND controller 10 according to the present embodiment includes a configuration register 11, a communication interface 12, a group identifying unit 13, data processing units 14a to 14c, and channel control units 15a to 15f.
Here, in the present embodiment, a case will be described in which the NAND controller 10 receives commands and data processing requests from a CPU 4; however, the NAND controller 10 may receive commands and data from other units of the information processing apparatus 1. Each unit of the information processing apparatus 1 that issues commands or processing requests to the NAND controller 10 may be referred to as a host device.
Further, respective transmission paths that connect the channel control units 15a to 15f and the NAND flash memories 7a to 7f are referred to as channels 8a to 8f. Furthermore, hereinafter, when the channel control units 15a to 15f are not discriminated from one another, they may be referred to as the channel control unit 15. Moreover, when the NAND flash memories 7a to 7f are not discriminated from one another, they may simply be referred to as the NAND flash memory 7. Furthermore, when the channels 8a to 8f are not discriminated from one another, they may be referred to as the channel 8.
Incidentally, in the present embodiment, a case will be described in which the three data processing units 14a to 14c and the six channels 8a to 8f are provided; however, the numbers of the data processing units and the channels are not limited to these examples.
Further, while the data processing units 14a to 14c have the same configurations,
The configuration register 11 stores therein a group configuration determined by a user. Further, the configuration register 11 stores therein assignment of each group to any of the data processing units 14a to 14c.
Specifically, upon receiving a write command or a read command from the CPU 4, the configuration register 11 performs write or read according to the command. For example, an operator instructs the CPU 4 to write a group configuration by using management software or a debugging interface that operates on the information processing apparatus 1. Then, the configuration register 11 stores therein the group configuration by writing it according to a write command transmitted by the CPU 4 in response to the instruction from the operator.
For example, the configuration register 11 stores therein the group configuration in a format as illustrated in
The configuration register 11 stores therein a group configuration by storing information indicating which of the channels 8a to 8f belong to each group as illustrated in
Incidentally,
In the present embodiment, the group configuration stored in the configuration register 11 is set such that all of the channels belong to any of the groups without duplication. Further, the number of the groups is equal to or smaller than the number of the data processing units 14a to 14c, and, in the present embodiment, the maximum number of the groups is three. Hereinafter, three, which is the maximum number of the groups, may be referred to as the number of all of the groups.
The communication interface 12 is an interface that mediates communication between the CPU 4 and each unit of the NAND controller 10. In reality, the CPU 4 and each unit of the NAND controller 10 communicate with each other via the communication interface 12; however, hereinafter, mediation of the communication interface 12 may be omitted for convenience of explanation.
The group identifying unit 13 receives a group configuration from the configuration register 11. Further, the group identifying unit 13 receives a data processing request, such as read or write, from the CPU 4.
In the present embodiment, as will be described later, the CPU 4 stores information on a destination group of the processing request in the data processing request.
The group identifying unit 13 selects the data processing unit 14 corresponding to the destination group designated by the processing request from among the data processing units 14a to 14c by using information that is about the data processing units 14a to 14c assigned to each of the groups and that is received from the configuration register 11. Then, the group identifying unit 13 transmits the received data processing request to the selected data processing unit 14.
Thereafter, the group identifying unit 13 returns, to the CPU 4, a response to the processing request. For example, if the processing request is a data write request, the group identifying unit 13 returns, to the CPU 4, a response indicating that transfer of the data processing request to the data processing unit 14 is completed. Further, if the processing request is a read request, the group identifying unit 13 receives, as a response to the processing request, data designated by the read request from a data buffer 143 to be described later. Then, the group identifying unit 13 transmits the received data to the CPU 4 as a response to the processing request. The group identifying unit 13 corresponds to an example of a “control unit”.
The data processing unit 14 includes a request buffer 141, an address translation control unit 142, the data buffer 143, a RAID control unit 144, and a channel selecting unit 145.
