JP2020102237A - Information processing system - Google Patents

Abstract

To provide an information processing system capable of shortening a start time.SOLUTION: An information processing system of an embodiment comprises a host device and a storage device connected to the host device. The storage device includes a nonvolatile semiconductor memory storing data transmitted from the host device, and a controller controlling communication with the host device, and data writing operation to and data reading operation from the nonvolatile semiconductor memory. The controller separates a space of logical addresses associated with data into a plurality of subspaces for management, and manages start information required when the storage device is started and required to convert logical addresses and physical addresses of the nonvolatile semiconductor memory so as to enable identification of start priorities per subspace.SELECTED DRAWING: Figure 20

Description

Embodiments of the present invention relate to an information processing system.

There is an initialization process that is required for each unit capacity when the storage device is started, but with the recent increase in the capacity of the storage device, the amount of initialization process that is required for the entire storage device increases, and the startup time is long. It has become to. In addition, a technique for dividing a logical address space of a storage device into a plurality of partial spaces and managing the divided space has been widely used.

JP, 2009-176200, A

The problem to be solved by the present invention is to provide an information processing system capable of shortening the startup time.

In order to achieve the above object, the information processing system according to the embodiment includes a host device and a storage device connected to the host device. The storage device includes a non-volatile semiconductor memory that stores data transmitted from the host device, a controller that controls communication with the host device, and controls a data write operation and a data read operation to the non-volatile semiconductor memory, The controller divides the logical address space associated with the data into a plurality of partial spaces and manages the partial address space. The controller is startup information necessary for booting the storage device, and converts the logical address and the physical address of the nonvolatile semiconductor memory. The information required for the above is managed so that the starting priority for each subspace can be identified.

3 is a block diagram illustrating a configuration of a storage device according to an embodiment. FIG. It is a figure explaining the structure of the physical page and physical block of embodiment. It is a figure explaining the structure of the logical page and logical block of embodiment. FIG. 3 is a diagram illustrating a logical address space divided into subspaces in the embodiment. It is a figure explaining the memory configuration of an embodiment. It is a figure explaining control of read/write of user data using translation information in an embodiment. It is a figure explaining an example of a data structure of position information of an embodiment. It is a figure explaining an example of the log of an embodiment. It is a figure explaining the 1st example of the allocation method of the log area of an embodiment. It is a figure explaining the 2nd example of the allocation method of the log area of an embodiment. It is a figure explaining the 3rd example of the allocation method of the log area of an embodiment. It is a figure explaining the 4th example of the allocation method of the log area of an embodiment. It is a figure explaining composition of an NS management table of an embodiment. It is a figure explaining operation which restores position information in an embodiment. It is a figure explaining the cache operation of the translation information to RAM in an embodiment. It is a figure explaining an example of the completion report of starting to the host in an embodiment. It is a flow chart explaining control of FW at the time of starting of an embodiment. FIG. 6 is a diagram illustrating an example of an operation of a command that specifies a startup priority in the embodiment. It is a figure explaining an example of the starting priority and starting completion report for every namespace in an embodiment. It is a figure explaining another example of the starting priority and starting completion report for every namespace in an embodiment. FIG. 7 is a diagram illustrating an example of an operation when there is an overlap in designated startup priorities in the embodiment. It is a figure explaining another example of operation|movement when there exists duplication in designated starting priority in embodiment. 7 is a flowchart illustrating FW control when a startup priority designation command is received in the embodiment. FIG. 7 is a diagram illustrating an example of an operation of a command for inquiring about a startup priority according to the embodiment.

Hereinafter, a storage device according to an embodiment will be described with reference to the drawings. In addition, in the following description, common reference numerals are given to components having the same function and configuration.

FIG. 1 is a block diagram illustrating the configuration of the storage device according to the embodiment.

The storage device 1 stores a controller 10 that controls the storage device 1 as a whole, a non-volatile storage medium 20 that stores data, and a temporary storage of various information groups, and between the host 2 and the non-volatile storage medium 20. A RAM (Random Access Memory) 30 used as a buffer for temporarily storing data, and a CPU (Central Processing Unit) 40 that controls the entire storage device 1 based on FW (Firmware) are included.

In the description of the present embodiment, the host 2 is a computer that supports an interface of the NVMe (NVM Express) (registered trademark) standard, but other standards, for example, SAS (Serial Attached SCSI) standard and SATA (Serial ATA) standard. It may be a computer that supports the interface.
The controller 10 is, for example, a semiconductor integrated circuit configured as an SoC (System on a Chip).

In the present embodiment, the non-volatile storage medium 20 is a NAND flash memory and is configured to include a plurality of NAND flash memory chips 100 (hereinafter referred to as memory chips 100), but other types of magnetic disks such as a magnetic disk may be used. A non-volatile storage medium may be used. Hereinafter, the nonvolatile storage medium 20 will be referred to as a NAND flash memory 20 (hereinafter, also simply referred to as NAND). The NAND flash memory 20 of the present embodiment has, for example, a 16-channel (Ch) memory chip 100. Hereinafter, each of the memory chips 100 will be referred to as a memory chip Ch0 to a memory chip Ch15. The number of channels may be more or less than 16.

In the present embodiment, the RAM 30 is a DRAM (Dynamic Random Access Memory), but other types of volatile memory such as an SRAM (Static Random Access Memory) may be adopted.

The RAM 30 and the CPU 40 may be built in the controller 10 instead of being separate semiconductor integrated circuits. Further, in the following description, some or all of the functions executed by the FW can also be executed by a dedicated HW (Hardware), and some or all of the functions executed by the HW should be executed by the FW. Is also possible.

The controller 10 includes a host interface (IF) control unit 200 that controls communication with the host 2 and manages a plurality of connected hosts 2, a buffer control unit 210 that controls read/write of the RAM 30, and a NAND flash memory 20. And a NAND control unit 220 for controlling read/write of. The host IF control unit 200 can report execution results of various commands received from the host 2 to the host 2 according to an instruction from the CPU 40 based on FW.

