WO2021120218A1

WO2021120218A1 - Method for improving utilization rate of ssd, storage device, and ssd controller

Info

Publication number: WO2021120218A1
Application number: PCT/CN2019/127180
Authority: WO
Inventors: 贾学超; 李敏秋; 褚艳旭
Original assignee: 华为技术有限公司
Priority date: 2019-12-20
Filing date: 2019-12-20
Publication date: 2021-06-24
Also published as: CN114787760A

Abstract

A device for improving the utilization rate of a solid-state drive (SSD), comprising: an SSD controller and a storage medium array. The SSD controller comprises a receiving unit and a processing unit. The storage medium array comprises at least one PU. The PU is divided into a main area (MA) and/or a lease area (LA). The receiving unit is used for receiving at least one service distributed by at least one host or virtual machine. The processing unit is used for determining the number of required least concurrent unit PUs for each service, and is further used for allocating the service to an MA or an LA of a PU having the least number of PUs.

Description

Method for improving SSD utilization rate, storage device and SSD controller

Technical field

This application relates to the storage field, and in particular to a method and device for improving a solid state drive (SSD).

Background technique

An SSD is a hard disk made of an array of solid-state electronic storage chips. Among them, the quality of service (QoS) of the SSD refers to the ability of the SSD device to provide a stable, consistent and predictable request response service to the host business. QoS is one of the key factors shaping the competitiveness of enterprise SSD products. In the main application scenarios of data centers, servers, or other SSDs, read latency is a key dimension to measure QoS. Among them, the utilization rate of the SSD in multiple dimensions, such as bandwidth, capacity, and lifespan. Among them, bandwidth can be the number of read and write input and output per second (input output per second, IOPS), read and write bandwidth (bandwidth, BW), etc.; and, the life span can be the total write byte (TWB), The number of erasing and writing (, PE), etc. However, in most usage scenarios, the utilization of each dimension of SSD is difficult to achieve both.

Since the SSD is designed and produced, the specifications of its dimensions have been determined, but the requirements for SSDs are also very different for different usage scenarios, such as various services in the data center or server (such as applications or users, etc.). In order to bridge the needs of different businesses and the differences in different hardware specifications, it is necessary to perform comprehensive management of hardware resources such as pooling and reallocation. After the hardware resource is pooled, it can be shared by multiple services under a certain allocation method, but it will inevitably introduce mutual interference between different services, for example, in a form of conflict or resource contention. The mutual interference between different services will reduce the QoS index.

In order to guarantee specific QoS indicators, the prior art generally suppresses mutual interference between services by means such as resource reservation and key resource isolation. However, this method makes it difficult for hardware resources to be fully utilized, which ultimately increases the total cost of ownership (TCO).

With the development of cloud applications, the current SSD demand for services is also polarized. For example, some applications that directly face users have higher and higher QoS requirements, such as information retrieval, social applications, etc.; and some applications that do not directly face users have higher and higher requirements for IOPS, such as big data. Application etc. In the non-volatile memory high-speed interface specification (non-volatile memory express, NVMe) 1.4 protocol, through non-volatile memory (non-volatile memory, NVM sets) and predictable latency mode (predictable latency mode, PLM) It provides isolation convenience for input and output (IO) in the dimensions of space and time, thereby improving the QoS of SSD. Among them, the concept of groups in NVM sets is not to group different memories, but to group different units at the same level in the memories. However, simple IO isolation will inevitably affect IOPS indicators while improving QoS indicators and reduce the bandwidth utilization of SSD products.

Summary of the invention

The embodiment of the present application provides a method and device for improving the utilization rate of a solid-state hard disk, so as to improve the utilization rate of a storage device under the premise of guaranteeing the QoS of the SSD service.

In the first aspect, the present application provides a storage device. The device includes: a storage medium array and a solid state drive SSD controller coupled to the storage medium array; the storage medium array includes a plurality of concurrent units PU, and the PU includes a master area (master area, At least one of MA) or lease area (LA), where MA is used to be allocated and used by the previous service, and LA is used to be allocated and used by at least one service; SSD controller is used to receive the characteristics of at least one service Information, where the characteristic information includes a service quality QoS indicator; the SSD controller is also used to allocate the MA of the PU to the service with the largest QoS indicator according to the QoS indicator, and allocate the LA of the PU to the rest of the service.

In a possible implementation manner, the characteristic information also includes typical read/write input/output number IOPS requirements per second, and service storage capacity indicators or typical capacity requirements. The SSD controller is also used to determine service requirements based on typical IOPS requirements. The first number of PUs; the second number of PUs required by the business is determined according to the service storage capacity index or typical capacity requirements; the MA of the PU is allocated to the service with the largest QoS index, and the LA of the PU is allocated to the rest of the service. : Allocate the MA of the first PU among the multiple PUs to the service with the largest QoS index, and allocate the LA of the first PU to the remaining services, where the number of PUs included in the first PU is the first number and The maximum value of the second number.

In a possible implementation manner, the characteristic information further includes the write volume requirements of the business, the write amplification factor of the business, and the programming and erasing PE life of the solid-state hard disk. The SSD controller is also used to: according to typical capacity requirements or business storage capacity To determine the OP requirements of oversupply redundant space, including OP requirements and typical capacity requirements or service storage capacity indicators, determine the number of PUs required by the service, as well as the service write volume requirements, the service write amplification factor, and the PE lifetime of solid-state drives. Two quantity.

In a possible implementation manner, in the same PU, the SSD controller controls the priority of the input and output IO operations that access the MA is higher than the priority of the IO operations that access the LA; where the IO operations are business IO operations .

In a possible implementation manner, the priority scheduling mode is strict priority scheduling, probabilistic priority scheduling, deprivation priority scheduling, or non-deprivation priority scheduling.

In a possible implementation manner, strict priority scheduling is that IO operations with high priority are scheduled before IO operations with low priority; probabilistic priority scheduling is that the probability of scheduling IO operations with high priority is higher than that of scheduling priority. Probability of low-level IO operations; deprivation-priority scheduling is high-priority IO operations under preset conditions, allowing interruption of low-priority IO operations being executed; non-deprivation-priority scheduling is high-priority IO Under any conditions, the operation is not allowed to interrupt the low priority IO operation being executed.

In a possible implementation manner, the SSD controller is also used to: sort at least one service from high to low according to the QoS index; according to the sort, sequentially assign LAs with different priorities to the remaining services, where the priority of the LA is equal to The QoS indicators of the allocated services are positively correlated.

In a possible implementation manner, the SSD controller is also used to adjust the ratio of MA and LA in the PU if the service has a margin in the allocated PU to achieve the service index.

In a possible implementation manner, the SSD controller is also used to: if there is unused storage space in the PU and the service allocated to the PU has margin to achieve the service storage capacity index and the service quality QoS index, adjust the PU internal The ratio of MA to LA.

In a possible implementation manner, achieving the business storage capacity index with margin includes: the storage capacity allocated to the business is greater than the storage capacity required by the business; or the storage capacity allocated to the business is equal to the storage capacity required by the business, but The storage capacity allocated to the business at the current moment is not fully used by the business.

In a possible implementation manner, achieving the service QoS index with margin includes: the actual QoS of the service is greater than or equal to the QoS index; or the service ignores the requirement of the QoS index within a certain time.

In a possible implementation manner, the SSD controller is also used to: determine the first ratio value of the PU according to the service storage capacity index, where the first ratio value is the ratio value of the MA and LA in the PU; through a pilot or model Analyze and estimate to determine the first QoS indicator of the service; if the first QoS indicator is less than the service QoS indicator, adjust the first ratio value, and determine the second QoS indicator of the adjusted service through pilot or model analysis and estimation; if the second If the QoS index is less than the service QoS index, continue to adjust the first ratio value of the PU, and determine whether the new first QoS index is greater than or equal to the service QoS index; if the new first QoS index is still less than the service QoS index, repeat In the above steps, continue to adjust the first ratio value until the QoS index of the service is greater than or equal to the service QoS index.

In a possible implementation manner, when the PU is multiple PUs, after determining that the first QoS indicator is less than the service QoS indicator, the SSD controller is further configured to: keep the first ratio constant and adjust the multiples of the multiple PUs. The third QoS index of the adjusted service is determined through pilot or model analysis and estimation, where the second ratio value is the ratio of MA and LA of one PU among multiple PUs; if the third If the QoS index is less than the service QoS index, adjust the first ratio value, and adjust the second ratio value of one or more PUs in multiple PUs at the same time, and determine the second QoS index of the adjusted service through pilot or model analysis and estimation ; If the second QoS index is less than the service QoS index, keep the adjusted first ratio value unchanged, continue to adjust the second ratio value of the PU, and determine whether the new third QoS index is greater than or equal to the service QoS index; if new When the third QoS index is still less than the service QoS index, repeat the above steps and continue to adjust the first ratio value until the service QoS index is greater than or equal to the service QoS index.

In a possible implementation manner, the SSD controller is further configured to: if the PU allocated by the service has an LA, and part of the LA is not used, then the unused part of the LA is allocated to other services.

In a possible implementation manner, the SSD controller is also used to: if other services ignore the requirements of QoS indicators within a certain period of time, determine to allocate the unused part of the LA to other services; or through pilot or model analysis Estimate and determine whether the QoS indicators of other services are greater than or equal to the business QoS indicators after receiving the allocated unused part of the LA; if the QoS indicators of other services are greater than or equal to the business QoS indicators, then determine the unused part of the LA Assign to other businesses.

In a possible implementation manner, the storage medium array is also used to: in the same PU, if a specific condition is met, the service located in the MA actively triggers the physical location rotation.

In a possible implementation manner, the storage medium array is also used to: select the physical block to be rotated from the MA; select an idle block in the LA in the same PU, or select the MA in the same PU Write the valid data in the block to be rotated into the free block; if you select an free block in the LA in the same PU, modify the attribute of the free block to MA, and change the block to be rotated Mark the data as invalid, and modify the attribute to LA; if you select an idle block in the MA in the same PU, mark the data in the block to be rotated as invalid.

In a possible implementation manner, the specific conditions include: the service triggers the rotation daemon; or the service ignores the requirements of the service QoS index within a specific time; or the PU assigned by the service is operating in the background and does not affect the service QoS index Or at least one host or virtual machine actively initiates it, where the host or virtual machine is also used to deliver at least one service; or the PU fails.