The request buffer 141 receives, from the group identifying unit 13, input of data processing requests to a group corresponding to the data processing unit 14 provided therewith. The request buffer 141 accumulates the received processing requests. Then, if the address translation control unit 142 is not performing a process, the request buffer 141 outputs the oldest processing request to the address translation control unit 142.
The address translation control unit 142 receives input of the data processing request from the request buffer 141. The address translation control unit 142 translates an address designated by the processing request into an address of the NAND flash memory 7 on the channel connected to the data processing unit 14 provided therewith. Then, the address translation control unit 142 outputs the translated address and the processing request to the channel selecting unit 145.
If the processing request is a data write request, the data buffer 143 receives, from the group identifying unit 13, pieces of data that are targets of the data processing request directed to the group corresponding to the data processing unit 14 provided therewith. The data buffer 143 accumulates the received pieces of the data. Then, if the RAID control unit 144 is not performing a process, the data buffer 143 outputs the oldest piece of data to the RAID control unit 144.
Further, if the processing request is a data read request, the data buffer 143 receives input of pieces of data from the RAID control unit 144. Then, the data buffer 143 accumulates the received pieces of the data. Thereafter, if the group identifying unit 13 is in a state in which a process is executable, the oldest piece of data is output to the group identifying unit 13.
If the processing request is a write request, the RAID control unit 144 receives input of data from the data buffer 143. Then, the RAID control unit 144 performs RAID processing on the data in accordance with a RAID configuration of the NAND flash memory 7 serving as a destination of the data. For example, in the case of a RAIDS, the RAID control unit 144 calculates a parity by using the received data. Thereafter, the RAID control unit 144 outputs the data subjected to the RAID processing to the channel selecting unit 145.
Further, if the processing request is a read request, the RAID control unit 144 receives input of data from the channel selecting unit 145. Then, the RAID control unit 144 performs RAID processing, such as data reconfiguration. Thereafter, the RAID control unit 144 outputs the data subjected to the RAID processing to the data buffer 143.
The channel selecting unit 145 receives the group configuration from the configuration register 11. Then, the channel selecting unit 145 connects to the channel belonging to the group to which the data processing unit 14 provided therewith is assigned. Herein, a case will be described in which the channels 8a and 8b are the channels belonging to the group to which the data processing unit 14 provided with the channel selecting unit 145 is assigned.
The channel selecting unit 145 receives input of a processing request from the address translation control unit 142. Further, if the processing request is a write request, the channel selecting unit 145 receives input of data from the RAID control unit 144.
Then, the channel selecting unit 145 outputs the processing request to the channel control unit 15 connected to the channel having the NAND flash memory 7 corresponding to the address designated by the processing request among the channel control units 15 connected to the channels 8a and 8b. Further, if the processing request is a write request, the channel selecting unit 145 outputs data to the channel control unit 15 that has output the processing request.
Thereafter, if the processing request is a data read request, the channel selecting unit 145 receives, from the channel control units 15, data that the channel control units 15 belonging to the same group have read from the NAND flash memories 7. Then, the channel selecting unit 145 outputs the received data to the RAID control unit 144.
The channel control units 15 acquire addresses designated by the processing request. Subsequently, the channel control units 15 perform processes on the designated addresses of the NAND flash memories 7 on the channels 8a and 8b.
For example, if the processing request is a write request, the channel control units 15 write the received data in the designated addresses of the NAND flash memories 7 on the channels 8a and 8b in accordance with the RAID configuration. Furthermore, if the processing request is a read request, the channel control units 15 read data from the the designated addresses of the NAND flash memories 7 on the channels 8a and 8b. Then, the channel control units 15 output the data read from the NAND flash memories 7 to the channel selecting unit 145 of the data processing unit 14 connected to the channel control units 15.
The NAND flash memories 7a to 7f transmit and receive data to the channel control units 15 belonging to the groups to which the corresponding channels 8a to 8f are assigned.
As described above, the data processing unit 14 performs data read and write processes. Further, as described above, the channels 8 are grouped for each of the data processing units 14. Namely, the information processing apparatus 1 according to the present embodiment is able to perform independent data processing for each group.