Next, configurations of the physical page 300 and the physical block 310 of the memory chip 100 will be described with reference to FIGS. 2A and 2B.

As shown in FIG. 2A, the minimum management unit for reading/writing data from/to the memory chip 100 is called a cluster 320. In this embodiment, the size of the cluster 320 is 4 kB. Further, the smallest unit of circuit configuration capable of reading/writing data at once inside the memory chip 100 is called a physical page 300. In this embodiment, the size of the physical page 300 is 16 clusters (4 kB×16 clusters=64 kB).

In addition, as shown in FIG. 2B, a unit of the minimum circuit configuration capable of erasing data of the memory chip 100 is called a physical block 310. In this embodiment, the size of the physical block 310 is 256 clusters, that is, 16 physical pages (64 kB×16 physical pages=1024 kB). It should be noted that the size of each of these units is an example, and is not limited to these values.

Next, the configurations of the logical page 400 and the logical block 410 will be described with reference to FIGS. 3A and 3B.

In this embodiment, as shown in FIG. 3A, a set including one physical page 300 of each of the memory chips Ch0 to Ch15 is a logical page 400. The controller 10 controls reading/writing of data to/from the NAND flash memory 20 using the logical page 400 as a logical unit. Further, as shown in FIG. 3B, the controller 10 controls the erase of the data of the NAND flash memory 20 using the logical block 410 which is the data of 16 logical pages 400 as a logical unit. That is, the controller 10 performs the erase process for each logical block 410, not for each physical block 310, which is the minimum unit for erasing data in the memory chip 100.

Next, the subspace will be described with reference to FIG.

The logical address (Logical Block Address, LBA) provided by the storage device 1 to the host 2 and the physical address of the NAND flash memory 20 in which the user data is stored correspond to each other on a one-to-one basis. Here, the user data refers to the data transmitted from the host 2 together with the write request. Hereinafter, a logical address corresponding to a physical address on a one-to-one basis is referred to as a drive LBA.

The drive LBA is divided into a plurality of partial spaces and provided to the host 2. The subspace is called a namespace in the NVMe standard, and the subspace is called a Logical Unit in the SCSI standard. In the following, the subspace will be referred to as a namespace. Each namespace is identified by a namespace ID (NSID). In this embodiment, the drive LBA is divided into (n+1) namespaces identified as NSID0 to NSIDn.

Each namespace can be treated as a storage area independent of the host 2, and the drive LBA of the storage device 1 is designated by the NSID and the host LBA starting from 0 for each namespace. Here, the host LBA is a logical address specified by the host 2 when a command is issued in order to specify the storage destination of the data in the storage device 1.

The host IF control unit 200 performs conversion between the host LBA designated by the command and the drive LBA. For example, as shown in FIG. 4, when the host 2 accesses the host LBA=0 of NSID2, the host IF control unit 200 treats the drive LBA=B as being accessed.

Each namespace is created by a command from the host 2. That is, the host 2 issues a command to the storage device 1 to specify the NSID of the namespace to be created and the storage capacity to be assigned to that namespace. It should be noted that the drive LBA may not have a namespace assigned thereto and may have an empty area to which a namespace cannot be assigned.

Further, when the storage device 1 is connected to a plurality of hosts 2, the name space accessible from each host 2 may be limited. For example, host #0 can access only NSID0 and NSID2, and host #1 can access only NSID3.

Next, the memory configuration of the present embodiment will be described with reference to FIG.

The NAND flash memory 20 stores user data 500, a translation information group 520a, a log 530, and an NS management table 540.

The translation information group 520a includes translation information 510a, and each translation information 510a is information indicating the correspondence between the drive LBA and the physical address. The log 530 is a log for recording a change in the correspondence between the drive LBA and the physical address. The NS management table 540 is a table for managing the activation priority order for each namespace.

The controller 10 includes (n+1) user data areas in which the user data 500 is stored, translation information group areas in which the translation information group 520a is stored, (m+1) log areas in which the log 530 is stored, and NS. The NAND flash memory 20 is managed including the NS management table area in which the management table 540 is stored. Each user data area has a one-to-one correspondence with each namespace. Although one namespace corresponds to each log area, each namespace may correspond to a plurality of log areas. That is, for example, the log area 0 does not correspond to both the NSID 0 and the NSID 1, but the NSID 0 may correspond to the log area 0 and the log area 1.

Although the controller 10 manages one logical page 400 as including both the user data area and the log area, the logical page 400 dedicated to the user data area and the logical page 400 dedicated to the log area may exist. ..

If the controller 10 can recognize the relationship between the host LBA and the drive LBA corresponding to each namespace, the controller 10 does not have to manage the user data area in one-to-one correspondence with the namespace. Good. For example, the user data 500 of a plurality of namespaces may be stored in one user data area, and the log 530 of the namespace may be stored in a log area allocated for each namespace.

Further, the controller 10 may include the NSID in the log 530 stored in the log area, instead of managing the log area in association with the namespace. In this case, in the following description, for example, the log area corresponding to NSID0 refers to a set of log areas in which the log 530 having NSID0 is stored.

In the RAM 30, each translation information 510a is copied from the translation information group area of the NAND flash memory 20 and stored as translation information (cache) 510b. Hereinafter, a set of translation information (cache) 510b stored in the RAM 30 will be referred to as a translation information group (cache) 520b. The RAM 30 also stores a position information group 550 for each namespace. Although details will be described later, the position information group 550 includes position information 560, and each position information 560 records information indicating a storage destination of the translation information 510a and the translation information (cache) 510b.

