Method and system for facilitating a physically isolated storage unit for multi-tenancy virtualization

Abstract

One embodiment provides a system which facilitates organization of data. During operation, the system allocates, to a function associated with a host, a number of block columns to obtain a physical storage space for the function, wherein a block column corresponds to a block from each of a plurality of dies of a non-volatile storage device. In response to processing an incoming host write instruction and an internal background write instruction, the system allocates a first block column to the incoming host write instruction and a second block column to the internal background write instruction, thereby extending a lifespan of the non-volatile storage device by recycling the first block column when deleting a namespace or virtual machine associated with the function.

Description

BACKGROUND

Field

This disclosure is generally related to the field of data storage. More specifically, this disclosure is related to a method and system for facilitating a physically isolated storage unit for multi-tenancy virtualization.

Related Art

Today, various storage systems are being used to store and access the ever-increasing amount of digital content. A storage system can include various storage devices which can provide persistent memory, e.g., a solid state drive (SSD) and a hard disk drive (HDD). A cloud service can provide access to a storage system by using virtualization, in which a single physical storage drive can be used by multiple virtual machines (VMs). When a single VM is destroyed, the system may physically remove all data corresponding to the single VM, e.g., to prevent subsequent access to the data associated with the single VM. The performance of the single physical drive may be required to be of sufficient reliability to eliminate tails associated with latency distribution. Furthermore, an accelerated recycling of physical space in the single physical storage drive may extend the usage of the storage drive, which can result in a revenue increase. Additionally, providing reliability in performance may be beneficial for fulfillment of service level agreements (SLAs).

One current virtualization method involves implementing the input/output (I/O) virtualization to provide logical drives for multiple VMs, using a single root I/O virtualization (SRIOV). This method can expose multiple virtual functions (VFs), which can be instantiated by different VMs to form the logical drives. However, this method can result in some constraints, e.g.: data and I/O from different VMs may be stored in the same NAND block or page, which can result in a time-consuming process for data destruction, and can also trigger garbage collection; I/Os from multiple VMs may be placed in a random layout across the physical storage drives, which can create difficulties in balancing the I/O performance among the multiple VMs; the I/O distribution may be spread randomly across the multiple physical storage drives, which can result in hot spots and a traffic imbalance; and a single storage drive may not provide data recovery protection among the multiple physical storage drives.

Thus, while the SRIOV method can provide logical drives for multiple VMs, the above-described constraints can result in a decrease in the efficiency and performance of the overall storage system.

SUMMARY

One embodiment provides a system which facilitates organization of data. During operation, the system allocates, to a function associated with a host, a number of block columns to obtain a physical storage space for the function, wherein a block column corresponds to a block from each of a plurality of dies of the non-volatile storage device. In response to processing an incoming host write instruction and an internal background write instruction, the system allocates a first block column to the incoming host write instruction and a second block column to the internal background write instruction, thereby extending a lifespan of the non-volatile storage device by recycling the first block column when deleting a namespace or virtual machine associated with the function.

In some embodiments, the function is a virtual function. In response to receiving a command to delete a virtual machine associated with the virtual function, the system erases the number of block columns of the physical storage space allocated for the virtual function and returns the number of block columns to a block column pool.

In some embodiments, allocating the number of block columns comprises obtaining the number of block columns from a block column pool.

In some embodiments, in response to receiving the incoming host write instruction, the system writes data associated with the host write to at least the first block column allocated to the function.

In some embodiments, the system identifies a sealed block column which is filled with data. The system executes the internal background write instruction as a garbage collection process based on the second block column, by: copying valid data from blocks of the sealed block column to blocks of the second block column; erasing data stored in the blocks of the sealed block column; and returning the sealed block column to a block column pool.

In some embodiments, the non-volatile storage device is one of a plurality of non-volatile storage devices which communicate with a global flash translation layer. The global flash translation layer allocates the number of block columns to the function, and the allocated block columns correspond to at least two of the non-volatile storage devices. The function is one of a plurality of virtual functions to which the global flash translation layer allocates block columns.

In some embodiments, the global flash translation layer maps each virtual function to an allocated physical storage space, and each physical storage space includes block columns corresponding to at least two of the non-volatile storage devices.