With reference to
In this case, a destination group is stored in the data processing request as in processing requests 401 to 403 illustrated in
The NAND controller 10 acquires a destination group stored in the processing request 401, and determines that the processing request 401 is a process on the group #1. Then, the NAND controller 10 reads data from the NAND flash memory 7 on the channel 8 belonging to the group #1.
Further, the NAND controller 10 acquires a destination group stored in the processing request 402, and determines that the processing request 402 is a process on the group #2. Then, the NAND controller 10 writes data to the NAND flash memory 7 on the channel 8 belonging to the group #2. Furthermore, the NAND controller 10 acquires a destination group stored in the processing request 403, and determines that the processing request 403 is a process on the group #3. Then, the NAND controller 10 writes data to the NAND flash memory 7 on the channel 8 belonging to the group #3.
As in the present embodiment, when a group is designated by a host device side, such as the CPU 4, the host device recognizes each of the groups as an individual storage unit.
Next, with reference to
For example, as illustrated in
In
For example, the CPU 4 selects a group according to the degree of importance of a process. The CPU 4 classifies the degree of importance of a process into a high level, a medium level, or a low level. If the degree of importance of a process is the high level, the CPU 4 decides to cause the group 91, to which the channels 8a and 8b with the highest performances belong, to perform the process of the processing request. Further, if the degree of importance of a process is the medium level, the CPU 4 decides to cause the group 92, to which the channels 8c and 8d with the medium performances belong, to perform the process. Furthermore, if the degree of importance of a process is the low level, the CPU 4 decides to cause the group 93, to which the channels 8e and 8f with the lowest performances belong, to perform the process.
In
Further, as another group configuration, as illustrated in
In
In this case, for example, the CPU 4 selects a group according to the degree of importance of a process. The CPU 4 classifies the degree of importance of a process into a high level, a medium level, or a low level. Then, if the degree of importance is the high level, the CPU 4 decides to cause the group 901, in which the RAID1 with a higher reliability and a higher process speed is formed, to perform the process. Further, if the degree of importance is the medium level, the CPU 4 decides to cause the group 902, in which the RAID5 with a higher reliability and a lower processing speed is formed, to perform the process. Furthermore, if the degree of importance is the low level, the CPU 4 decides to cause the group 903, in which the RAID0 with a lower reliability is formed, to perform the process.
In
However, the group configurations illustrated in
Next, with reference to
The configuration register 11 stores therein information on a group configuration (Step S1).
The configuration register 11 notifies the group identifying unit 13 and the channel selecting unit 145 of the group configuration (Step S2).
The channel selecting unit 145 connects to the channel 8 belonging to a group to which the data processing unit 14 provided therewith belongs (Step S3).
The group identifying unit 13 receives a data processing request (Step S4).
Subsequently, the group identifying unit 13 acquires a destination group of the processing request, and selects the destination group as a processing target group (Step S5).
The address translation control unit 142 translates an address designated by the processing request into an address of the NAND flash memory 7 (Step S6).
The data processing unit 14 performs data processing on the the channel 8 belonging to the assigned group, by using the channel control unit 15 connected by the channel selecting unit 145 (Step S7).
As described above, the information processing apparatus according to the present embodiment classifies the channels into groups and performs data processing for each of the groups. Therefore, the information processing apparatus according to the present embodiment can perform individual data processing for each of the groups. Further, the channels are classified into groups according to processing performances. Therefore, even when there are processing performance differences between the channels, the channels with different performances can operate in parallel without being interfered with one another. As a result, it becomes possible to implement the performances of storage devices on all of the channels depending on the intended use, without any waste.
Further, even when the performance of each of the channels changes due to a change in mounting states of the memory devices or a change in the RAID configurations, it becomes possible to, only by changing the group configuration, prevent a reduction in processing capabilities caused by performance differences between the channels. That is, it becomes possible to improve the performance of the entire system and the resource use efficiency without redesigning hardware, such as a large scale integration (LSI).