In a possible implementation manner, the storage medium array is also used to: if the area divided by the current PU cannot match the priority of the service, divide the MA into a first priority MA and a second priority LA; and/or LA is divided into a third priority LA and a fourth priority LA; if the area of the divided PU still cannot match the priority of the service, the area within the PU will continue to be binary division until it matches the priority of the service; , The priority of the area after each binary division does not exceed the priority before the division of the area.

In a possible implementation manner, the SSD controller and the storage medium array are packaged on an integrated chip; or the SSD controller includes a first SSD controller and a second SSD controller; the first SSD controller is located in at least one host The second SSD controller and storage medium array are packaged on an integrated chip.

In a possible implementation manner, the storage medium array includes at least one storage medium, and the storage medium is any of flash memory, dynamic random access memory DRAM, static random access memory SRAM, phase change memory PCM, and storage class memory SCM. One.

In the second aspect, this application provides a solid state drive SSD controller, which is coupled with a storage medium array; wherein the storage medium array includes a plurality of concurrent units PU, and the PU includes at least one of the main area MA or the rental area LA. One, where MA is used to be allocated and used by one service, and LA is used to be allocated and used by at least one service; SSD controller is used to receive characteristic information of at least one service, where the characteristic information includes service quality QoS indicators; SSD control The device is also used to allocate the PU of the PU to the service with the largest QoS index according to the QoS index, and to allocate the PU's LA to the rest of the service.

In a possible implementation manner, the SSD controller is also used to: determine the first ratio value of the PU according to the service storage capacity index, where the first ratio value is the ratio value of the MA and LA in the PU; through a pilot or model Analyze and estimate to determine the first QoS indicator of the service; if the first QoS indicator is less than the service QoS indicator, adjust the first ratio value, and determine the second QoS indicator of the adjusted service through pilot or model analysis and estimation; if the second If the QoS index is less than the service QoS index, continue to adjust the first ratio of multiple PUs in the PU, and determine whether the new first QoS index is greater than or equal to the service QoS index; if the new first QoS index is still less than the service QoS index At this time, repeat the above steps and continue to adjust the first ratio value until the QoS index of the service is greater than or equal to the service QoS index.

In a possible implementation manner, when the PU is multiple PUs, after determining that the first QoS indicator is less than the service QoS indicator, the SSD controller is further configured to: keep the first ratio constant and adjust the multiples of the multiple PUs. The third QoS index of the adjusted service is determined through pilot or model analysis and estimation, where the second ratio value is the ratio of MA and LA of one PU among multiple PUs; if the third If the QoS index is less than the service QoS index, adjust the first ratio value, and adjust the second ratio value of one or more PUs in multiple PUs at the same time, and determine the second QoS index of the adjusted service through pilot or model analysis and estimation ; If the second QoS index is less than the service QoS index, keep the adjusted first ratio value unchanged, continue to adjust the second ratio value of the PU, and determine whether the new second QoS index is greater than or equal to the service QoS index; if new When the second QoS index of is still less than the service QoS index, repeat the above steps and continue to adjust the first ratio value until the service QoS index is greater than or equal to the service QoS index.

In a possible implementation manner, the SSD controller is also used to: in the same PU, the SSD controller controls the storage medium array, and if a specific condition is met, the service located in the MA actively triggers the physical location rotation.

In a possible implementation manner, the SSD controller is also used to: control the storage medium array, select the physical block to be rotated from the MA; select an idle block in the LA in the same PU, or select the same An idle block in the MA in the PU; write the valid data in the block to be rotated into the idle block; if an idle block in the LA in the same PU is selected, modify the attribute of the idle block to MA, and Mark the data in the block to be rotated as invalid and modify the attribute to LA; if an idle block in the MA in the same PU is selected, the data in the block to be rotated is marked as invalid.

In a possible implementation manner, the SSD controller is also used to: if the area divided by the current PU cannot match the priority of the service, the SSD controller controls the storage medium array and divides the MA into the first priority MA and the second priority Level LA, and/or divide the LA into a third priority LA and a fourth priority LA; if the area of the divided PU still cannot match the priority of the service, continue to divide the area within the PU into binary division until Match the priority of the service; where the priority of the area after each binary division does not exceed the priority before the division of the area.

In a third aspect, the present application provides a method for improving the utilization rate of a solid state drive. The method includes: receiving characteristic information of at least one service, wherein the characteristic information includes a service quality QoS index; and dividing the main area of the concurrent unit PU according to the QoS index The MA is allocated to the service use with the largest QoS index, and the leased area LA of the PU is allocated to the rest of the service use, where the PU includes at least one of the MA or the LA.

In a possible implementation manner, the characteristic information also includes typical read/write input/output number IOPS requirements per second, and business storage capacity indicators or typical capacity requirements; according to typical IOPS requirements, determine the first number of PUs required by the business; Service storage capacity index or typical capacity requirement, to determine the second number of PUs required by the service; assigning the PU’s MA to the service with the largest QoS index, and assigning the PU’s LA to the rest of the service use includes: The MA of the first PU is allocated to the use of the service with the largest QoS index, and the LA of the first PU is allocated to the use of the remaining services, where the number of PUs included in the first PU is the maximum of the first number and the second number .

In a possible implementation manner, the characteristic information also includes the write volume requirements of the business, the write amplification factor of the business, and the programming and erasing PE life of the solid-state hard disk; the second PU number is determined according to the business storage capacity index or typical capacity requirements. Including: According to typical capacity requirements or business storage capacity indicators, business write volume requirements, business write amplification factor, solid-state drive PE life, determine the excess redundant space OP requirements; combine OP requirements and typical capacity requirements or business storage The capacity index determines the second number of PUs required by the business.

In a possible implementation manner, the method further includes: located in the same PU, the SSD controller controls the priority of the input and output IO operations that access the MA is higher than the priority of the IO operations that access the LA; wherein, the IO operations are services IO operations.

In a possible implementation manner, allocating the PU's LA to other services includes: sorting at least one service from high to low according to QoS indicators; according to the sorting, sequentially assigning LAs of different priorities to the remaining services, where LA The priority of is positively correlated with the QoS index of the assigned service.

In a possible implementation manner, the method further includes: adjusting the ratio of MA and LA in the PU if the service has a margin in the allocated PU to achieve the service index.

In a possible implementation, if the business has a margin in the allocated PU to achieve the business index, adjusting the ratio of MA and LA in the PU includes: if the PU has unused storage space and is allocated to the business of the PU If there is margin to achieve the business storage capacity index and the business service quality QoS index, the ratio of MA and LA in the PU will be adjusted.

In a possible implementation manner, adjusting the ratio of MA and LA in the PU includes: determining a first ratio value of the PU according to a service storage capacity index, where the first ratio value is the ratio value of MA and LA in the PU; Pilot or model analysis and estimation to determine the first QoS index of the service; if the first QoS index is less than the service QoS index, adjust the first ratio value, and determine the adjusted second QoS index of the service through pilot or model analysis and estimation; If the second QoS index is less than the service QoS index, continue to adjust the first ratio value of the PU, and determine whether the new first QoS index is greater than or equal to the service QoS index; if the new first QoS index is still less than the service QoS index , Repeat the above steps and continue to adjust the first ratio value until the QoS index of the service is greater than or equal to the service QoS index.

In a possible implementation manner, when the PUs are multiple PUs, after determining that the first QoS indicator is less than the service QoS indicator, the method further includes: keeping the first ratio constant and adjusting multiple second ones in the multiple PUs. The third QoS index of the adjusted service is determined through pilot or model analysis and estimation, where the second ratio value is the ratio of the MA and LA of one PU among the multiple PUs; if the third QoS index is less than For service QoS indicators, adjust the first ratio value, and simultaneously adjust the second ratio value of one or more PUs in multiple PUs, and determine the second QoS indicator of the adjusted service through pilot or model analysis and estimation; 2. If the QoS index is less than the service QoS index, keep the adjusted first ratio value unchanged, continue to adjust the second ratio value of the PU, and determine whether the new second QoS index is greater than or equal to the service QoS index; if the new second QoS index is greater than or equal to the service QoS index; When the QoS index is still less than the service QoS index, repeat the above steps and continue to adjust the first ratio value until the service QoS index is greater than or equal to the service QoS index.

In a possible implementation manner, the method further includes: if the PU allocated by the service has an LA and part of the LA is not used, allocating the part of the LA that is not used to other services.

In a possible implementation manner, allocating part of the unused LA to other services includes: if other services ignore the requirements of QoS indicators within a certain period of time, then determining to allocate the part of the unused LA to other services; Or through pilot or model analysis and estimation, it is determined whether the QoS indicators of other services are greater than or equal to the business QoS indicators after receiving the allocated unused part of the LA; if the QoS indicators of other services are greater than or equal to the business QoS indicators, it is determined The unused part of LA is allocated to other services.

In a possible implementation manner, the method further includes: in the same PU, if a specific condition is met, the service located in the MA actively triggers the physical location rotation.

In a possible implementation manner, the physical position rotation includes: selecting the physical block to be rotated from the MA; selecting an idle block in the LA in the same PU, or selecting one of the MA in the same PU Free block; write the valid data in the block to be rotated into the free block; if you select an free block in the LA in the same PU, modify the attribute of the free block to MA; mark the data in the block to be rotated It is invalid, and the attribute is changed to LA; if a free block in the MA in the same PU is selected, the data in the block to be rotated will be marked as invalid.

In a possible implementation manner, the method further includes: if the area divided by the current PU cannot match the priority of the service, dividing the MA into a first priority MA and a second priority LA; and/or dividing the LA into The third priority LA and the fourth priority LA; if the area of the divided PU still cannot match the priority of the service, continue to divide the area within the PU until the priority of the service is matched; among them, every time The priority of the area after the binary division does not exceed the priority before the division of the area.

In a fourth aspect, the present application provides a computer-readable storage medium, including instructions, which when run on a computer, cause the computer to execute the method according to any one of the third aspects.

In the fifth aspect, the present application provides a computer program product containing instructions, which when run on a computer, causes the computer to execute the method as in any one of the third aspect.