In some embodiments, an erasure coding (EC) encoding/decoding module in a controller performs EC encoding/decoding for the functions. Data associated with the function is stored in the allocated number of block columns across the at least two non-volatile storage devices. The system performs, by the EC encoding/decoding module, EC encoding on the data prior to the data being stored in the allocated number of block columns to obtain an EC codeword. The system distributes the EC codeword to be stored in block columns in the allocated number of block columns across the at least two non-volatile storage devices.

In some embodiments, the system divides a physical storage capacity of a non-volatile storage device into a plurality of block groups, wherein a block group comprises a plurality of block columns.

BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 illustrates an architecture of an exemplary virtualization in a system, in accordance with the prior art.

FIG. 2 illustrates physically isolated storage space based on the storage capacity divided into block groups which include block columns, in accordance with an embodiment of the present application.

FIG. 3 illustrates an exemplary environment for virtualization and allocation of block columns from a block column pool, in accordance with an embodiment of the present application.

FIG. 4 illustrates three exemplary block columns, including a sealed block column for a host write, an open block column for a host write, and an open block column for a garbage collection process, in accordance with an embodiment of the present application.

FIG. 5 illustrates an exemplary storage controller with a global flash translation layer module which maps virtual functions to block columns from at least two storage devices, in accordance with an embodiment of the present application.

FIG. 6 illustrates an exemplary environment for erasure code (EC) encoding in a physical storage space, in accordance with an embodiment of the present application.

FIG. 7A presents a flowchart illustrating a method for facilitating organization of data, in accordance with an embodiment of the present application.

FIG. 7B presents a flowchart illustrating a method for facilitating organization of data, in accordance with an embodiment of the present application.

FIG. 7C presents a flowchart illustrating a method for facilitating organization of data, in accordance with an embodiment of the present application.

FIG. 8 illustrates an exemplary computer system that facilitates organization of data, in accordance with an embodiment of the present application.

FIG. 9 illustrates an exemplary apparatus that facilitates organization of data, in accordance with an embodiment of the present application.

In the figures, like reference numerals refer to the same figure elements.

DETAILED DESCRIPTION

The following description is presented to enable any person skilled in the art to make and use the embodiments, and is provided in the context of a particular application and its requirements. Various modifications to the disclosed embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments and applications without departing from the spirit and scope of the present disclosure. Thus, the embodiments described herein are not limited to the embodiments shown, but are to be accorded the widest scope consistent with the principles and features disclosed herein.

Overview

The embodiments described herein facilitate multi-tenancy virtualization by using physically isolated storage spaces across multiple storage drives.

As described above, virtualization is a technology in which a single physical storage drive can be used by multiple virtual machines (VMs). When a single VM is destroyed, the system may physically remove all data corresponding to the single VM, e.g., to prevent subsequent access to the data associated with the single VM. The performance of the single physical drive may be required to be of sufficient reliability to eliminate tails associated with latency distribution. Furthermore, an accelerated recycling of physical space in the single physical storage drive may extend the usage of the storage drive, which can result in a revenue increase. Additionally, providing reliability in performance may be beneficial for fulfillment of service level agreements (SLAs).

One current virtualization method involves implementing the input/output (I/O) virtualization to provide logical drives for multiple VMs, using a single root I/O virtualization (SRIOV). This method can expose multiple virtual functions (VFs), which can be instantiated different VMs to form the logical drives. An exemplary system based on the SRIOV method is described below in relation to FIG. 1.

However, this method can result in some constraints, e.g.: data and I/O from different VMs may be stored in the same NAND block or page, which can result in a time-consuming process for data destruction, and can also trigger garbage collection; I/Os from multiple VMs may be placed in a random layout across the physical storage drives, which can create difficulties in balancing the I/O performance among the multiple VMs; the I/O distribution may be spread randomly across the multiple physical storage drives, which can result in hot spots and a traffic imbalance; and a single storage drive may not provide data recovery protection among the multiple physical storage drives. Thus, while the SRIOV method can provide logical drives for multiple VMs, the above-described constraints can result in a decrease in the efficiency and performance of the overall storage system.