Next, referring to FIG. 6, the conversion from the host LBA to the physical address using the translation information 510 will be described. In this embodiment, the translation information 510a in the translation information group 520a and the translation information (cache) 510b in the translation information group (cache) 520b have the same data structure. Note that the translation information 510a and the translation information (cache) 510b may have different data structures as long as the translation information 510a and the translation information (cache) 510b represent the same content. For example, the translation information 510a may store compressed translation information data.

In FIG. 6 and the following description, the translation information group 520a and the translation information group (cache) 520b are collectively referred to as the translation information group 520, and the translation information 510a and the translation information (cache) 510b are collectively referred to as the translation information. It is described as 510.

When the storage device 1 receives a command from the host 2 (S100), the host IF control unit 200 extracts the NSID and the host LBA from the command (S101) and calculates the drive LBA using these (S102).

In this embodiment, the physical address of the NAND flash memory 20 is recorded in each translation information 510, and each translation information 510 is arranged in the order of the drive LBA associated with the user data 500. That is, the CPU 40 can acquire the physical address associated with each user data 500 by searching the translation information group 520 with the drive LBA as an index based on the FW (S103). The translation information does not have to be arranged in the order of the drive LBA as long as the CPU 40 can recognize the correspondence between the drive LBA and the translation information.

In the present embodiment, the translation information 510 corresponding to the drive LBA that is not associated with the user data 500 records information indicating “Unmapped”. In addition, one translation information 510 may record the correspondence between a plurality of drive LBAs and a plurality of physical addresses.

Each translation information 510 is stored in the NAND flash memory 20 while the power for operating the storage device 1 is not supplied. Since the access to the NAND flash memory 20 is generally slower than the access to the RAM 30, each translation information 510 can be copied (cached) to the RAM 30 as soon as possible after the power supply to the storage device 1. desirable. Note that the translation information 510 may be copied to the RAM 30 by using the access to the corresponding drive LBA as a trigger, instead of copying it immediately after power is supplied.

Next, an example of the data structure of the position information group 550 in the RAM 30 will be described with reference to FIG. 7.

FIG. 7 shows the position information group 550 of NSID0, but the position information groups 550 of other namespaces have the same configuration. The position information group 550 includes a plurality of position information 560. Each position information 560 corresponds to one translation information 510 on a one-to-one basis, and the location where each translation information 510 is stored (that is, the RAM 30 used as a cache or the NAND flash memory 20) and its physical address are It is recorded.

Each position information 560 is arranged in the same order as the corresponding translation information 510, that is, in the order of the drive LBA associated with the user data 500. Therefore, the CPU 40 searches the position information group 550 using the drive LBA as an index to obtain the physical address of the RAM 30 or the NAND flash memory 20 in which the translation information 510 corresponding to each user data 500 is stored. ..

In the example of FIG. 7, the drive LBA corresponding to the host LBA “L0” of NSID0 is “La”, and the position information 560 corresponding to this is the physical address “Pb” of the NAND flash memory 20. In the area of the physical address “Pb” of the NAND flash memory 20, “physical address=Pa” of the NAND flash memory 20 is stored as the translation information 510a. That is, the user data 1 corresponding to the host LBA “L0” of NSID0 is stored in the area of the physical address “Pa” of the NAND flash memory 20.

The drive LBA corresponding to the host LBA “L1” of NSID0 is “Lc”, and the position information 560 corresponding to this is the physical address “Pd” of the RAM 30. In the area of the physical address “Pd” of the RAM 30, “physical address=Pc” of the NAND flash memory 20 is stored as translation information (cache) 510b. That is, the user data 2 corresponding to the host LBA “L1” of NSID0 is stored in the area of the physical address “Pc” of the NAND flash memory 20.

Next, an example of the log 530 will be described with reference to FIG.

In the example shown in FIG. 8, the drive LBA corresponding to the host LBA “L1” of NSID 0 is “Lc”, and in the initial state of FIG. 8, the user data 2 corresponding to the drive LBA “Lc” is the NAND flash memory 20. Is stored in the area of the physical address “Pc”. Further, translation information (cache) 510b indicating the relationship between the drive LBA “Lc” and the physical address “Pc” is stored in the physical address “Pd” of the RAM 30.

In this state, the user data 3 is sent together with the write request from the host 2 to the host LBA "L1" of NSID0. At this time, the CPU 40 controls the controller 10 to store the user data 3 in the area of the physical address “Pe” of the NAND flash memory 20 where valid data is not stored (S200). Further, the CPU 40 adds new translation information (cache) 510b (physical address "Pe") to the physical address "Pf" of the RAM 30 (S201). The CPU 40 rewrites the position information 560 corresponding to the drive LBA “Lc” from the physical address “Pd” of the RAM 30 to “Pf” (S202).

The added translation information (cache) 510b (physical address "Pe") is stored in the translation information group 520a of the NAND flash memory 20 by the time the power supply to the storage device 1 is cut off at the latest (S203). .. Here, it is assumed that the translation information (cache) 510b (physical address “Pe”) corresponding to the drive LBA “Lc” is stored in the physical address “Pk” of the NAND flash memory 20.

The CPU 40 stores the log 530 indicating that the translation information (cache) 510b corresponding to the drive LBA “Lc” is stored at the physical address “Pk” in the log area 0 (S204).

The logs 530 are sequentially stored in the log areas 0 to m as the translation information (cache) 510b corresponding to each namespace is stored. In each of the log areas 0 to m, logs 530 are stored in chronological order from oldest to newest. As described above, only the log 530 of one namespace is stored in each log area 0 to m, and the logs 530 of a plurality of namespaces are not mixed in one log area. The log 530 is used to restore the position information group 550 when the storage device 1 is activated, as will be described later.