The method and device for improving the utilization rate of solid-state hard disks provided in this application are performed by inputting different services or dimensions of service QoS, IOPS, capacity, etc., and assigning the number of PUs as needed. At the same time, follow the following principles: first assign high QoS demand services, then assign low QoS demand services; high priority logical partitions are assigned to high QoS demand services first, and low priority logical partitions are first assigned to low QoS demand services; the higher the priority Logical partition, the less the number of businesses involved in sharing. While allocating the number of PUs on demand, it also allocates services to different logical partitions, which solves the problem of mutual interference when multi-tenant/multi-service co-location, and ensures that the utilization of storage media is maximized under the premise of meeting various business needs化. At the same time, by adjusting the proportions of different logical partitions in the concurrent unit, the proportion of PUs occupied by services can be reduced, while free resources are released to further improve the utilization of storage media. And through physical rotation, the wear leveling of the physical blocks in the PU is realized. This application also divides the concurrent units into different logical partitions and assigns different priority dimensions to refine management granularity, increase management flexibility, and improve storage media utilization while ensuring service QoS indicators.

Description of the drawings

FIG. 1 is a schematic diagram of an application scenario provided by an embodiment of the application;

2 is a schematic structural diagram of a flash storage medium structure provided by an embodiment of the application;

FIG. 3a is a schematic diagram of a non-NVM Sets isolation data arrangement provided by an embodiment of the application;

FIG. 3b is a schematic diagram of an isolated data arrangement with NVM Sets according to an embodiment of the application;

FIG. 4 is a schematic diagram of an arrangement of concurrent units provided by an embodiment of the application;

FIG. 5 is a schematic diagram of a concurrent unit area division provided by an embodiment of the application;

FIG. 6 is a structural diagram of a storage device provided by an embodiment of this application;

FIG. 7a is a schematic diagram of service distribution of a concurrent unit according to an embodiment of the application;

FIG. 7b is a schematic diagram of another concurrent unit service distribution provided by an embodiment of this application;

FIG. 8 is a schematic diagram of a concurrent unit area adjustment provided by an embodiment of the application;

FIG. 9 is a schematic diagram of a physical location rotation of logical partitions in a concurrent unit according to an embodiment of the application;

10 is a schematic diagram of another physical location rotation of logical partitions in a concurrent unit provided by an embodiment of the application;

FIG. 11 is a schematic diagram of a dichotomy of a logical partition of a concurrent unit according to an embodiment of the application;

FIG. 12 is a schematic diagram of a concurrent unit allocation service provided by an embodiment of this application;

FIG. 13 is a schematic diagram of another application scenario provided by an embodiment of the application;

FIG. 14 is a schematic diagram of another application scenario provided by an embodiment of this application;

FIG. 15 is a schematic diagram of yet another application scenario provided by an embodiment of the application;

FIG. 16 is a flowchart of a method for improving the utilization rate of a solid state drive provided by an embodiment of the application;

FIG. 17 is a flowchart of a method for adjusting the proportion of logical partitions of a concurrent unit according to an embodiment of the application;

FIG. 18 is a flowchart of a method for rotating the physical position of a concurrent unit according to an embodiment of the application;

FIG. 19 is a flowchart of a method for refining logical partitions of a concurrent unit according to an embodiment of the application.

Detailed ways

The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

This application is mainly applied to the host accessing data to the storage medium array through the SSD controller. FIG. 1 is a schematic diagram of an application scenario provided by an embodiment of the application. As shown in Figure 1, this scenario may include one or more hosts, one or more SSD controllers, and one or more storage medium arrays. Among them, one or more hosts communicate with various interfaces such as NVMe, serial attached small computer system interface (SAS), and high-speed serial computer expansion bus interface (peripheral component interconnect express, PCIe). The front end of the SSD controller is connected. And, the back end of each SSD controller is connected to the corresponding storage medium array through a non-flash interface (NFI), so that the SSD controller can read, write, and erase the data in the storage medium array. operating.

Those in the art should note that the storage medium in this application can be flash memory (flash), dynamic random access memory (dynamic random access memory, DRAM), static random access memory (static random-access memory, SRAM), or relative Any storage medium with the concept of a concurrent unit, such as phase change memory (PCM) and storage-class memory (SCM). This application is not limited again. For ease of description, the embodiment of the present application will take the storage medium as flash as an example for description.

FIG. 2 is a schematic structural diagram of a flash storage medium provided by an embodiment of the application. FIG. 2 shows the storage medium array in FIG. 1. If the storage medium is selected as flash, a schematic diagram of the structure of the flash. Among them, the flash array includes multiple flashes, each flash in the flash array can include multiple channels (channels), each channel can be composed of multiple modules (die), and each die can be composed of multiple planes. (plane), each plane can be composed of multiple blocks. In the scenarios shown in Figure 1 and Figure 2, SSDs will face the need to mix the deployment of "high QoS, low IOPS requirements" and "high IOPS, low QoS requirements" services. At the same time, there may be situations where it is necessary to temporarily increase the IOPS of NVM Sets.

In some solutions, QoS is improved by using spatial isolation (or called concurrent isolation). For example, in the related technology of NVM Sets, the SSD disks are separated into multiple regions according to the concurrency of the media to solve the problem of IO conflicts caused by multi-tenant access, which leads to the problem of deterioration of delay. This problem can be regarded as the main reason for the deterioration of QoS. Among them, the tenant refers to the customer who uses the system or computer computing resources, including all data that can be identified as a designated user in the system, such as account and statistical information, various data created by the user in the system, and the user's own customization Application environment, etc., belong to the scope of tenants. In this solution, each isolated NVM Sets is equivalent to a sub-disk of the SSD disk, which is exclusively shared by a single tenant or shared by a small number of tenants, thereby reducing IO conflicts and improving QoS.

FIG. 3a is a schematic diagram of a non-NVM Sets isolation data arrangement provided by an embodiment of the application. Each block shown in FIG. 3a can be understood as a die, and FIG. 3a is only an example showing a 4*4 flash array, in which 4 die in each column can be regarded as a channel. It can be seen that when NVM Sets are not used to isolate data, each die can be called by different tenant data. Among them, different tenants are distinguished by different patterns. For example, in Figure 3a, 8 different patterns are overlapped on each die, which means that a die in the area can be called by 8 different tenants at the same time. It can be seen that when different tenants need to use the same die at the same time, they will interfere with each other, thereby affecting QoS. FIG. 3b is a schematic diagram of an isolated data arrangement with NVM Sets according to an embodiment of the application. In addition, each block shown in FIG. 3b can be understood as a die, and FIG. 3b is only an example showing a 4*4 flash array, in which 4 die in each column can be regarded as a channel. It can be seen that when using NVM Sets to isolate data, each die is only called individually by a certain tenant, that is, it is exclusively shared by a specific tenant. Among them, different tenants are still distinguished by different patterns. However, it can be seen that the data of a specific tenant in FIG. 3b is hit and stored on a specific part of the die. For example, as shown in FIG. 3b, every four adjacent die are exclusively shared by a specific tenant. It can be seen that the 4*4 flash array shown in FIG. 3b is divided into 8 areas, and the die in each area is used for exclusive use by specific tenants. After the data access between tenants is completely isolated in space, mutual interference is minimized.

However, compared with all tenants sharing all concurrent media channels, in the above scheme, each tenant is isolated from each other, resulting in a reduction in the maximum number of concurrent media channels for each tenant, and thus the maximum IOPS that each tenant can achieve is also greatly reduced. At the same time, since the services of each tenant are independent of each other, the physical bandwidth of the SSD disk cannot be fully utilized. And when dealing with different customer services such as "high QoS, low IOPS demand" and "high IOPS, low QoS demand", two or more systems need to be designed accordingly to meet the business needs of different customers.

This application divides the concurrent units in the storage medium array into a main area MA and a rental area LA. According to the business needs of the business, the services with high QoS requirements are allocated to the MA of the storage media array; for the services with low QoS requirements, they can be allocated to the LA, so that multiple tenants or multiple services can be co-location (co-location) Mutual interference problems arising from time to time. At the same time, because the LA of the storage medium can be shared by multiple services with low QoS, the utilization rate of the storage device is improved while ensuring the QoS of the service. Those skilled in the art should note that the business involved in this application may also be a business set that includes multiple businesses.

The technical solutions in the embodiments of the present application will be described in detail below in conjunction with the drawings in the embodiments of the present application.

FIG. 4 is a schematic diagram of an arrangement of concurrent units provided by an embodiment of the application.

As shown in Figure 4, it can be seen that the storage medium located in the storage medium array contains multiple parallel units (PU). Among them, each PU displayed in each column can be regarded as being in the same channel, as shown by the channel arrow, each PU in the horizontal direction represents a PU in a different channel. The PU in each column represents each die in the channel. Those skilled in the art should note that in this application, the smallest unit that can concurrently and independently execute different operation types or different operation parameter commands is used as the PU. For example, taking mainstream flash storage media as an example, the smallest level that can perform "read/write/erase" operations independently of each other is die, then die is the PU in this application. It is understandable that for more advanced storage devices, if it can support independent and concurrent "read/write/erase" operations between plane levels, then for this type of storage media, plane can be involved in this application的PU.

FIG. 5 is a schematic diagram of a concurrent unit area division provided by an embodiment of the application.

As shown in Fig. 5, it can be seen that this application further divides each PU into a detailed division, for example, into two logical areas, namely the main area MA and the rental area LA. Among them, the MA in this application is only allowed to be allocated to a specific service or service set. In other words, the storage medium MA can only be used exclusively by a single service or a collection of services. For LA, it can be allowed to be allocated to one or more services or service sets. In other words, the storage medium LA can only be shared by one or more services or service sets.

FIG. 6 is a structural diagram of a storage device provided by an embodiment of the application.

A storage device is shown in FIG. 6. Wherein, the device may include: a storage medium array and a solid state hard disk SSD controller coupled to the storage medium array. Among them, the SSD controller may include a receiving unit and a processing unit.

In an example, the device may also include a host. Used to deliver at least one service or service set to the SSD controller. In another example, the host may be one or more hosts. In another example, one or more virtual machines (VM) or virtual functions (VF) may also run on the host. Those skilled in the art should note that the SSD controller can receive services or service sets delivered from one or more hosts and one or more virtual machines or virtual functions, which is not limited in this application.