Next, with reference to
A NAND controller 10 according to the present modified example differs from the first embodiment in that, as illustrated in
In the present modified example, the request buffer 141 outputs a processing request to the address translation control unit 142. Thereafter, the request buffer 141 receives the processing request subjected to an address translation from the address translation control unit 142.
Then, the request buffer 141 transmits the processing request subjected to the address translation to the channel selecting unit 145 to issue a request for an access to the NAND flash memory 7.
The address translation control unit 142 receives a processing request from each of the request buffers 141 of the data processing units 14a to 14c. Then, the address translation control unit 142 performs arbitration to determine priorities on a first-come, first-served basis or in an arbitrary manner, and performs an address translation on each processing request. Subsequently, the address translation control unit 142 returns the processing request subjected to the address translation to the request buffer 141 serving as a transmission source of the processing request.
As described above, the NAND controller according to the present modified example is configured such that the address translation control units of the data processing units of the first embodiment are formed into a common unit and the other functions are provided to each of the data processing units. The address translation does not cause a delay in processes on other processing requests; therefore, even when an address translation function is shared as described above, it becomes possible to maintain the same advantageous effects as those of the first embodiment. Further, the present modified example is an example of sharing of functions, and other functions may be shared as long as the functions do not cause a delay in processes on other processing requests in the data processing unit.
Further, while the NAND controller that controls the NAND flash memory has been described as an example in the above explanation, the present invention may be applied to other controllers as long as the controllers control storage devices, such as memory devices. For example, the present invention may be applied to a memory controller that controls the DIMMs 3 in
Next, a second embodiment will be described. An information processing apparatus according to the present embodiment differs from the first embodiment in that a group is selected by a NAND controller side. A block diagram of the NAND controller 10 according to the present embodiment is the same as
In the present embodiment, the CPU 4 determines a priority of a process in the case of a write process. Then, the CPU 4 adds the priority of the process to a processing request.
The group identifying unit 13 receives a processing request from the CPU 4. If the processing request is a read request, the group identifying unit 13 outputs the processing request to the data processing unit 14 assigned to a group, to which the channel 8 corresponding to the NAND flash memory 7 having a storage area corresponding to an address designated by the processing request belongs.
On the contrary, if the processing request is a write request, the group identifying unit 13 acquires the priority stored in the processing request. The group identifying unit 13 stores therein association between each priority and a group. For example, a case will be described in which the groups 91 to 93 are provided by classification as illustrated in
The group identifying unit 13 stores therein a first threshold and a second threshold for associating the priority of the processing request with a group. The second threshold is a threshold for a priority lower than that of the first threshold.
If the priority is equal to or greater than the first threshold, the group identifying unit 13 outputs the processing request to the data processing unit 14a assigned to the group 91 to which the channels 8a and 8b belong. Further, if the priority is smaller than the first threshold and equal to or greater than the second threshold, the group identifying unit 13 outputs the processing request to the data processing unit 14b assigned to the group 92 to which the channels 8c and 8d belong. Furthermore, if the priority is smaller than the second threshold, the group identifying unit 13 outputs the processing request to the data processing unit 14c assigned to the group 93 to which the channels 8e and 8f belong.
As described above, the group identifying unit 13 according to the present embodiment determines a group that processes the processing request according to the priority of the processing request.
Further, with reference to
A processing request 411 in
On the contrary, in the case of the write request, information indicating a priority is added as in the processing requests 412 and 413. Herein, the processing request 412 has a priority of 1 and the processing request 413 has a priority of 2.
The NAND controller 10, upon receiving the processing request 411, makes copies of the request and transfers the requests to all of the groups #1 to 3. Each of the groups performs an address translation, and if an address designated by the processing request 411 has been registered, the NAND controller 10 reads data from the NAND flash memory 7 having a storage area corresponding to this address. On the contrary, if the address designated by the processing request 411 has not been registered, the group that has received the request determines that the processing request is not directed thereto and discards the request.