Note that, for example, when the power supply to the storage device 1 is cut off, not all the logs 530 may be stored in the log area at once. For example, the log 530 may be stored every time the total size of the log 530 associated with the storage of the translation information (cache) 510b for a plurality of times becomes one logical page. In this case, the CPU 40 needs to manage that the translation information (cache) 510b that is not reflected in the log 530 is stored, for example, by adding a flag to each of the translation information (cache) 510b.

Next, an example of the operation of allocating one logical block 410 to one log area will be described with reference to FIGS. 9a and 9b. One log 530 in FIG. 9a shows a set of logs 530 for one logical page.

The log 530 is collected in units of logical blocks 410 for each namespace, and is stored in the order of the numbers of the logical pages 400 in each logical block 410. In the example shown in FIG. 9a, the logical block #0 is allocated as the log area 0, and the log 530 of NSID0 is stored therein. The logical block #1 is assigned as the log area 1, and the log 530 of NSID2 is stored therein. Further, the log 530 of NSID1 is stored in the logical block #2 as the log area 2, the log 530 of NSID0 is stored in the logical block #3 as the log area 3, and the log 530 of NSID1 is stored in the logical block #4 as the log area 4. The log 530 is stored.

As a result, as shown in FIG. 9B, the numbers (log block numbers) of the logical blocks 410 storing the log 530 of NSID0 are #0 and #3, and the log block numbers of NSID1 are #2 and #4. The log block number of NSID2 is #1, respectively.

Next, an example of operation of dividing one logical block 410 into a plurality of areas and allocating each divided area to one log area will be described with reference to FIGS. 10a and 10b. In FIG. 10A, as in FIG. 9A, one log 530 shows a set of logs 530 for one logical page.

Each logical block 410 is allocated as one log area for every eight logical pages 400. Hereinafter, a set of logical pages #0 to #7 is referred to as a lower page, and a set of logical pages #8 to #15 is referred to as an upper page. Note that the number of logical pages 400 that can be assigned as one log area is not limited to eight.

In the example shown in FIG. 10A, the Lower Page of the logical block #0 is assigned as the log area 0, and the log 530 of NSID 0 is stored therein. The upper page of the logical block #0 is assigned as the log area 1, and the log 530 of the NSID 2 is stored therein. The lower page of the logical block #1 is allocated as the log area 2, and the log 530 of NSID0 is stored therein. The Upper Page of the logical block #1 is assigned as the log area 3, and the log 530 of NSID1 is stored therein.

As a result, as shown in FIG. 10B, the numbers (log block numbers) of the logical blocks 410 storing the log 530 of NSID 0 are #0 (L) and #1 (L), and the log block number of NSID 1 is #. The log block numbers of 1 (U) and NSID2 are #0 (U), respectively.

Next, an example of an operation of allocating a set including one logical page 400 of each logical block 410 as a log area of one namespace will be described with reference to FIGS. 11a and 11b. In FIG. 11A, as in FIG. 9A, one log 530 shows a set of logs 530 for one logical page.

The log 530 is stored in the logical page 400 at a fixed position (number) in each logical block 410 for each namespace. For example, the logical page #0 of each logical block 410 is allocated as the log area 0 for storing the log 530 of NSID0. Further, the logical page #1 of each logical block 410 is allocated as the log area 1 for storing the log 530 of NSID1. Furthermore, the logical page #2 of each logical block 410 is stored as the log area 2 for storing the log 530 of NSID2, and the logical page #3 of each logical block 410 is stored as the log area 3 for storing the log 530 of NSID3. Assign each.

In the example illustrated in FIG. 11A, the log 530 of NSID0 is in the logical page #0 of the logical blocks #0 to #4, the log 530 of NSID1 is in the logical page #1 of the logical blocks #0 to #2, and the log 530 of NSID2 is The logical page #2 of the logical blocks #0 to #2 and the log 530 of NSID3 are stored in the logical page #3 of the logical blocks #0 to #3, respectively.

As a result, as shown in FIG. 11B, the numbers (log block numbers) of the logical blocks 410 storing the log 530 of NSID0 are #0 to #4, the log block numbers of NSID1 are #0 to #2, and NSID2. The log block numbers are #0 to #2, and the log block numbers of NSID3 are #0 to #3.

Although a set including one logical page 400 of each logical block 410 is assigned as one log area in the above, a set including a plurality of logical pages 400 of each logical block 410 is set as one log area. It is also possible to assign. For example, the logical page #0 and the logical page #1 of each logical block 410 are assigned as the log area 0 for storing the log 530 of the NSID 0, and the logical page #2 and the logical page #3 of each logical block 410 are set as the log area 1. It is also possible to allocate the log 530 of NSID1 for storage and allocate the logical page #4 and logical page #5 of each logical block 410 as the log area 2 for storage of the log 530 of NSID2.

Next, an example of an operation of dividing one logical page 400 into a plurality of areas and allocating each divided area to one log area will be described with reference to FIGS. 12a and 12b. In FIG. 12 a, one log 530 does not represent a set of logs 530. It is assumed that each logical page 400 can store up to 16 logs 530.

In the example shown in FIG. 12a, the area starting from the address Pa of the logical page #0 is allocated as the log area 0, and eight logs 530 of NSID0 are stored therein. The area starting from the address Pb of the logical page #0 is allocated as the log area 1, and two logs 530 of NSID2 are stored therein. The area starting from the address Pc of the logical page #0 is allocated as the log area 2, and four logs 530 of NSID1 are stored therein.

Further, the area starting from the address Pd of the logical page #1 is allocated as the log area 0, and four logs 530 of NSID0 are stored therein. The area starting from the address Pe of the logical page #1 is assigned as the log area 3, and seven logs 530 of NSID3 are stored therein.

As a result, as shown in FIG. 12B, the numbers (log page numbers) of the logical pages 400 storing the log 530 of NSID 0 are #0 (start address=Pa, length=8) and #1 (start address= Pd, length=4), the log page number of NSID1 is #0 (start address=Pc, length=4), and the log page number of NSID2 is #0 (start address=Pb, length=2). , NSID3, the log page number is #1 (start address=Pe, length=7). Information on the start address and length of each log area as shown in FIG. 12b may be stored in the logical page 400 as header information.