In one embodiment, one or more hosts or virtual machines are connected to the SSD controller through various interfaces such as NVMe, SAS, PCIe, and so on. The receiving unit in the SSD controller receives the characteristic information of at least one service or service set issued by at least one host or virtual machine. In an example, the characteristic information includes service quality QoS indicators. The processing unit in the SSD controller can also be used to sort at least one service or service set according to the size of the QoS index of each service or service set. Then the PU's MA is allocated to the service or service set with the largest QoS index for use, and the PU's LA is allocated to the remaining services or service sets for use according to the ranking, so as to perform corresponding operations.

In another embodiment, the characteristic information further includes a typical read/write input/output number IOPS requirement per second, and a service storage capacity index or typical capacity requirement. Of course, it can be understood that the characteristic information may also include any other information that may be used to describe the characteristics of a service or a service set. The SSD controller determines the first number of PUs required by the service or service set according to typical IOPS requirements, which can also be referred to as the first number of PUs. After that, according to the service storage capacity index or typical capacity requirements, the second number of PUs required by the service or service set is determined, which can also be referred to as the second number of PUs. Then, the largest number of PUs in the first number of PUs and the second number of PUs is used as the first PU, that is, the first PU is the minimum number of PUs required. The SSD controller then allocates the MA of the least number of PUs to the service or service set with the largest QoS index, and allocates the LA of the least number of PUs to the rest of the service or service set according to the ordering. For the convenience of description, this application divides the minimum number of PUs in the storage medium array into a specific access area. Those skilled in the art should understand that the specific access area is divided only for the convenience of description, and there is no operation of dividing the specific access area in the actual service distribution process. For the convenience of description, the following uses "specific access area" to mean "the minimum number of PUs in the storage medium array". For example, as shown in FIG. 6, the processing unit divides the first 3 PUs in the first row and the first 3 PUs in the second row in the storage medium 1 into a specific access area. Among them, the specific access area is framed with bold lines. And the service or service set received by the receiving unit is allocated to the MA or LA in the area through the NFI.

Those skilled in the art should note that the logical partition in each PU may only include MA, or only LA, and may also include MA and LA. Those skilled in the art should also note that “assigning a certain service to a PU” in this application has the same meaning as “allocating the PU to a certain service for use”.

In one example, when there is only one service or service set, the service or service set is allocated to the MA of the PU in the specific access area. For example, as shown in FIG. 7a, the service or service set is allocated to the MA in a specific access area, that is, the gray area in FIG. 7a. In another example, when there are at least two services or service sets, sorting is performed according to the size of the QoS indicators of the at least two services or service sets. The service or service set with the largest QoS index is allocated to the MA of the PU in the specific access area. For example, as shown in FIG. 7a, the service or service set is allocated to the MA in a specific access area, that is, the gray area in FIG. 7a. And according to the above sorting sequence, the remaining services or service sets are sequentially allocated to the LA of the PU in the specific access area. For example, as shown in FIG. 7b, the service or service set is allocated to the LA in the specific access area, that is, the gray area in FIG. 7b. Among them, in another example, the remaining services or service sets can be assigned to LAs with different priorities in sequence. The priority of LA can be positively correlated with the QoS index of the assigned service or service set. In another example, the services or service sets with high QoS requirements in the above ranking may be allocated first, and then the services or service sets with low QoS requirements may be allocated.

In another example, each PU may include one or more LAs, and different LAs may have different priorities. Those skilled in the art should note that each PU can only have one MA, and the priority of the MA is the highest. For each PU, a logical partition with a higher priority has a smaller number of services or service sets that it participates in sharing. It is understandable that the MA with the highest priority is only assigned to one service or service set. For LAs with different priorities, the higher the priority LA, the fewer services or service sets can be shared.

In another embodiment, in the embodiment of the present application, when there are multiple services or service sets, the MA and LA in the same PU are respectively called by different services or service sets. At this time, in the same PU, all IO priorities for accessing MA are higher than those for accessing LA. Where IO is the IO operation of the corresponding business or business set.

Compared with other solutions, for example, by calculating the reserved time label corresponding to each tenant's IO request, combined with the information of the tenant game set in advance, and responding according to the priority from high to low, the reserved time label is not greater than The IO request at the current time. At the same time, other solutions are only applicable to closed systems and are vulnerable to some denial of service (DoS) attacks. And directly give users priority, for low-priority tenants may face process starvation (starvation) problem. However, in this application, there is no need to compare the time of each IO, and the implementation complexity is low. At the same time, it is suitable for open systems and is a storage medium with different priorities assigned according to the business or business set. Therefore, for high-priority services of low-priority tenants, the starvation problem is avoided.

In an example, the priority scheduling method of IO operation can be any priority scheduling method or multiple priority scheduling methods such as strict priority scheduling, probabilistic priority scheduling, deprivation priority scheduling, non-deprivation priority scheduling, etc. The combination form.

In one example, strict priority scheduling may be that IO operations with high priority are scheduled before IO operations with low priority. Probabilistic priority scheduling may be that the probability of scheduling an IO operation with a high priority is higher than the probability of scheduling an IO operation with a low priority. The deprivation priority scheduling can be that IO operations with high priority are allowed to interrupt the IO operations with low priority being executed under preset conditions. Non-deprivation priority scheduling can be that high-priority IO operations are not allowed to interrupt the low-priority IO operations being performed under any conditions. Those skilled in the art should note that the priority scheduling mode may also be any other priority scheduling mode, or a combination of multiple priority scheduling modes, which is not limited in this application.

In an embodiment, determining the minimum number of PUs required for the processing unit in the SSD controller can be specifically determined in the following manner. The processing unit in the SSD controller first determines the typical IOPS requirements of the service or service set issued by one or more hosts or virtual machines. According to the typical IOPS requirements of the service or service set, the first PU number is determined. It can be understood that the first number of PUs is the number of least used PUs determined according to the IOPS required by the service or service set. The processing unit in the SSD controller also needs to determine the service storage capacity index or typical capacity requirements of the service or service set. And according to the service storage capacity index or typical capacity requirement of the service or service set, the second PU number is determined. It can be understood that the second number of PUs is the number of least used PUs determined according to the service storage capacity index or typical capacity requirements of the service or service set requirements. Finally, the processing unit will select the largest number of the first PU number and the second PU number as the required minimum number of PUs.

Wherein, the service storage capacity index may be used to indicate how much capacity (capacity) the service or service set needs to use. As for the typical capacity requirements, it is the capacity requirements that SSDs actually need when performing services or business IO operations. Because when performing an IO operation, the storage medium needs to write the information on the medium to other spare mediums first when erasing the storage medium. Then the entire medium is erased, that is, the part that needs to be erased and rewritten at this time is much larger than the new data actually needs. At this time, this part of the capacity requirement needs to be taken into consideration, that is, the typical capacity requirement.

In another embodiment, the characteristic information further includes information such as the write volume requirements of the service or the service set, the write amplification factor of the service or the service set, and the program erase (PE) life of the solid state disk. When the processing unit determines the second number of PUs, it can be based on typical capacity requirements or service storage capacity indicators, terabyte write (TBW) requirements of the service or service set, write amplification factor of the service or service set, SSD To determine the over-provision (OP) requirements for the longevity of the PE. The processing unit then combines OP requirements and typical capacity requirements or business storage capacity indicators to determine the second PU number.

In this application, by entering dimensions such as different services or service QoS, IOPS, capacity, etc., the number of PUs is allocated on demand. At the same time, follow the following principles: first assign high QoS demand services, then assign low QoS demand services; high priority logical partitions are assigned to high QoS demand services first, and low priority logical partitions are first assigned to low QoS demand services; the higher the priority Logical partition, the less the number of businesses involved in sharing. While allocating the number of PUs on demand, it also allocates services to different logical partitions, which solves the problem of mutual interference when multi-tenant/multi-service co-location, and ensures that the utilization of storage media is maximized under the premise of meeting various business needs化.

FIG. 8 is a schematic diagram of a concurrent unit area adjustment provided by an embodiment of the application.

As shown in FIG. 8, in one embodiment, if the service or service set has a margin to achieve the business index in the assigned specific access area, the ratio of MA and LA in the PU of the specific access area is adjusted. As shown in the left half of Figure 8, in a specific access area, if multiple PUs are jointly allocated to a certain service or service set at this time, if the service or service set has margin to achieve the service index, Then, the proportions of MA and LA of the PU in the specific access area can be adjusted statically or dynamically. After adjustment, for example, as shown in the right half of FIG. 8, the MA of each PU is divided into a part of the storage space and classified as LA.

In an example, the service index may include a service storage capacity index and a service QoS index.

In one example, the processing unit in the SSD controller can adjust the ratio of MA and LA in the PU statically or dynamically. The processing unit determines whether there is unused storage space in the PU in the specific access area. If the processing unit determines that there is unused storage space, it continues to determine whether the service or service set allocated to the specific access area can achieve the service storage capacity index and the service QoS index with margin. If the processing unit determines that the service storage capacity index and the service quality QoS index can be achieved with margin, the ratio of MA and LA in the PU of the specific access area is adjusted.

In another example, achieving a service storage capacity index with margin may be that the storage capacity allocated to the service or service set allocated to the specific access area is greater than the storage capacity required by the service or service set. It is understandable that for different storage media, the storage capacity of a single PU is also different. For a PU with a large capacity, there is still unused storage capacity after the service or service set is allocated. Or, the storage capacity allocated to the business or business set allocated to the specific access area is equal to the storage capacity required by the business or business set, but the storage capacity allocated to the business or business set at the current moment has not been allocated by the business or business set. The business or business collection is fully used. For example, during a certain period of time, the business or business set does not fully use the allocated storage capacity temporarily; or because the business or business set will only be fully utilized in the later stage of execution, then the business or business set will not fully use the allocated storage capacity for the time being. The allocated storage capacity cannot be used in the early stage of the business collection. Or it may be that after the initial adjustment of the MA and LA in the specific access area, the storage capacity allocated to the service or service set allocated to the specific access area is still greater than the storage capacity required by the service or service set. In one example, after a service or service set is allocated to a specific access area, the processing unit will compress logical partitions to achieve preliminary adjustments according to the storage capacity required by the service or service set. Normally, the initial adjustment of the processing unit is to compress the logical partition to just meet the storage capacity requirements of the service or service set at one time, and use the remaining part as the LA area. It is understandable that for the remaining LA, the allocated service or service set cannot be perfectly matched, so the storage capacity for the service or service set subsequently allocated to the LA may still be excessive.