Further, upon receiving the processing request 412, the NAND controller 10 checks information on the priority of the processing request 412, and confirms that the priority of the processing request 412 is 1. Then, the NAND controller 10 selects a group to which the channel with the highest performance belongs from among the groups #1 to 3, and processes the processing request 412 by using the data processing unit 14 assigned to the selected group.
Furthermore, upon receiving the processing request 413, the NAND controller 10 checks information on the priority of the processing request 413, and confirms that the priority of the processing request 413 is 2. Then, the NAND controller 10 selects a group to which the channel with the second highest performance belongs from among the groups #1 to 3, and processes the processing request 413 by using the data processing unit 14 assigned to the selected group.
In this case, the NAND controller 10 selects a group based on the priority; therefore, the CPU 4 need not recognize the state of division of the groups.
The method of selecting a group on the NAND controller 10 side is not limited to the method using the priority. For example, it may be possible to use accompanying information other than the priority added to the processing request.
In this case, in the case of a write process, a host device, such as the CPU 4, adds the accompanying information to the processing request. For example, the accompanying information may be a type of a host device that has output the processing request.
The group identifying unit 13 receives the processing request from the CPU 4. If the processing request is a write request, the group identifying unit 13 acquires the accompanying information stored in the processing request.
The group identifying unit 13 stores therein association between the accompanying information and a group. Herein, a case will be described in which the type of a host device that has output the processing request is used as the accompanying information. If the processing request is output from the CPU 4, it is presumed that the processing request is important; therefore, the group identifying unit 13 causes a group with a channel having a higher performance to process the processing request. On the contrary, for example, if the processing request is output from a local area network (LAN), it is presumed that the degree of importance of the processing request is low; therefore, the group identifying unit 13 causes a group with a channel having a lower performance to process the processing request.
Further, while the type of the host device serving as an output source of the processing request is used as the accompanying information, the accompanying information is not specifically limited as long as it is information that enables processing requests to be distinguished from one another. For example, the NAND controller 10 may use a logical address, a data size, and a type of software of an issuance source as the accompanying information, and may select a group by analyzing these pieces of accompanying information.
Furthermore, with reference to
A processing request 421 in
On the contrary, in the case of the write request, accompanying information is stored in a header or the like as in the processing requests 422 and 423.
The NAND controller 10, upon receiving the processing request 421, makes copies of the request and transfers the requests to all of the groups #1 to 3. Each of the groups performs an address translation, and if an address designated by the processing request 421 has been registered, the NAND controller 10 reads data from the NAND flash memory 7 having a storage area corresponding to the address designated by the processing request 421. On the contrary, if the address designated by the processing request 421 has not been registered, the group that has received the request determines that the processing request is not directed thereto and discards the request.
Further, upon receiving the processing request 422 or 423, the NAND controller 10 acquires the accompanying information of the processing request 422 or 423. Then, the NAND controller 10 selects a group to which data is written from among the groups #1 to #3 according to the acquired accompanying information.
Even in this case, the NAND controller 10 selects a group based on the accompanying information; therefore, the CPU 4 need not recognize the state of division of the groups.
As described above, the group identifying unit 13 according to the present embodiment can determine a group that processes the processing request according to the accompanying information.
As described above, in the information processing apparatuses according to the present embodiment and modified example, the NAND controller selects a group according to a processing request. Therefore, it becomes possible to centralize processes related to the groups into the NAND controller side and suppress a change to be added to processes performed by the CPU or the like, enabling to easily improve the performance of the entire system and the resource use efficiency.
Next, a third embodiment will be described. An information processing apparatus according to the present embodiment differs from the first embodiment in that a NAND controller automatically configures groups.
Each of the NAND flash memories 7a to 7f has a specification storage area in which specification information on itself is stored. The specification information is, for example, a manufacturer, a speed, or the like. Further, while the NAND flash memories are described as examples in the present embodiment, other memory devices may have specification storage areas. For example, in the case of a double-data-rate (DDR) 3 DIMM, a serial presence detect (SPD) serves as the specification storage area. Namely, in the case of other memory devices, the same advantageous effects as those of the present embodiment can be obtained as long as they have specification storage areas.