In the above description, the log area of each logical page 400 is divided into a plurality of areas, and the log 530 for each namespace is stored in each area. On the other hand, in the storage of the user data 500, the user data area of each logical page 400 may be divided into a plurality of areas, and the user data 500 for each namespace may be stored in each area. User data 500 in multiple namespaces may be stored therein.

Next, the configuration of the NS management table 540 will be described with reference to FIG.

In the NS management table 540, the physical address of the RAM 30 in which the position information group 550 is stored, the number (log block number) of the logical block 410 in which the log 530 is stored, and the startup priority order described later are namespaces. It is recorded for each.

The NS management table 540 is stored in the NAND flash memory 20 while the power for operating the storage device 1 is not supplied, but after the power is supplied to the storage device 1, prior to other information. The data is copied to the RAM 30 or a memory area (not shown) in the CPU 40 or the controller 10.

Even when the logical page 400 at a fixed position (number) in each logical block 410 is allocated as a log area as shown in FIG. 11a, the logical page number allocated as a log area should be fixed in advance for each namespace. For example, the logical page number storing the log 530 may not be recorded in the NS management table 540.

Further, when the log area is allocated as shown in FIG. 12, the NS management table 540 records the log page number instead of the log block number. Alternatively, when the header information is stored in the logical page 400 as described above, the log block number may be recorded in the NS management table 540. In this case, the CPU 40 can determine at which address in each logical block 410 the log 530 of which namespace is stored by reading the header information when reading the log 530.

Next, with reference to FIG. 14, an operation of restoring the position information group 550 when the storage device 1 is activated will be described.

When the storage device 1 is activated, the CPU 40 reads the activation priority order from the NS management table 540, and reads the log 530 from the namespace having the higher activation priority order according to the log block number. As described above, the log 530 is stored in the order of saving the translation information (cache) 510b. Therefore, the latest position information 560 corresponding to some of the drive LBAs can be restored by reading the log 530 in the storage order.

In the example illustrated in FIG. 14, the CPU 40 reads the log 530 in the log area 0 in the order of storage, and restores the physical address “Pb” as the position information 560A from the log 530A corresponding to the drive LBA “La”. Further, the physical address “Pd” is restored as the position information 560C from the log 530B corresponding to the drive LBA “Lc”. The position information 560C is overwritten in a later operation as described later. Next, the physical address “Pc” is restored as the position information 560D from the log 530C corresponding to the drive LBA “Lb”. A physical address “Pf” corresponding to the drive LBA “Lc” is recorded in the log 530D read next. In this case, the position information 560C is overwritten from the physical address "Pd" to the physical address "Pf". A physical address “Pe” corresponding to the drive LBA “Lf” is recorded in the log 530E read thereafter, and the position information 560D is restored as the physical address “Pe”.

At this point, since each translation information 510 has not been copied from the NAND flash memory 20 to the RAM 30, the location where each translation information 510 is stored in the restored position information 560 is the NAND flash memory 20. Become.

Even if the logs 530 are read out in the reverse order of the storage order, the latest position information 560 corresponding to some drive LBAs can be restored. In this case, the physical address recorded in the log 530 that is read later and corresponds to the same drive LBA is not reflected in the position information 560. In the example shown in FIG. 14, as for the log 530 corresponding to the drive LBA “Lc”, the log 530D is read first and the log 530B is read later. The physical information “Pf” recorded in the log 530D is reflected in the position information 560C, but the physical address “Pd” recorded in the log 530B is not reflected.

When the log area is not managed for each namespace, the CPU 40 determines which namespace corresponds to the log 530 based on the NSID included in the log 530. In this case, if the read log 530 corresponds to the name space to be restored, the CPU 40 uses the log 530 to restore the position information group 550. On the other hand, if the read log 530 does not correspond to the namespace to be restored, the CPU 40 reads the next log 530 without using the log 530.

In this embodiment, the location information group 550 is restored for each namespace according to the activation priority order of the NS management table 540. That is, after starting the storage device 1, the CPU 40 first restores the position information group 550 of the namespace having the highest starting priority. When this is completed, the CPU 40 sequentially starts the restoration of the position information group 550 of the namespace with the second highest activation priority.

When all the logs 530 of a certain namespace are read and the restoration of the position information group 550 is completed, the user data 500 can be read/written for that namespace. That is, in the present embodiment, the restoration of the position information group 550 is completed, and thus the activation of the namespace is completed, so that each log 530 can be considered as the activation information of the namespace.

When the translation information 510 is stored in the NAND flash memory 20, it may be stored separately for each namespace. In this case, after the storage device 1 is activated, the location information group 550 of a certain namespace is restored, and the translation information 510 of the relevant namespace is copied to the RAM 30 to complete the activation of the relevant namespace.

Next, the cache operation of the translation information 510 in the RAM 30 will be described with reference to FIG.

In the present embodiment, the translation information 510a is copied from the NAND flash memory 20 to the RAM 30 as translation information (cache) 510b after the restoration of the position information group 550 for all namespaces is completed. When the copy of the translation information 510 to the RAM 30 is completed, the reference destination of the corresponding position information 560 becomes the RAM 30.

It should be noted that the translation information 510 may be copied to the RAM 30 after the restoration of the position information group 550 of the entire storage device is completed, and the translation information of the namespace is updated every time the restoration of the position information group 550 of a certain namespace is completed. 510 may be copied to RAM 30.

Next, an example of the activation completion report to the host 2 will be described with reference to FIGS. 16a to 16c.

As shown in FIG. 16a, NSID0 is the third highest, NSID1 is the second highest, NSID2 is the second highest, and NSID3 is the fourth highest, as shown in FIG. 16a.