In another example, the service QoS index with margin can be that the actual QoS of the service or service set assigned to the specific access area is greater than or equal to the QoS index of the service or service set; or, the actual QoS assigned to the specific access area is greater than or equal to the QoS index of the service or service set; Within a certain period of time, the service or service set in the visit area can ignore the QoS index requirements of the service or service set. For example, if the storage system supports the NVMe protocol, the non-deterministic window (NDWIN) interval in the PLM of the input and output determinism (IOD) in the protocol will not affect the business. Or the QoS of the service set is restricted.

In an example, if the above-mentioned service storage capacity index and service QoS index are satisfied at the same time, the processing unit in the SSD controller can perform the logical partitioning of the service or service set allocated to the specific access area. Proportion adjustment. For example, the logical partition occupied by the service or service set can be compressed until the service storage capacity index and the service QoS index just meet the requirements. Mark the remaining logical partitions as LA and continue to allocate them to other services or service sets for use.

Those skilled in the art can understand that the proportion of logical partitions to which each PU belongs may be inconsistent.

In another embodiment, if the service QoS index of the assigned service or service set does not meet the QoS index of the service or service set, the processing unit in the SSD controller can perform the processing of the MA and LA in the PU of the specific access area. The proportion is adjusted.

In an example, the processing unit in the SSD controller determines the first ratio value in the specific access area according to the service storage capacity index. Wherein, the first ratio value is the ratio value between MA and LA in the specific access area. Determine the first QoS index of the service or service set allocated to a specific access area through pilot or model analysis and estimation. If the first QoS indicator is less than the service QoS indicator, the first ratio value is adjusted. Those skilled in the art should note that at this time, since the processing unit traverses all combinations of the first ratio values, but still cannot meet the QoS index of the service or service set, the first ratio value needs to be adjusted to ensure that the QoS of the service or service set is satisfied. index. Through pilot or model analysis and estimation, determine the adjusted second QoS index of the service or service set, and determine whether the second QoS index is greater than or equal to the service QoS index. If the second QoS indicator is still less than the service QoS indicator, continue to adjust the first ratio value of the PU in the specific access area, and determine whether the new first QoS indicator is greater than or equal to the service QoS indicator. If the new first QoS index is still smaller than the service QoS index, repeat the above steps and continue to adjust the first ratio value. Until the QoS index of the service or service set is greater than or equal to the service QoS index.

In another example, when there are multiple PUs, after determining that the first QoS indicator is less than the service QoS indicator, the first ratio value may be kept unchanged, and the second ratio value of the multiple PUs in the specific access area may be adjusted. Those in the field should note that adjusting the second ratio value at this time can be understood as the ratio of MA to LA in the special access area unchanged, and the ratio of MA to LA of each PU in the special access area is adjusted. Determine the third QoS index of the adjusted service or service set through pilot or model analysis and estimation. Wherein, the second ratio value is the ratio value of the MA and LA of each PU in the specific access area. If the third QoS index is less than the service QoS index, adjust the first ratio value, and adjust the second ratio value of one or more PUs in multiple PUs at the same time, and determine the adjusted service or service through pilot or model analysis and estimation Set the second QoS indicator; if the second QoS indicator is less than the service QoS indicator, keep the adjusted first ratio unchanged, continue to adjust the second ratio of PU, and determine whether the new third QoS indicator is greater than or equal to Service QoS index; if the new third QoS index is still less than the service QoS index, repeat the above steps and continue to adjust the first ratio value until the QoS index of the service or service set is greater than or equal to the service QoS index.

In another embodiment, the processing unit in the SSD controller may also have an LA in the PU in the specific access area allocated by the service or service set, and when part of the LA is not used, the unused part of the LA is allocated to other Business or business collection. It should be understood that the LA in this embodiment may be, for example, the LA divided after adjusting the scale in FIG. 8. The processing unit can then redistribute the divided LA to other services or service sets for use.

Of course, what needs to be understood is that for other businesses or business sets, certain conditions must also be met before the allocated LA can be leased. In an example, the processing unit may determine that if other services or service sets ignore the requirements of QoS indicators within a specific time, it may allocate the unused part of LA to other services or service sets. It is understandable that all unused LAs can also be allocated to other services or service groups, and then the ratio of MA to LA can be adjusted according to the service indicators of other services or service groups as shown in Figure 8 above. This application is not limited here. In another example, it may be determined through pilot projects or model analysis and estimation to determine whether the QoS index of the other service or service group is greater than or equal to the service QoS index after the other service or service group accepts the allocated unused part of the LA. When the QoS index of other services or service groups is greater than or equal to the service QoS index, it is determined to allocate the unused part of the LA to other services or service groups. It is understandable that when renting unused LAs for other services or service groups, it is necessary to ensure that the QoS indicators of other services or service groups are not affected. Only then can the LA of other concurrent units be rented to achieve the corresponding purpose.

In another example, the purpose may be to lease a concurrent unit as an OP for garbage collection (garbage collection, GC) or for wear leveling between different PU collections. Or it can be used to defragment the disk or rebuild data. This application is not limited here.

Those skilled in the art should note that for other businesses or business sets, the leased logical partition can have a small capacity ratio relative to the logical partitions allocated by other businesses or the business set itself. At the same time, the priority of the logical partition is lower than that of other businesses or business sets. The priority of the logical partition assigned by the service set itself. For other businesses or business collections, the logical partitions of other PUs can be leased to improve the write concurrency. After temporarily improving the write concurrency, the data can be moved from the leased logical partitions to free logic in a short period of time. Partition area to avoid data loss.

In this application, by adjusting the proportions of different logical partitions in the concurrent unit, the proportion of PUs occupied by services can be reduced, and idle resources can be released at the same time, thereby further improving the utilization of storage media. At the same time, for certain services or service sets, by leasing the LA of other PUs, the IOPS of the service is temporarily increased, and the co-location of the “high QoS, low IOPS demand” service and the “low QoS, high IOPS demand” service is realized and solved The problem of repeated construction in the deployment of mixed services is solved.

FIG. 9 is a schematic diagram of a physical location rotation of logical partitions in a concurrent unit provided by an embodiment of the application.

As shown in FIG. 9, in one embodiment, for PUs in the storage medium array, which are located in the same PU, if certain conditions are met, the service or service set assigned to the MA can actively trigger the physical location rotation.

In an example, the specific conditions can include: a service or service set triggers a specific round-robin daemon; or a service or service set within a specific time, ignoring the requirements of service QoS indicators. It can be understood that, for example, if the storage system supports the NVMe protocol , Then in the NDWIN interval in the PLM of the IOD in the agreement, there is no restriction on the QoS of the service or service set; or one or more PUs allocated by the service or service set are performing background operations and do not affect the service QoS The indicator, for example, may be a necessary background operation such as GC, but the operation does not affect the achievement of the QoS indicator of the service or service set allocated on the MA; or at least one host or virtual machine actively initiates it; or the PU fails and needs to be necessary Applause for detection or repair, etc. Those skilled in the art should note that it may also include other arbitrary preset specific conditions, which are not limited in this application.

In another example, the physical position rotation can be performed in the following manner. The PU in the storage medium array first selects the physical block to be rotated from the MA. After that, select an idle block in the LA in the same PU. As shown in ① in Figure 9, first select the physical block to be rotated from the selected MA, that is, the grid area in the white MA area in ①. And then select a free block in the LA in the same PU, that is, the slashed area in the dark LA area in ①. It is understandable that the slash indicates that there is no valid data in idle, the grid indicates valid data, the white indicates that the area attribute is MA, and the dark color indicates that the area attribute is LA. After that, the valid data in the block to be rotated is written into the free block, which is shown in ② in Fig. 9. It can be seen that the free block in the LA area has written valid data, that is, the dark grid area. Then, modify the attribute of the free block to MA, mark the data in the block to be rotated as invalid, and modify the attribute to LA. That is, as shown in Fig. 9 ③, modify the block attribute of the valid data to MA, that is, the white grid MA area in ③, and mark the area data to be rotated in the original MA as invalid, and modify the attribute, that is, in ③ The dark slash LA area.

FIG. 10 is a schematic diagram of another physical location rotation of logical partitions in a concurrent unit provided by an embodiment of the application.

As shown in FIG. 10, in another example, the physical position rotation can also be performed in the following manner. The PU in the storage medium array first selects the physical block to be rotated from the MA. After that, select an idle block in the MA in the same PU. It is understandable that this method is a rotation of the physical position inside the MA. As shown in ① in Figure 10, first select the physical block to be rotated from the selected MA, that is, the grid area in the white MA area in ①. And then select a free block in the MA in the same PU, that is, the slashed area in the white MA area in ①. It is understandable that the slash indicates that there is no valid data in idle, the grid indicates valid data, and the white indicates that the area attribute is MA. After that, the valid data in the block to be rotated is written into the free block, which is shown in ② in Fig. 9. It can be seen that the free block in the MA area has written valid data, that is, the white grid area. Then, the data in the block to be rotated is marked as invalid. That is, the area data to be rotated in the original MA shown in ③ in FIG. 10 is marked as invalid, that is, the white diagonal MA area in ③.

Those skilled in the art should note that for the physical rotation modes shown in FIG. 9 and FIG. 10, the rotation sequence can be arbitrarily adjusted according to actual conditions, which is not limited in this application.

It is understandable that through the physical rotation mode shown in Figure 9 and Figure 10, the rotation is actively triggered by the service or service set allocated in the high-priority logical partition after meeting specific conditions, and the block used to be exchanged can come from the logic Partition or logical partition with lower priority. In this way, the wear leveling of the physical components of the storage medium is realized. For some static PU logical partition ratios, the dynamic adjustment of the partition ratio can also be achieved through the rotation of physical blocks and data movement.

In this application, physical rotation is performed to achieve the wear leveling of the physical blocks in the PU. For some static proportions of logical partitions, the dynamic adjustment of the proportions of logical partitions can be achieved more conveniently through physical rotation.