The embodiments described herein address these issues by providing a system which divides the physical storage capacity of a non-volatile storage drive into block groups which include block columns, where each block column corresponds to a block from a die of the storage drive. For each virtual function (VF) associated with a host (e.g., an incoming host write instruction), the system can allocate a number of block columns, where the allocated number is based on requirements of a respective VF or its associated virtual machine (VM), as described below in relation to FIG. 2.

Similar to this allocation of block columns in processing an incoming host write instruction, the system can also allocate block columns to an internal background write instruction, as described below in relation to FIGS. 2 and 3. This allows the data access to remain independent in the allocated block columns of a given physical storage space, while also allowing the background operations to occur within its allocated block columns of another given physical storage space.

Moreover, the division and organization of the physical storage media into the block columns can result in a more efficient destruction of data, which can accelerate the readiness of the physical storage media to provide service to other VMs, as described below in relation to FIGS. 3 and 4. The block columns allocated to a VF can be associated with dies in a single storage device or across multiple storage devices, as described below in relation to FIG. 5. In addition, the organization of the physical storage media can provide a flexible erasure coding process, which can result in improvement in the agility and flexibility of data protection in handling issues relating to random defect patterns.

Thus, by dividing and organizing the physical storage media into block columns which can be allocated to various host applications or VFs, and by allocating block columns to a host write instruction and an internal background write instruction, the described system can provide physically isolated storage spaces which can facilitate a more efficient multi-tenancy virtualization.

A “storage system infrastructure,” “storage infrastructure,” or “storage system” refers to the overall set of hardware and software components used to facilitate storage for a system. A storage system can include multiple clusters of storage servers and other servers. A “storage server” refers to a computing device which can include multiple storage devices or storage drives. A “storage device” or a “storage drive” refers to a device or a drive with a non-volatile memory which can provide persistent storage of data, e.g., a solid state drive (SSD), a hard disk drive (HDD), or a flash-based storage device.

A “non-volatile storage device” refers to a computing device, entity, server, unit, or component which can store data in a persistent or a non-volatile memory. In the embodiments described herein, the non-volatile storage device is depicted as a solid state drive (SSD), which includes a plurality of dies which can be accessed over a plurality of channels, but other non-volatile storage devices can be used.

A “computing device” refers to any server, device, node, entity, drive, or any other entity which can provide any computing capabilities.

A physical storage capacity of a non-volatile storage device can be divided or organized into “block groups.” A block group can include a plurality of “block columns.” A block column can correspond to a block from each of a plurality of dies of the non-volatile storage device.

“Allocating” block columns to a function can also be referred to as “assigning,” “mapping,” or “associating” block columns to a function.

A “sealed block column” refers to a block column which is filled with data and in a state which is ready to be recycled. An “open block column” refers to a block column which includes pages which are available to be written to or programmed. An open block column can be associated with a host write instruction (e.g., block column 432 in FIG. 4) or with an internal background write instruction (e.g., block column 422 in FIG. 4).

A “virtual machine” or “VM” can be associated with a host, and a VM can instantiate a corresponding “virtual function” or “VF.” The embodiments described herein refer to allocating block columns to a function associated with a host, and can also refer to allocating block columns to a virtual machine corresponding to the function, which can be a VF. Some references to a VF or VM may be described as “VF/VM” or “VM/VF.”

Architecture of Exemplary Virtualization in a System in the Prior Art

FIG. 1 illustrates an architecture of virtualization in an exemplary system 100, in accordance with the prior art. System 100 can depict the single root I/O virtualization (SRIOV), which can expose multiple virtual functions (VFs). The VFs can be instantiated by different virtual machines (VMs) to form the logical drives. For example, system 100 can include multiple VMs, such as VMs 102, 104, and 106. Each VM can be associated with a VF: VM 102 can be associated with a VF 122; VM 104 can be associated with a VF 124; and VM 106 can be associated with a VF 126. System 100 can thus include logical drives, which can include a physical function (PF 120) and VFs 122, 124, and 126. System 100 can receive incoming I/O requests or data from VMs 102-106 (and via associated VFs 122-126), which requests/data can be processed by queue pairs 130. Queue pairs 130 can include submission queues (SQs) and completion queues (CQs). System 100 can input the I/O requests/data into a round robin module 140, which can subsequently transmit the I/O requests/data to non-volatile memory express (NVMe) SSDs 150 (via a communication 160). Based on the logical block addresses (LBAs) of the data, NVMe SSDs 150 may mix the data from different VMs together in a random manner.