In the example shown in FIG. 16b, when power is supplied to the storage device 1, restoration of the position information group 550 of NSID2 having the first boot priority is started. Then, when the restoration of the position information group 550 of NSID2 is completed, the storage device 1 reports to the host 2 that the activation of NSID2 is completed. After that, according to the activation priority of the NS management table 540, the restoration of the location information group 550 of NSID1 and the activation completion report to the host 2, the restoration of the location information group 550 of NSID0 and the activation completion report to the host 2, the location information of NSID3 The restoration of the group 550 and the report of the completion of activation to the host 2 are performed. When the restoration of the position information group 550 is completed for all the namespaces, the fact that the storage device 1 has been activated is reported to the host 2.

In FIG. 16B, the storage device 1 voluntarily reports to the host 2 that the activation of the relevant namespace has been completed upon completion of the restoration of the position information group 550 of each namespace. On the other hand, in FIG. 16c, as a response to the inquiry command from the host 2, whether the activation is completed or not is reported.

That is, as illustrated in FIG. 16c, with the supply of power to the storage device 1, restoration of the position information group 550 of the NSID 2 having the first boot priority is started. When a command inquiring whether or not the activation of NSID2 is completed is received from the host 2 before the restoration of the position information group 550 of NSID2 is completed, the storage device 1 reports that it is incomplete. Similarly, when the command for inquiring whether or not the activation of NSID0, which has not started the restoration of the position information group 550, is completed during the restoration of the position information group 550 of NSID2 is received, the storage device 1 similarly Report that it is incomplete. Then, after the restoration of the location information group 550 of NSID2 is completed, when a command inquiring whether the activation of NSID2 is completed is received, the storage device 1 reports that it is completed. In this example, when the storage device 1 completes the restoration of the position information group 550 of one namespace, the storage device 1 starts to restore the position information group 550 of the namespace of the next activation priority. When the restoration of the position information group 550 is completed for all the namespaces, the fact that the storage device 1 has been activated is reported to the host 2.

Next, with reference to FIG. 17, FW control at the time of starting the storage device 1 will be described.

With the power supply to the storage device 1 (S300), the CPU 40 first reads the NS management table 540 from the NAND flash memory 20 (S301). Next, the CPU 40 initializes the variable i indicating the activation priority to 1 (S302), and logs from the log area of the namespace having the highest activation priority (that is, i=1) according to the activation priority of the NS management table 540. 530 is read (S303), and the position information group 550 is restored on the RAM 30 (S304). Until the restoration of the location information group 550 of the namespace is completed (S305: No), the CPU 40 updates the read position of the log 530 (S306) and repeats the reading of the log 530 and the restoration to the RAM 30 (S303 to S306). ).

When the restoration of the position information group 550 of the namespace is completed (S305: Yes), the CPU 40 instructs the host IF control unit 200 to report the completion of activation of the namespace to the host 2 (S307).

When the activation of all namespaces is not completed (S308: No), the CPU 40 increments the variable i indicating the activation priority (S309), and outputs the log 530 from the log area of the namespace whose activation priority is i. The position information group 550 is read out (S303) and restored on the RAM 30 (S304).

When the activation of all namespaces is completed (S308: Yes), the CPU 40 instructs the host IF control unit 200 to report the completion of activation of the storage device 1 to the host 2 (S310).

After that, the translation information 510 is cached from the NAND flash memory 20 to the RAM 30 (S311).

Next, with reference to FIG. 18a and FIG. 18b, an example of the operation of the command that specifies the startup priority will be described.

FIG. 18a is an example of an operation of a command for designating an activation priority for an existing namespace. As shown in FIG. 18a, for example, the host 2 designates the boot priority of the NSID 2 as the first priority. The storage device 1, which has received this command, reports the normal end to the host 2 after setting the activation priority of NSID 2 in the NS management table 540 to the first rank.

FIG. 18b is an example of the operation of the command that specifies the activation priority at the same time as the creation of the namespace. As shown in FIG. 18a, for example, the host 2 requests to generate the NSID 2 having the first boot priority. Upon receipt of this command, the storage device 1 allocates an area of NSID2 to the logical address space, sets the activation priority of NSID2 in the NS management table 540 to first, and then reports a normal end to the host 2. ..

It should be noted that in response to any of the above commands, if the operation requested by the host 2 cannot be performed, or if this operation fails, the storage device 1 reports an error to the host 2.

As described so far, the host 2 can designate the namespace to be preferentially used when the storage device 1 is activated by designating the activation priority for each namespace to the storage device 1.

Next, with reference to FIG. 19a and FIG. 19b, an example of the activation priority order for each namespace, the use for each namespace, and an activation completion report from the storage device 1 will be described.

The host 2 determines the usage for each namespace and how to set the activation priority. The host 2 stores, for example, the use of each namespace (a) a boot record, a namespace for storing an OS kernel image, (b) a namespace for storing a system file necessary for booting the OS, and (c) an application program. To be stored, (d) a namespace for storing user data, and (e) a namespace for storing backup data. Further, the host 2 determines that the activation priority of these namespaces is higher in the order of (a)(b)(c)(d)(e). Then, the host 2 issues a command for designating the boot priority order to the storage device 1, and designates the boot priority order for each namespace.

In the example shown in FIG. 19a, NSID0 has the startup priority set to the third rank and is used to store the application program. NSID1 has a startup priority of second and is used for storing system files. The NSID 2 has the boot priority set to the first rank and is used to store the boot record and the OS kernel image. The NSID 3 has a startup priority of 4th and is used to store user data. The NSID 4 has a startup priority of 5th and is used to store backup data.