FIG. 11 is a schematic diagram of a dichotomy of a logical partition of a concurrent unit according to an embodiment of the application.

As shown in FIG. 11, in one embodiment, each PU in the storage medium array can also be used to: if the area divided by the current PU cannot match the priority of the service or service set, the MA is divided into the first priority MA and second priority LA. For example, as shown in ① in Figure 11, if the PU is currently divided into two categories only based on the QoS indicators of the service or service set, that is, MA and LA1. The priority of the service or service set allocated on the MA is higher than the priority of the service or service set allocated on LA1. If there is a third service or service set at this time, but the priority of the third service or service set cannot match the priority of MA and LA 1 in ①. At this time, the MA can be divided into two again, that is, as shown in Figure 11 ②, the original MA is divided into MA and LA 1, and the original LA 1 becomes LA 11. At this time, the priority order of the logical partitions in the PU is MA>LA1>LA11. In another example, the original LA 1 can also be divided into two correspondingly, and the LA can be divided into a third priority LA and a fourth priority LA. That is, the original LA 1 is divided into LA 11 and LA12 in ②. At this time, the priority order of the logical partitions in the PU is MA>LA1>LA11>LA12. Those skilled in the art should note that the priority of the new logical partition is a subdivision of the priority of the original logical partition, wherein the priority of the area after each binary division does not exceed the priority before the division of the area. In other words, if the priority of other logical partitions was originally higher than the priority of the original logical partition, the priority of other logical partitions is still higher than the priority of all new logical partitions. Similarly, if the priority of other logical partitions was originally lower than the priority of the original logical partition, the priority of other logical partitions is still lower than the priority of all new logical partitions.

In another example, if the priority of the logical partition of the PU after the division meets the requirements of the service or service set, the division may not be continued. If the logical partition of the PU still does not meet the requirements of the service or service set after the division, the logical partitions in the PU are continuously divided into binary divisions until the divided logical partition meets the requirements of the priority of the service or service set. For example, as shown in ③ in Figure 11, you can continue to divide the MA into two, that is, divide the original MA into MA and LA 1. Similarly, divide the original LA 1 into LA 11 and LA12, and divide the original LA 11 into LA 111 And LA112, divide the original LA 12 into LA 121 and LA122. At this time, the priority order of the logical partitions in the PU is MA>LA1>LA11>LA12>LA111>LA112>LA121>LA122. If you continue to divide into two, you can repeat the above process, for the convenience of description, I will not repeat it again.

This application divides the concurrent units into different logical partitions and assigns different priority dimensions to refine management granularity, increase management flexibility, and improve storage media utilization while ensuring service QoS indicators.

Fig. 12 is a schematic diagram of a concurrent unit distribution service provided by an embodiment of the application.

As shown in Figure 12, in a more specific example, the NVMe protocol is supported in this scenario. The figure shows multiple different services, and it is understandable that the service may be a service set that includes multiple services. For example, NVM set A service set, NVM set B service set, and NVM set C service set may be service sets with high QoS requirements, and NVM set D service set and NVM set E service sets may be service sets with lower QoS requirements. As shown in FIG. 12, after the NVM set A service set is preferentially allocated to the specific access area of the service set, the processing unit of the SSD controller determines that the capacity index and QoS index of the service set at this time are both achieved but not rich. Then the logical partition of the PU in the specific access area may only contain the MA. For example, in Fig. 12, the NVM set A service set is represented by PU a1 to PU an framed by the solid line frame of the service set.

After the NVM set B service set and the NVM set C service set are allocated to the specific access area of the corresponding service set, the processing unit of the SSD controller determines that the capacity index and QoS index of the corresponding service set are both affluent at the same time. Then the processing unit can compress the logical partitions allocated by the corresponding service set. And assign the divided logical partitions as LA to other services or service sets for use. As shown in Fig. 12, the MA occupied by the NVM set B service set and the NVM set C service set are composed of the MAs from PU b1 to PU bn and the MAs from PU c1 to PU cn by the solid line boxes of the respective service sets.

After the NVM set D service set is allocated to the specific access area of the corresponding service set, the processing unit of the SSD controller determines the capacity index and QoS index of the corresponding service set, which cannot be achieved. At this time, the processing unit can improve the concurrency and maximum bandwidth capacity of the NVM set D service set in the short term by renting the LA of other PUs. For example, the LA occupied by the NVM set D service set shown in FIG. 12 is composed of the dashed box of its service set, including the LAs from PU d1 to PU dn that originally belonged to the NVM set D service set, as well as the LA leased from other PUs. For example, part LA from PU b1 to PU bn and part LA from PU c1 to PU cn.

For the NVM set E service set, the service set requires the highest IOPS, but there is no clear requirement for the QoS index. At this time, as much LA that is rich in the PU can be leased to the NVM set E service set for use. To ensure that the IOPS and capacity utilization of the storage medium are maximized. For example, the LA occupied by the NVM set E service set shown in Figure 12 is composed of the dashed box of its service set, including the LA leased from other PUs, such as the part of the LA from PU b1 to PU bn and the part of the LA from PU c1 to PU cn. .

Those skilled in the art should note that in Figures 4 to 12 in this application, the storage medium is preferentially allocated based on services or service sets with high QoS requirements. It is understandable that services or service sets with low QoS requirements can also be assigned preferentially, which is not limited in this application.

In an embodiment, if the storage system running on the storage medium provided in FIGS. 4 to 12 does not support the NVMe protocol, the concept of NVM Set and multi-user/multi-service does not exist. In this scenario, by combining multi-stream technology, data with high QoS requirements can be stored in the MA area of the PU, and data with lower QoS requirements can be stored in the LA area of the PU. Among them, the multi-streaming technology can be to mark data feature tags through the host or to split streams based on the data in the disk. The usage mode of the PU in the storage medium in this scenario is the same as that shown in FIG. 4 to FIG. 12, and for convenience of description, it is not repeated here.

This application also reduces the management complexity and satisfies the application requirements of most scenarios by dually partitioning the PU models. Through the recursive method, the multi-partition is converted into a binary partition problem, which reduces the application complexity.

The present application also provides an SSD controller, which is coupled with the storage medium array. The SSD controller is the same as the SSD controller in FIGS. 2-12, and can execute any function or method of the SSD controller in FIGS. 2-12. For the convenience of description, I will not repeat them here.

FIG. 13 is a schematic diagram of another application scenario provided by an embodiment of the application.

As shown in FIG. 13, in another embodiment, in the application scenario shown in the figure, one or more VMs or VFs can also be run on the host that delivers the service or service set. It is understandable that the host is still connected to the SSD controller through various interfaces such as NVMe, SAS, PCIe. The SSD controller is connected to the corresponding storage medium array through the NFI. Those in the art should note that the SSD controller can choose to connect to the storage medium through other interfaces according to the interface specifications opened by the actual manufacturer of the storage medium, so as to ensure that the host can perform data on the storage medium array through the SSD controller. Read, write, wipe and other operations The usage mode of the PU in the storage medium in this scenario is the same as that shown in FIG. 4 to FIG. 12, and for convenience of description, it is not repeated here.

FIG. 14 is a schematic diagram of another application scenario provided by an embodiment of the application.

As shown in FIG. 14, in an embodiment, in the application scenario shown in the figure, the controller circuit for controlling the storage medium array and the storage medium array may also be packaged on an integrated chip. The host is connected to the integrated chip through interfaces such as universal flash storage (UFS) or embedded multimedia card (eMMC), and accesses data. The usage mode of the PU in the storage medium in this scenario is the same as that shown in FIG. 4 to FIG. 12, and for convenience of description, it is not repeated here.

FIG. 15 is a schematic diagram of another application scenario provided by an embodiment of the application.

As shown in FIG. 15, in one embodiment, the SSD controller may include a first SSD controller and a second SSD controller. The first SSD controller is located in one of the at least one host, and the second SSD controller and the storage medium array are packaged on an integrated chip.

In an example, an "open channel storage system" can also be used on the storage medium. In this type of system, for the controller that controls the storage medium array, part of its functions are moved to the host for implementation, such as the controller 1 shown in FIG. 15. At the same time, another part of the controller functions and the storage medium array are packaged on an integrated chip, such as the controller 2 shown in FIG. 15. Those skilled in the art can understand that the connection mode between the controller 1 and the controller 2 can be any connection mode in the prior art, which is not limited in this application. The usage mode of the PU in the storage medium in this scenario is the same as that shown in FIG. 4 to FIG. 12, and for convenience of description, it is not repeated here.

FIG. 16 is a flowchart of a method for increasing the utilization rate of a solid state drive provided by an embodiment of the application.

As shown in FIG. 16, the present application also provides a method for improving the utilization rate of a solid state drive. The method may include the following steps:

S1601. Receive characteristic information of at least one service or service set.

In an example, the characteristic information of at least one service or service set delivered by at least one host or virtual machine may be received. Among them, the characteristic information includes service quality QoS indicators.

S1602: Sort at least one service or service set according to the size of the QoS index of the service or service set.

In an example, one or more at least one service or service set can be sorted according to the size of the QoS index of the service or service set.

In another example, it is also possible to determine the minimum number of concurrent unit PUs required for the characteristic information of each service or service set.

In one embodiment, the characteristic information includes typical IOPS requirements, service storage capacity indicators, and typical capacity requirements. Determine the number of PUs of the first concurrent unit according to the typical IOPS requirements of the service or service set. And according to the service storage capacity index or typical capacity requirements of the service or service set, the second PU number is determined. The largest number of the first number of PUs and the second number of PUs is taken as the minimum number of PUs required.

In another embodiment, the characteristic information further includes the write volume requirement of the service or the service set, the write amplification factor of the service or the service set, and the program-erased PE life of the solid-state hard disk. According to the typical capacity requirements or business storage capacity indicators, the write volume requirements of the business or business set, the write amplification factor of the business or business set, and the PE lifetime of the solid state drive, determine the OP demand for the excess redundant space. Combine OP requirements and typical capacity requirements or business storage capacity indicators to determine the second PU number.

In an example, in the same PU, the priority of the IO operation that accesses the MA is higher than the priority of the IO operation that accesses the LA. Among them, the IO operation is the IO operation of a business or a business set.