The method depicted in conventional system 100 can result in some constraints. First, data and I/O from different VMs may be stored in the same NAND block or even NAND page, which can result in a time-consuming process for data destruction, and can also trigger garbage collection. The garbage collection is an internal background write operation which can result in interference with incoming write I/Os. Second, I/Os from multiple VMs may be placed in a random layout across the physical storage drives, which can create difficulties in balancing the I/O performance among the multiple VMs. Third, the distribution of the I/O may not be managed by one physical server for the multiple NVMe SSDs 150, and the usage of physical capacity on neighboring storage drives may differ greatly. Because the I/O distribution may be spread randomly across the multiple physical storage drives, this can result in hot spots and a traffic imbalance. Fourth, a single storage drive may not provide sufficient data recovery protection among the multiple physical storage drives.

Thus, while the SRIOV method can provide logical drives for multiple VMs, the above-described constraints can result in a decrease in the efficiency and performance of the overall storage system.

Physically Isolated Storage Units: Block Groups with Block Columns Across Multiple Dies

FIG. 2 illustrates an exemplary environment 200 with physically isolated storage space based on the storage capacity divided into block groups which include block columns, in accordance with an embodiment of the present application. Environment 200 can include multiple channels, e.g., channels 1 210, 2 217, N−1 223, and N 229. Each channel can be associated with one or more NAND dies, and each NAND die can be accessed via a channel. For example: NAND dies 212 and 242 can be accessed via channel 1 210; NAND dies 218 and 248 can be accessed via channel 2 217; NAND dies 224 and 254 can be accessed via channel N−1 223; and NAND dies 230 and 260 can be accessed via channel N 229. In some embodiments, each channel can correspond to a single NAND die.

The system can divide or organize the physical space of the depicted NAND dies (i.e., 212, 218, 224, 230, 242, 248, 254, and 260) into a plurality of block groups, where a block group can include a plurality of block columns and where each block column corresponds to a block from each of a plurality of dies. The division or organization of the physical storage space of the storage media can be depicted by a communication 268. For example, a block group a 270 can include a block column 1 272 and a block column k 274. Block column 1 272 can correspond to the following blocks: block group a (“Ga”), block 1 (i.e., “Ga1”) 214 of NAND die 212; Ga2 220 of NAND die 218; GaN−1 226 of NAND die 224; and GaN 232 of NAND die 230. Similarly, block column k 274 can corresponding to the following blocks: block group z (“Gz”), block 1 (i.e., “Gz1”) 216 of NAND die 212; Gz2 222 of NAND die 218; GzN−1 228 of NAND die 224; and GzN 234 of NAND die 230. Thus, each of the 1 through k block columns of block group a 270 can include N number of blocks from N number of NAND dies. Furthermore, the system can divide or organize the storage space of the physical storage media into a plurality of block groups, e.g., block group a 270 through block group z 280. Similar to block group a 270, block group z 280 can include 1 through k number of block columns, where each block column correspond to blocks from NAND dies 1-N.

Each block column can be considered a physical storage space, and the system can allocate a certain number of block columns from the same or different block groups to a certain host, application, virtual function, or virtual machine associated with a host, e.g., in handling an incoming host I/O or write instruction. The depicted block groups with the corresponding block columns can thus form a physically isolated storage space. Any number of block columns from any number of block groups can also form a physical isolated storage space. For example, the NAND blocks from Ga1 to GaN from N different NAND dies on the N channels can form block column 1 272 of block group a 270. The system may allocate each block column only to one VF/VM, and can also allocate multiple block columns to one VF/VM. The system can allocate multiple block columns for multiple VFs/VMs in parallel. Moreover, the system can allocate block columns from the same block group or from different block groups. That is, the allocated block columns may or may not be associated with the same block group.