By setting each namespace in this way, the host 2 can perform the OS boot process prior to other user data. That is, as shown in FIG. 19B, the host 2 starts the OS boot process by receiving the boot completion report of the NSID 2 having the first boot priority. The host 2 starts the use of the system file by receiving the activation completion report of NSID1 having the second highest activation priority. The host 2 starts using the application program by receiving the activation completion report of NSID0 having the third highest activation priority. As a result, the boot time of the entire system can be shortened as compared with a system in which the host 2 starts the OS boot process after waiting for the boot of the entire storage device 1 to be completed.

Further, it is possible to preferentially activate a namespace in which user data that is desired to be preferentially used when the system is activated is stored. In the example of FIG. 19b, the host 2 starts using the user data by receiving the activation completion report of NSID3 having the fourth activation priority. The host 2 starts to use the backup data by receiving the activation completion report of NSID 4 having the fifth activation priority.

Next, with reference to FIG. 20a and FIG. 20b, another example of activation priority order for each namespace, usage for each namespace, and activation completion report from the storage device 1 will be described.

In the example shown in FIG. 20a, NSID0 has the activation priority set to the third rank and is used to store the user data A. The NSID 1 has the activation priority set to the second rank and is used to store the user data B. The NSID 2 is set to have the first boot priority and is used to store the operating system (OS). The NSID 3 has the activation priority set to the fourth rank and is used to store the user data C.

User data A to user data C correspond to virtual machine images of a plurality of users, for example. The user data B corresponds to data used by all users, for example. The user data A corresponds to data used by a user who has a high-priority license contract, for example. The user data C corresponds to data used by a user who has a low-priority license contract, for example.

By setting each namespace in this way, the host 2 can preferentially activate the namespace in which user data desired to be preferentially used at system startup is stored. In the example of FIG. 20b, the host 2 starts to use the user data B by receiving the activation completion report of NSID1 whose activation priority is second. The host 2 starts to use the user data A by receiving the activation completion report of NSID0 whose activation priority is the third highest. The host 2 starts to use the user data C by receiving the activation completion report of the NSID 3 having the fourth activation priority.

Next, with reference to FIGS. 21a to 21c, an example of the operation in the case where the start priority orders designated by the commands overlap will be described.

As shown in FIG. 21B, for example, the activation priority order recorded in the NS management table 540 is as follows: NSID0 is third, NSID1 is second, NSID2 is first, NSID3 is fourth, and NSID4 is fourth. Fifth place.

In this state, it is assumed that the host 2 specifies the startup priority of NSID 4 to the third rank by the command (command of FIG. 18a) that specifies the startup priority for the existing namespace. However, at this time, the third rank of the activation priority is already assigned to NSID0. In such a case, the storage device 1 first sets the command-specified activation priority (3rd) to NSID4. In addition, among the existing namespaces, the startup priority of the namespace that is not the command-specified namespace and has a startup priority equal to or lower than the command-specified startup priority (3rd) is set one by one. It That is, the activation priority of NSID0 having the third activation priority is set to the fourth, and the activation priority of NSID3 having the fourth activation priority is set to the fifth. When the setting of the activation priority of each namespace is completed, the storage device 1 reports the normal end of the command to the host 2.

After the operation requested by this command is completed, the NS management table 540 becomes as shown in FIG. 21c, where NSID0 is fourth, NSID1 is second, NSID2 is first, NSID3 is fifth, and NSID4 is fifth. Is the third place.

Next, with reference to FIGS. 22a to 22c, another example of the operation in the case where the start priorities designated by the commands overlap will be described. Here, the operation when a command (command in FIG. 18b) that specifies the activation priority is received at the same time when the namespace is generated will be described. Note that FIG. 22b showing the state of the NS management table 540 before command reception is the same as FIG. 21b.

At the same time as the creation of the namespace, the host 2 requests to generate NSID5 having the third boot priority by the command specifying the boot priority, but the third boot priority is already assigned to NSID0. ing. In such a case, the storage device 1 first sets the command-specified activation priority (3rd) to NSID 5. In addition, among the existing namespaces, the startup priority of the namespace that is not the command-specified namespace and has a startup priority equal to or lower than the command-specified startup priority (3rd) is set one by one. It That is, the startup priority of NSID0 having the third startup priority is set to the fourth, the startup priority of NSID3 having the fourth startup priority is set to the fifth, and the startup priority is set to the fifth. The activation priority of NSID4, which was ranked fifth, is set to sixth. When the generation of the NSID 5 and the setting of the activation priority order of each namespace are completed, the storage device 1 reports the normal end of the command to the host 2.

After the operation requested by this command is completed, the NS management table 540 becomes as shown in FIG. 22c, where NSID0 is fourth, NSID1 is second, NSID2 is first, NSID3 is fifth, and NSID4 is fourth. Is the 6th place, and NSID 5 is the 3rd place.

21a to 21c and 22a to 22c, in the case where the storage device 1 cannot perform the operation requested by the host 2 or fails in this operation, in particular, the boot priority order. When the duplication of the above is not permitted, the storage device 1 may report an error to the host 2.

If the start priority order specified by the command is reflected in the NS management table 540 and as a result, the start priority order recorded in the NS management table 540 becomes discontinuous, the storage device 1 stores the start priority order. The activation priority order may be changed so that For example, when the start priority order designated by the command is reflected in the NS management table 540 and the start priority order recorded in the NS management table 540 is the first, second, and fourth positions, the NS The management table 540 may be rewritten to change the activation priority order to be first, second, and third.

Next, with reference to FIG. 23, the control of the FW when the command designating the activation priority order is received will be described.

When the command designating the activation priority is received (S400), the CPU 40 sets the activation priority designated by the command in the namespace (designated namespace) designated by the command (S401). Next, the CPU 40 checks whether or not the designated boot priority order overlaps with the existing boot priority order (S402).

If the startup priority specified by the command overlaps with any of the existing startup priorities (S402: Yes), the startup is in a namespace other than the specified namespace and that is lower than the startup priority specified by the command. For a namespace having a priority, the activation priority is set lower by 1 (S403).