The configuration analyzing unit 16 acquires specification information from each of the specification storage areas of the NAND flash memories 7a to 7f.
As illustrated in the device information table 161, the configuration analyzing unit 16 acquires, as the specification information on the NAND flash memories 7a to 7f, a manufacture, a type, a speed, and a volume per page for example. Then, the configuration analyzing unit 16 performs setting of a group configuration of each of the channels in the configuration register 11 by using the acquired information.
The configuration analyzing unit 16 sets a variable i, which represents a group number of a group #i, to an initial value of 1 (Step S101).
Subsequently, the configuration analyzing unit 16 sets all of the channels 8a to 8f as processing targets (Step S102).
Then, the configuration analyzing unit 16 selects one of the channels 8a to 8f set as the processing targets, and sets the selected channel as a reference channel (Step S103).
Subsequently, the configuration analyzing unit 16 extracts channels whose performance difference from the reference channel is equal to or smaller than a predetermined threshold from among the processing target channels (Step S104).
Then, the configuration analyzing unit 16 sets the reference channel and the extracted channels as channels belonging to the group #i (Step S105).
Subsequently, the configuration analyzing unit 16 excludes the reference channel and the extracted channels from the processing targets (Step S106).
Then, the configuration analyzing unit 16 increments the value of the variable i representing the group number by one (Step S107).
Subsequently, the configuration analyzing unit 16 determines whether a processing target channel has been left (Step S108). If the processing target channel has been left (Step S108: YES), the configuration analyzing unit 16 determines whether i is equal to or smaller than the number of all of the groups (Step S109). In other words, the configuration analyzing unit 16 determines whether there is a vacant group for which setting is possible.
If it is equal to or smaller than the number of all of the groups (Step S109: YES), the configuration analyzing unit 16 returns the process to Step S103.
On the contrary, if it is greater than the number of all of the groups (Step S109: NO), the configuration analyzing unit 16 resets the configuration register 11 (Step S110).
Then, the configuration analyzing unit 16 increases the threshold for determining a performance difference from the reference channel (Step S111), that is, eases the determination condition, and returns the process to Step S101.
On the contrary, if no processing target channel has been left (Step S108: NO), the configuration analyzing unit 16 ends the setting of the group configuration in the configuration register 11.
By performing the above described processes, the configuration analyzing unit 16 can perform setting such that all of the channels 8a to 8f belong to any of the groups without duplication. Further, in the present embodiment, when all of the channels are classified into a smaller number of groups than the number of all of the groups, the process of setting the group configuration is terminated; however, other methods may be employed. For example, if the number of groups is smaller than the number of all of the groups when no processing target is left, the configuration analyzing unit 16 may set a stricter threshold and perform a group configuration again. The configuration analyzing unit 16 corresponds to an example of a “group configuring unit”.
The configuration register 11 stores therein the group configuration set by the configuration analyzing unit 16.
The group identifying unit 13 and the data processing units 14a to 14c receive the group configuration set by the configuration analyzing unit 16 from the configuration register 11, and processes data processing requests.
As described above, the information processing apparatus according to the present embodiment automatically generates a group configuration from the specifications of storage devices. Therefore, an operator need not determine the group configuration, so that loads on the operator can be reduced. Further, if a storage device is changed, the group configuration is automatically changed, so that it becomes possible to easily improve the performance of the entire system and the resource use efficiency.
Furthermore, in the above description, a function to automatically generate a group configuration is provided on the NAND controller; however, the group configuration may be generated by others rather than the NAND controller. For example, it may be possible to cause an external computer to acquire performance information on channels, cause the computer to run a program that executes the flow in
Next, a fourth embodiment will be described. An information processing apparatus according to the present embodiment differs from the first embodiment in that groups are configured such that duplication of channels among multiple groups is allowed. The NAND controller 10 according to the present embodiment is represented by the block diagram in
The configuration register 11 stores therein, as illustrated in
In
The configuration register 11 stores therein which of the channels 8a to 8f are duplicated among the groups. Then, the configuration register 11 notifies the group identifying unit 13 and the channel selecting unit 145 of the group configuration and information on the duplicated channel 8.