While processing the incoming host I/O or write instruction, the system can also perform background write instructions by using the physically isolated storage spaces. For example, in a communication 288, the system can “pair” two different types of block columns. The system can allocate a host I/O instruction (and its corresponding VF/VM) to a column i 292, which corresponds to blocks 292.1 to 292.N. The system can also allocate a garbage collection (“GC”) process 293 (i.e., an internal background write instruction) to a column i+1 294. While environment 200 depicts the allocated pair of block columns as neighboring or sequential (i and i+1) and in a 1-to-1 ratio, the system can allocate any two block columns to these two different processes, and the ratio of allocated blocks can be any ratio other than 1-to-1. Different scenarios of application and system usage may result in various combinations (i.e., ratios) in the allocation of host I/O block columns (e.g., as allocated to host I/O 291) and background write block columns (e.g., as allocated to GC 293).

Exemplary Environment for Virtualization and Block Column Allocation

FIG. 3 illustrates an exemplary environment 300 for virtualization and allocation of block columns from a block column pool, in accordance with an embodiment of the present application. Environment 300 can include a plurality of virtual machines (VMs) 312, 322, and 342, which correspond, respectively, to virtual functions (VFs) 304, 324, and 344. As described above in relation to FIG. 2, the system can allocate a physical storage space to each VF, where the physical storage space is associated with a number of block columns of the physical storage capacity of a non-volatile storage device. Thus, one physical storage space can include a plurality of block columns and can serve only one VM via a corresponding VF.

The system can allocate block columns to each VF based on various conditions, including but not limited to: a size of incoming data associated with a VF/VM; a demand of the VM; a requested capacity, bandwidth, storage capacity, or other factor associated with and received from the VM; any performance, latency, or bandwidth requirements associated with a VF/VM; a historical, current, or predicted amount of physical storage space required by a VF/VM; and any other factor which can affect the amount of physical storage space required by a VF/VM. Each physical storage space can thus be a flexible domain with a variable number of block columns based on these exemplary factors and the utilization of a given VM. The system can dynamically allocate block columns to a VM/VF based on these factors, and can also adjust the allocated number of block columns based on these factors and other real-time factors or conditions detected, determined, or monitored by the system.

When the system receives incoming new data to be stored in the non-volatile storage device, or when certain background operations result in moving data to a new block column, the system can allocate a free block column (e.g., with all data erased) from a block column pool 390. The system can allocate the block columns to each VF from block column pool 390, as shown by block columns 392 and 394. For example: the system can allocate, from block column pool 390, block columns 308, 310, 312, and 314 to obtain a physical storage space 306 for VF 304; the system can allocate, from block column 390, block columns 328, 330, 332, 334, 336, and 338 to obtain a physical storage space 326 for VF 324; and the system can allocate, from block column pool 390, block columns 348 and 350 to obtain a physical storage space 346 for VF 344.

Furthermore, the system can perform a garbage collection process 362 on an identified block column, and upon recycling and erasing the data in the blocks corresponding to the identified block column, the system can return the recycled block column back to block column pool 390 for future use. Similarly, the system can perform a namespace destruction 364, which can involve removing a VM and deleting all data associated with the VM. Upon deleting the data in the blocks of all the block columns associated with the given namespace (i.e., all data associated with a given VM that is stored in the allocated block columns of a physical storage space for the given VM), the system can also return those block columns back to block column pool 390 for future use. Because each physical storage space contains only data corresponding to its given VM (and does not contain any data corresponding to other VMs), all of the data associated with a given VM is stored in its respective physical storage space, which eliminates the need to implement a garbage collection process on the respective physical storage space or on any other physical storage spaces. This is an improvement over the conventional system, in which overprovisioned space can result in more complicated data erasure procedures due to the data being scattered in various locations in the physical storage media (as depicted above in relation to the prior art environment of FIG. 1).

Exemplary Block Columns while Processing Incoming Host Write and Internal Background Process

As described above in relation to FIG. 3, all the data associated with a given VM is stored in its corresponding physically isolated storage space, which includes block columns corresponding to blocks from a plurality of dies via a plurality of channels. This can result in a decreased interference among the multiple VMs when handling I/O access. Indeed, in a given physical storage space, the system can process and perform both an incoming host write instruction and an internal background write instruction (such as a garbage collection procedure) in parallel, or at the same or a similar time, which can reduce the interference caused by an internal background write in the conventional system, as described above in relation to FIG. 1.