In an example, the priority scheduling method may be any one or a combination of strict priority scheduling, probabilistic priority scheduling, deprivation priority scheduling, or non-deprivation priority scheduling.

In one example, strict priority scheduling is that IO operations with high priority are scheduled before IO operations with low priority. Probabilistic priority scheduling is that the probability of scheduling IO operations with high priority is higher than the probability of scheduling IO operations with low priority. The deprivation priority scheduling is that IO operations with high priority are allowed to interrupt the IO operations with low priority being executed under preset conditions. Non-deprivation priority scheduling is that IO operations with high priority are not allowed to interrupt the IO operations with low priority being executed under any conditions.

S1603: Allocate the main area MA of the concurrent unit PU to the service or service set with the largest QoS index for use, and allocate the rental area LA of the PU to other services or service sets for use according to the ranking.

In one embodiment, when at least one host or virtual machine delivers the characteristic information of a service or service set, the MAs of at least PUs are allocated to the service or service set for use; when at least one host or virtual machine When issuing the characteristic information of at least two services or service sets, sort according to the size of the QoS indicators of the at least two services or service sets; assign the MA with the least number of PUs to the service or service set with the largest QoS indicator for use ; According to the sorting, the LAs of the least PUs are allocated to the remaining services or service sets for use.

In one embodiment, assigning the LAs of the least PU PUs to the remaining services or service sets according to the sorting includes: according to the sorting order, assigning LAs with different priorities to the remaining services or service sets in sequence, where the priority of the LA is the same as The QoS indicators of the assigned service or service set are positively correlated.

In one embodiment, if the minimum number of PUs allocated by a service or service set has an LA, and part of the LA is not used, then the unused part of the LA is allocated to other services or service sets.

In one example, if other services or service sets neglect the requirements of QoS indicators within a certain period of time, it is determined to allocate the unused part of LA to the other services or service sets; or through pilot or model analysis and estimation, it is determined After other services or service sets accept the allocated unused part of the LA, whether the QoS indicators of other services or service sets are greater than or equal to the business QoS indicators; if the QoS indicators of other services or service sets are greater than or equal to the business QoS indicators, it is determined to be The unused part of the LA is allocated to other services or service sets.

In another embodiment, if the system that executes the above method does not support the NVMe protocol, the host of this application can also combine the multi-stream technology to split the data by marking the data feature label of the host or based on the in-disk data splitting algorithm. The data stream contains one or more of the following characteristic information: QoS indicator characteristics, IOPS indicator characteristics, IO continuity characteristics, whether it is a user request characteristic, etc. And store the data with high QoS index requirements in the high-priority logical partition of the PU, such as MA. Store data with relatively low QoS index requirements in a low-priority logical partition of the PU, such as LA. Among them, those skilled in the art should note that the number of data streams that can be shared by a logical partition with a higher priority is smaller.

FIG. 17 is a flowchart of a method for adjusting the proportion of logical partitions of a concurrent unit according to an embodiment of the application.

As shown in FIG. 17, after S1603, the present application can also adjust the proportions of different logical partitions in the PU, that is, adjust the ratio between MA and LA. The method may further include the following steps:

S1701: Determine whether the service or service set has a margin to achieve the service index within the assigned minimum number of PUs.

In one embodiment, if the service or service set has a margin to achieve the service index in the assigned minimum number of PUs, the ratio of MA to LA in the minimum number of PUs is adjusted.

In another embodiment, if there is unused storage space in the least number of PUs and the service or service set allocated to the least number of PUs has margin to achieve the service storage capacity index and the service quality QoS index, then adjust The ratio of MA and LA in the minimum number of PUs.

In an example, a marginal achievement of the business storage capacity index may include: the storage capacity allocated to the business or the business set is greater than the storage capacity required by the business or the business set; or the storage capacity allocated to the business or the business set is equal to the business Or the storage capacity required by the service collection, but the storage capacity allocated to the service or service collection at the current moment is not fully used by the service or service collection.

In an example, achieving the service QoS index with margin may include: the actual QoS of the service or service set is greater than or equal to the QoS index; or the service or service set ignores the QoS index requirement within a specific time.

S1702: Determine a first ratio value of the least number of PUs according to the service storage capacity index, where the first ratio value is the ratio value of MA and LA in the least number of PUs.

S1703: Determine the first QoS indicator of the service or service set through pilot projects or model analysis and estimation.

S1704: Determine whether the first QoS indicator is less than the service QoS indicator.

In one embodiment, if the first QoS indicator is less than the service QoS indicator, then enter S1705; if the first QoS indicator is greater than or equal to the service QoS indicator, then enter S1711.

S1705: Keep the first ratio value unchanged, and adjust the second ratio values of multiple PUs in the least number of PUs.

S1706: Determine the adjusted second QoS index of the service or service set through pilot projects or model analysis and estimation, where the second ratio value is the ratio value of the MA and LA of each PU in the least number of PUs.

S1707: Determine whether the second QoS indicator is less than the service QoS indicator.

In one embodiment, if the second QoS indicator is less than the service QoS indicator, then enter S1708; if the second QoS indicator is greater than or equal to the service QoS indicator, then enter S1711.

S1708: Adjust the first ratio value and simultaneously adjust the second ratio value of one or more PUs in the least number of PUs.

S1709: Determine the third QoS indicator of the adjusted service or service set through pilot projects or model analysis and estimation.

S1710: Determine whether the third QoS indicator is greater than or equal to the service QoS indicator.

In one embodiment, if the third QoS indicator is less than the service QoS indicator, return to S1705, keep the adjusted first ratio value unchanged, continue to adjust the second ratio value of multiple PUs in the least number of PUs, and determine Whether the new second QoS indicator is greater than or equal to the service QoS indicator. If the third QoS indicator is greater than or equal to the service QoS indicator, enter S1711.

S1711, end the adjustment.

FIG. 18 is a flowchart of a method for rotating the physical position of a concurrent unit according to an embodiment of the application.

As shown in FIG. 18, after S1603, the present application can also rotate the physical locations of different logical partitions in the PU. The method may also include the following steps:

S1801: Determine whether a specific condition is met.

In one embodiment, in the same PU, if a specific condition is met, the service or service set located in the MA actively triggers the physical location rotation.

In an example, the specific conditions may include: the service or service set triggers the rotation daemon; or the service or service set ignores the requirements of the service QoS index within a specific time; or one or more PUs allocated by the service or service set are Perform background operations without affecting service QoS indicators; or at least one host or virtual machine actively initiates them; or PU fails.

S1802: If a specific condition is met, select a physical block to be rotated from the MA.

S1803: Select an idle block in the same PU.

In one embodiment, one free block in the LA in the same PU is selected.

In another embodiment, an idle block in the MA in the same PU is selected.

S1804: Write valid data in the block to be rotated into an idle block.

S1805: Mark the data in the block to be rotated as invalid.

In one embodiment, if the selected free block is from LA, it is also necessary to modify the attribute of the free block to which valid data is written to MA, and to modify the attribute of the original block to be rotated to LA.

As shown in Figure 19, after S1603 in this application, the logical partitions in the PU can be divided into more detailed divisions by recursive method, that is, the logical partitions are continuously divided into two logical partitions, thereby dividing logical partitions with different priorities. The method can also include the following steps:

S1901: Determine whether the area divided by the current PU can match the priority of the service or service set.

In one embodiment, if the area divided by the current PU cannot match the priority of the service or service set, S1902 is executed. In another embodiment, if the area divided by the current PU can match the priority of the service or service set, S1905 is executed.

S1902: Divide the MA into a first priority MA and a second priority LA.

S1903: Divide the LA into a third priority LA and a fourth priority LA.

S1904: Determine whether the area of the divided PU can match the priority of the service or service set.

In an embodiment, if the divided PU area cannot match the priority of the service or service set, S1902 is executed. In another embodiment, if the divided area of the divided PU can match the priority of the service or service set, S1905 is executed.

In another embodiment, it can be understood that if the divided PU area still cannot match the priority of the service or service set, the binary division of the area within the PU will continue until the priority of the service or service set is matched. level. Among them, the priority of the area after each binary division does not exceed the priority before the division of the area.

S1905, end subdivision.

Those skilled in the art should note that when there are more than two logical partitions in the PU, the service or service set with the higher QoS requirement is given priority to the logical partition with the higher priority. For a logical partition with a higher priority, the number of services shared on the logical partition is smaller.

Those skilled in the art should be aware that, in one or more of the foregoing examples, the functions described in the embodiments of the present application may be implemented by hardware, software, firmware, or any combination thereof. When implemented by software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on the computer-readable medium. The computer-readable medium includes a computer storage medium and a communication medium, where the communication medium includes any medium that facilitates the transfer of a computer program from one place to another. The storage medium may be any available medium that can be accessed by a general-purpose or special-purpose computer.

The steps of the method or algorithm described in combination with the embodiments disclosed herein can be implemented by hardware, a software module executed by a processor, or a combination of the two. The software module can be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or all areas in the technical field. Any other known storage media.

The specific implementations described above further describe the purpose, technical solutions, and beneficial effects of the application. It should be understood that the foregoing are only specific implementations of the application and are not intended to limit the scope of the application. The scope of protection, any modification, equivalent replacement, improvement, etc. made on the basis of the technical solution of this application shall be included in the scope of protection of this application.