Next, the CPU 40 checks whether or not the start priority order recorded in the NS management table 540 has discontinuity (S404). When there is discontinuity in the start priority order (S404: Yes), the CPU 40 may change the start priority order so as to eliminate the discontinuity (S405).

When the setting of the boot priority is completed for all namespaces by (S401), (S403), or (S405), the CPU 40 instructs the host IF control unit 200 to report the normal termination of the command to the host 2 (S406). ).

Next, with reference to FIGS. 24a to 24c, an example of the operation of the command for inquiring the activation priority will be described.

As shown in FIG. 24a, for example, the startup priority order recorded in the NS management table 540 is that NSID0 is third, NSID1 is second, NSID2 is first, NSID3 is fourth, and NSID4 is fourth. Fifth place.

In FIG. 24 b, the storage device 1 has received a command from the host 2 to inquire about the boot priority of NSID 3. The CPU 40 searches the NS management table 540, obtains the boot priority (4th) of the NSID 3, and then instructs the host IF control unit 200 to report the 4th boot priority of the NSID 3 to the host 2. To do.

In FIG. 24c, the storage device 1 has received from the host 2 a command for inquiring about the namespace ID having the fourth activation priority. The CPU 40 searches the NS management table 540, obtains the namespace ID (NSID3) having the fourth activation priority, and then instructs the host IF control unit 200 to give the host 2 the fourth activation priority. Report NSID3 as the namespace ID, which is the rank.

24b and FIG. 24c, if the namespace ID or the activation priority matching the inquiry from the host 2 is not recorded in the NS management table 540, the storage device 1 sends the host 2 And report the error.

According to the storage device of at least one embodiment described above, the host specifies the startup priority for each namespace, and the storage device performs the startup processing in order from the namespace with the highest startup priority. The startup time can be shortened.

Although some embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the spirit of the invention. These embodiments and modifications thereof are included in the scope and the gist of the invention, and are also included in the invention described in the claims and an equivalent range thereof.

1... Storage device, 2... Host, 10... Controller, 20... Nonvolatile storage medium, 30... RAM, 40... CPU, 100... Memory chip, 200... Host IF control unit, 210... Buffer control unit, 220... NAND control Part, 300... Physical page, 310... Physical block, 320... Cluster, 400... Logical page, 410... Logical block, 500... User data, 510a... Translation information, 510b... Translation information (cache), 520a... Translation information group, 520b... Translation information group (cache), 530... Log, 540... NS management table, 550... Position information group, 560... Position information

Claims (12)

A host device,
A storage device connected to the host device;
Equipped with
The storage device is
A non-volatile semiconductor memory for storing data transmitted from the host device;
A controller that controls communication with the host device and controls write operation and read operation of the data to the nonvolatile semiconductor memory;
Including,
The controller is
The space of the logical address associated with the data is managed by being divided into a plurality of partial spaces,
Starting information required for starting the storage device, which is necessary for conversion between the logical address and the physical address of the non-volatile semiconductor memory, can be identified by the starting priority of each partial space. Information management system managed by. The controller is
The information processing system according to claim 1, wherein the host device sets the boot priority in accordance with a designation of a first command that designates and transmits one of the subspaces. The information processing system according to claim 1, wherein the controller controls the startup information to be stored in the nonvolatile semiconductor memory. The non-volatile memory has a plurality of areas,
4. The one area of the plurality of areas stores a plurality of the startup information corresponding to one of the partial spaces, and the startup information corresponding to two or more different partial spaces is not stored. Information processing system described. The information processing system according to claim 4, wherein each of the plurality of areas is a unit for erasing data in the nonvolatile memory. The controller is
When starting the storage device,
The information processing system according to claim 1, wherein the boot information is read in the descending order of boot priority. The controller is
The information processing system according to claim 6, wherein the completion of the activation of the partial spaces is reported to the host device for each of the partial spaces. 7. The information processing system according to claim 6, wherein data associated with a logical address belonging to the partial space can be written and read based on a command from the host when the activation of the partial space is completed. The controller is
The host device sets the activation priority in accordance with a designation of a first command which is designated and transmitted by one of the subspaces,
The plurality of subspaces are
Having a first subspace and a second subspace,
The host device is
A boot priority order of the second subspace to which the logical address to which the operating system kernel is associated belongs is specified to be higher than the first subspace to which the logical address to which user data is associated belongs. Item 6. The information processing system according to item 6. The controller is
The host device sets the activation priority in accordance with a designation of a first command which is designated and transmitted by one of the subspaces,
The plurality of subspaces are
Has a third subspace and a fourth subspace,
The host device is
When the user who has the contract of the second priority has a higher priority than the user who has the contract of the first priority,
A user of a user who has a contract with the second priority rather than the third partial space to which the logical address associated with the user data of the user with the first priority contract belongs. The information processing system according to claim 6, wherein the start priority of the fourth subspace to which the logical address to which the data is associated belongs is designated to be high. The information processing system according to claim 1, wherein the host device and the storage device communicate with each other based on the NVMe standard, and the partial space is a namespace. A host device,
A storage device connected to the host device;
Equipped with
The storage device is
A non-volatile semiconductor memory for storing data transmitted from the host device;
A controller that controls communication with the host device and controls write operation and read operation of the data to the nonvolatile semiconductor memory;
Including,
The storage device is
For each of the plurality of subspaces, start-up is performed by the start-up information that is information necessary for conversion between the logical address and the physical address of the nonvolatile semiconductor memory,
The plurality of subspaces are divided from the space of the logical address associated with the data, and the activation priority order is defined in each of the plurality of subspaces,
The controller is
An information processing system that manages the boot information so that the storage device can be booted for each of the partial spaces according to the boot priority.

Images