The channel selecting unit 145, when there is the duplicated channel 8 in the group to which the data processing unit 14 provided therewith belongs, preferentially uses channels that are not duplicated. If it is difficult to perform data processing by only the channels 8 that are not duplicated, the channel selecting unit 145 uses the duplicated channel 8 for the data processing.
For example, the channel selecting unit 145 of the data processing unit 14c in
As described above, by allowing duplication of the channels among the groups, it becomes possible to use the duplicated channel as a buffer area when a processing load on each of the groups increases, enabling to further improve the resource use efficiency.
Further, in the above description, the explanation has been given based on the state in which the configuration register 11 stores therein a group configuration in which duplication of the channels 8 among the groups is allowed; however, even in the present embodiment, it is possible to automatically set the group configuration as in the third embodiment.
Therefore, setting of a group configuration when duplication of the channels 8 is allowed will be described below. Herein, similarly to the third embodiment, a case will be described in which the configuration analyzing unit 16 in
The configuration analyzing unit 16 sets a variable i, which represents a group number of the group #i, to an initial value of 1 (Step S201).
Subsequently, the configuration analyzing unit 16 sets all of the channels 8a to 8f as processing targets (Step S202).
Then, the configuration analyzing unit 16 selects one of the channels 8a to 8f set as the processing targets, and sets the selected channel as a reference channel (Step S203).
Subsequently, the configuration analyzing unit 16 extracts channels whose performance difference from the reference channel is equal to or smaller than a predetermined threshold from among all of the channels (Step S204).
Then, the configuration analyzing unit 16 sets the reference channel and the extracted channels as channels belonging to the group #i (Step S205).
Subsequently, the configuration analyzing unit 16 excludes the reference channel and the extracted channels from the processing targets (Step S206).
Then, the configuration analyzing unit 16 increments the value of the variable i representing the group number by one (Step S207).
Subsequently, the configuration analyzing unit 16 determines whether a processing target channel has been left (Step S208). If the processing target channel has been left (Step S208: YES), the configuration analyzing unit 16 determines whether i is equal to or smaller than the number of all of the groups (Step S209).
If it is equal to or smaller than the number of all of the groups (Step S209: YES), the configuration analyzing unit 16 returns the process to Step S203.
On the contrary, if it is greater than the number of all of the groups (Step S209: NO), the configuration analyzing unit 16 resets the configuration register 11 (Step S210).
Then, the configuration analyzing unit 16 increases the threshold for determining a performance difference from the reference channel (Step S211), that is, eases the determination condition, and returns the process to Step S201.
On the contrary, if there is no processing target channel (Step S208: NO), the configuration analyzing unit 16 ends the setting of the group configuration in the configuration register 11.
By performing the above described processes, the configuration analyzing unit 16 can automatically set a group configuration while allowing duplication of all of the channels 8a to 8f.
As described above, by automatically generating a group configuration while allowing duplication of the channels, it becomes possible to more flexibly improve the resource use efficiency.
According to an embodiment of the information processing apparatus, the storage device control circuit, and the storage device control method disclosed in this specification, it becomes possible to improve the processing performance and the resource use efficiency.
All examples and conditional language recited herein are intended for pedagogical purposes of aiding the reader in understanding the invention and the concepts contributed by the inventor to further the art, and are not to be construed as limitations to such specifically recited examples and conditions, nor does the organization of such examples in the specification relate to a showing of the superiority and inferiority of the invention. Although the embodiments of the present invention have been described in detail, it should be understood that the various changes, substitutions, and alterations could be made hereto without departing from the spirit and scope of the invention.
Number | Date | Country | Kind |
---|---|---|---|
2013-227724 | Oct 2013 | JP | national |