Claims

A storage device, characterized by comprising: a storage medium array and a solid-state hard drive (SSD) controller coupled to the storage medium array;

The storage medium array includes a plurality of concurrent units PU, the PU includes at least one of the main area MA or the lease area LA, wherein the MA is used to be allocated and used by one service, and the LA is used to be used by at least one service. Allocation

The SSD controller is configured to receive characteristic information of at least one service, where the characteristic information includes a service quality QoS indicator;

The SSD controller is further configured to allocate the MA of the PU to the service use with the largest QoS indicator according to the QoS indicator, and allocate the LA of the PU to the rest of the service use.
The device according to claim 1, wherein the characteristic information further includes a typical read/write input/output number IOPS requirement per second, and a service storage capacity index or typical capacity requirement, and the SSD controller is further used for:

Determine the first number of PUs required by the service according to the typical IOPS requirements;

Determine the second number of PUs required by the service according to the service storage capacity index or the typical capacity requirement;

The allocating the PU of the PU to the service use with the largest QoS index, and allocating the PU of the PU to the rest of the service use includes:

Allocate the MA of the first PU among the multiple PUs to the service use with the largest QoS index, and allocate the LA of the first PU to the rest of the service use, where the first PU includes The number of PUs is the maximum value of the first number and the second number.
The device according to claim 1, wherein in the same PU, the SSD controller controls the priority of the input and output IO operations that access the MA is higher than the priority of the IO operations that access the LA Level; wherein, the IO operation is the IO operation of the service.
The device according to claim 1, wherein the SSD controller is further configured to:

Sort at least one of the services from high to low according to the QoS index;

According to the ordering, LAs with different priorities are assigned to the remaining services in sequence, wherein the priority of the LA is positively correlated with the QoS index of the assigned service.
The device according to any one of claims 1-4, wherein the SSD controller is further configured to:

If the service has a margin in the allocated PU to achieve the service index, the ratio of MA and LA in the PU is adjusted.
The device according to claim 5, wherein the SSD controller is further configured to:

Determining a first ratio value of the PU according to a service storage capacity index, where the first ratio value is a ratio value of MA and LA in the PU;

Determine the first QoS indicator of the service through pilot or model analysis and estimation;

If the first QoS indicator is less than the service QoS indicator, adjust the first ratio value, and determine the adjusted second QoS indicator of the service through pilot or model analysis and estimation;

If the second QoS indicator is less than the service QoS indicator, continue to adjust the first ratio value of the PU, and determine whether the new first QoS indicator is greater than or equal to the service QoS indicator;

If the new first QoS index is still less than the service QoS index, repeat the above steps and continue to adjust the first ratio value until the QoS index of the service is greater than or equal to the service QoS index.
The device according to claim 6, wherein when the PU is multiple PUs, after determining that the first QoS index is less than the service QoS index, the SSD controller is further configured to:

Keep the first ratio value unchanged, adjust multiple second ratio values in the multiple PUs, and determine the adjusted third QoS indicator of the service through pilot or model analysis and estimation, where the The second ratio value is a ratio value of MA and LA of one PU of the plurality of PUs;

If the third QoS indicator is less than the service QoS indicator, adjust the first ratio value, and simultaneously adjust the second ratio value of one or more PUs in the multiple PUs, and pass a pilot or model Analyze and estimate, and determine the adjusted second QoS indicator of the service;

If the second QoS indicator is less than the service QoS indicator, keep the adjusted first ratio value unchanged, continue to adjust the second ratio value of the PU, and determine a new third QoS indicator Whether it is greater than or equal to the service QoS index;

If the new third QoS indicator is still less than the service QoS indicator, repeat the above steps and continue to adjust the first ratio value until the QoS indicator of the service is greater than or equal to the service QoS indicator.
The device according to any one of claims 1-7, wherein the SSD controller is further configured to:

If the PU allocated by the service has the LA, and part of the LA is not used, then the unused part of the LA is allocated to other services.
The device according to claim 8, wherein the SSD controller is further configured to:

If the other service ignores the requirements of the QoS indicator within a specific time, it is determined to allocate the unused part of the LA to the other service; or

Through pilot or model analysis and estimation, it is determined whether the QoS index of the other service is greater than or equal to the QoS index of the service after the other service receives the allocated unused part of the LA; if the QoS index of the other service is If it is greater than or equal to the service QoS index, it is determined to allocate the unused part of the LA to the other service.
The device according to any one of claims 1-9, wherein the storage medium array is further used for:

In the same PU, if a specific condition is met, the service located in the MA actively triggers the physical location rotation.
The device according to claim 10, wherein the storage medium array is further used for:

Select a physical block to be rotated from the MA;

Selecting an idle block in the LA in the same PU, or selecting an idle block in the MA in the same PU;

Writing valid data in the to-be-rotated block into the idle block;

If a free block in the LA in the same PU is selected, the attribute of the free block is modified to MA, the data in the block to be rotated is marked as invalid, and the attribute is modified to LA;

If an idle block in the MA in the same PU is selected, the data in the to-be-rotated block is marked as invalid.
The device according to claim 10 or 11, wherein the specific condition comprises:

The service triggers the rotation daemon; or

The service ignores the requirements of the service QoS index within a certain period of time; or

The PU allocated by the service is undergoing background operation and does not affect the service QoS index; or

At least one host or virtual machine actively initiates, wherein the host or virtual machine is also used to deliver at least one of the services; or

The PU is malfunctioning.
The device according to any one of claims 1-12, wherein the storage medium array is further used for:

If the area currently divided by the PU cannot match the priority of the service, divide the MA into a first priority MA and a second priority LA; and/or

Divide the LA into a third priority LA and a fourth priority LA;

If the divided area of the PU still cannot match the priority of the service, continue to divide the area in the PU into binary divisions until the priority of the service is matched; wherein, after each binary division The priority of the area does not exceed the priority before the area is divided.
The device according to any one of claims 1-13, wherein the SSD controller and the storage medium array are packaged on an integrated chip; or

The SSD controller includes a first SSD controller and a second SSD controller; the first SSD controller is located in at least one host, and the second SSD controller and the storage medium array are packaged in an integrated chip on.
The device according to any one of claims 1-14, wherein the storage medium array comprises at least one storage medium, and the storage medium is flash memory, dynamic random access memory DRAM, static random access memory SRAM, Any one of phase change memory PCM and storage-level memory SCM.
A solid state hard disk SSD controller, characterized in that the SSD controller is coupled with a storage medium array; wherein, the storage medium array includes a plurality of concurrent units PU, and the PU includes a main area MA or a rental area LA. At least one of, wherein the MA is used to be allocated and used by one service, and the LA is used to be allocated and used by at least one service;

The SSD controller is configured to receive characteristic information of at least one service, where the characteristic information includes a service quality QoS indicator;

The SSD controller is further configured to allocate the MA of the PU to the service use with the largest QoS indicator according to the QoS indicator, and allocate the LA of the PU to the rest of the service use.
The controller according to claim 16, wherein the characteristic information further includes a typical read/write input/output number IOPS requirement per second, and a service storage capacity index or typical capacity requirement, and the SSD controller is also used for:

Determine the first number of PUs required by the service according to the typical IOPS requirements;

Determine the second number of PUs required by the service according to the service storage capacity index or the typical capacity requirement;

The allocating the MA of the PU to the service use with the largest QoS index and allocating the LA of the PU to the remaining service uses include:

Allocate the MA of the first PU among the multiple PUs to the service use with the largest QoS index, and allocate the LA of the first PU to the rest of the service use, where the first PU includes The number of PUs is the maximum value of the first number and the second number.
The controller according to claim 16, wherein the SSD controller is further configured to control access to the MA input and output IO operations in the same PU with a higher priority than access to the LA IO operation The priority; wherein, the IO operation is the IO operation of the service.
The controller according to any one of claims 16-18, wherein the SSD controller is further configured to:

If the service has a margin in the allocated PU to achieve the service index, the ratio of MA and LA in the PU is adjusted.
The device according to any one of claims 16-18, wherein the SSD controller is further configured to:

If the area divided by the PU currently cannot match the priority of the service, the SSD controller controls the storage medium array to divide the MA into a first priority MA and a second priority LA, and/or Divide the LA into a third priority LA and a fourth priority LA;

If the divided area of the PU still cannot match the priority of the service, continue to divide the area in the PU into binary divisions until the priority of the service is matched; wherein, after each binary division The priority of the area does not exceed the priority before the area is divided.
A method for improving the utilization rate of a solid state drive, characterized in that the method includes:

Receiving characteristic information of at least one service, where the characteristic information includes a service quality QoS indicator;

According to the QoS index, the main area MA of the concurrent unit PU is allocated to the service use with the largest QoS index, and the rental area LA of the PU is allocated to the rest of the service use, where PU includes MA or LA at least one.
The method according to claim 21, wherein the characteristic information further includes a typical read/write input/output number IOPS requirement per second, and a service storage capacity index or typical capacity requirement, and the method further comprises:

Determine the first number of PUs required by the service according to the typical IOPS requirements;

Determine the second number of PUs required by the service according to the service storage capacity index or the typical capacity requirement;

The allocating the MA of the PU to the service use with the largest QoS index and allocating the LA of the PU to the remaining service uses include:

Allocate the MA of the first PU among the multiple PUs to the service use with the largest QoS index, and allocate the LA of the first PU to the rest of the service use, where the first PU includes The number of PUs is the maximum value of the first number and the second number.
The method according to claim 21, characterized in that, in the same PU, the SSD controller controls the priority of the input and output IO operations that access the MA is higher than the priority of the IO operations that access the LA ; Wherein, the IO operation is the IO operation of the service.
The method according to any one of claims 21-23, wherein the method further comprises:

If the service has a margin to achieve the service index in the assigned minimum number of PUs, the ratio of MA and LA in the minimum number of PUs is adjusted.
The method according to any one of claims 21-24, wherein the method further comprises:

If the PU allocated by the service has the LA, and part of the LA is not used, then the unused part of the LA is allocated to other services.
The method according to any one of claims 21-25, wherein the method further comprises:

In the same PU, if a specific condition is met, the service located in the MA actively triggers the physical location rotation.
The method according to claim 26, wherein the satisfying a specific condition comprises:

The service triggers the rotation daemon; or

The service ignores the requirements of the service QoS index within a certain period of time; or

One or more PUs allocated by the service are performing background operations and do not affect the QoS indicators of the service; or

At least one host or virtual machine actively initiates, wherein the host or virtual machine is also used to deliver at least one of the services; or

The PU is malfunctioning.
The method according to any one of claims 21-25, wherein the method further comprises:

If the area currently divided by the PU cannot match the priority of the service, divide the MA into a first priority MA and a second priority LA; and/or

Divide the LA into a third priority LA and a fourth priority LA;

If the divided area of the PU still cannot match the priority of the service, continue to divide the area in the PU into binary divisions until the priority of the service is matched; wherein, after each binary division The priority of the area does not exceed the priority before the area is divided.
A computer-readable storage medium, comprising instructions, which when run on a computer, cause the computer to execute the method according to any one of claims 21 to 28.
A computer program product containing instructions that, when run on a computer, causes the computer to execute the method according to any one of claims 21 to 28.