FIG. 4 illustrates three exemplary block columns, including a sealed block column 412 for a host write, an open block column 432 for a host write, and an open block column 422 for a garbage collection process, in accordance with an embodiment of the present application. Block column 412 can include blocks 414, 416, and 418, where each block can include a mixture of valid and invalid data pages. Block column 412 can be a sealed block column which includes data programmed in all pages via a host write 410, where block column 412 is subsequently marked as “sealed.” Block column 422 can include blocks 424, 426, and 428, where each block can include available pages (indicated with a blank cell) to which data can be written, as part of an internal background write instruction, e.g., a garbage collection write. Block column 422 can be an open block column which includes pages which can be programmed via a GC write 420. Block column 432 can include blocks 434, 436, and 438, where each block can include both programmed data (marked either as valid or invalid) and available pages (indicated with a blank cell). Block column 432 can be an open block column which includes pages available to be programmed via a host write 430. Sealed block column 412 associated with a host write instruction and open block column 432 associated with a host write instruction can both correspond to column i 292 for host I/O 291 as depicted in FIG. 2. Open block column 432 associated with a garbage collection write instruction can correspond to column i+1 294 for garbage collection 293.

During operation, the system can identify block column 412 with all pages programmed as a block column to be marked as a sealed block column, and the system can seal identified block column 412. The system can execute an internal background write instruction (e.g., a garbage collection process) based on block column 422 (associated with GC write 420). The system can copy valid data from blocks 414-418 of sealed block column 412 to available blocks 424-428 of open block column 422 (via, e.g., communications 442, 444, and 446). The system can erase the data stored in blocks 414-418 of sealed block column 412, and can return sealed block column 412 back to a block column pool (not shown in FIG. 4; as depicted by block column pool 390 of FIG. 3). At the same time, the system can receive and place incoming host data into open block column 432 (associated with host write 430).

In addition, the system may seal an open block upon detecting or based upon a certain condition, e.g., detecting that no data has been programmed or written to an open block within a predetermined time interval or period, and determining that the amount of data stored in the blocks of a block column is greater than a predetermined threshold.

Mapping of Virtual Functions to Block Columns Across Storage Devices

When the capacity of a single storage drive is divided into several physically isolated storage spaces (as described above), one challenge may be the upper limit of a physical capacity which may be obtained by a single VM. One solution for overcoming such a fixed upper limit is to extend the physical storage space associated with or allocated to a VM by distributing the physical storage spaces across multiple storage devices.

FIG. 5 illustrates an exemplary storage controller 502 with a global flash translation layer module 512 which maps virtual functions 1 506, 2 508, and N 510 to block columns from at least two storage devices, such as SSDs 520, 530, and 540, in accordance with an embodiment of the present application. Storage controller 502 can also include a physical function (PF) 504 which may be associated with VFs 1 506-N 510. Each storage device can include a block column pool and a plurality of allocated physical storages spaces which each correspond to a VF. For example, SSD 520 can include a block column pool 528, from which the system can allocate block columns to various physical storage spaces for a respective VF, such as a physical storage space (for VF 1) 522, a physical storage space (for VF 2) 524, and a physical storage space (for VF N) 526. Similarly, SSD 530 can include a block column pool 538, from which the system can allocate block columns to various physical storage spaces for a respective VF, such as a physical storage space (for VF 1) 532, a physical storage space (for VF 2) 534, and a physical storage space (for VF N) 536. Additionally, SSD 540 can include a block column pool 548, from which the system can allocate block columns to various physical storage spaces for a respective VF, such as a physical storage space (for VF 1) 542, a physical storage space (for VF 2) 544, and a physical storage space (for VF N) 546. Thus, the physical storage space allocated to each VF can include physical storage space across multiple SSDs. For example, using global flash translation layer 512, the system can allocate, for VF 1, the following three physical storage spaces across three different SSDs: physical storage space 522 in SSD 520; physical storage space 532 in SSD 530; and physical storage space 542 in SSD 540. Global flash translation layer 512 can store a mapping of the plurality of allocated physical storage spaces to a respective virtual function (e.g., a mapping of physical storage spaces 522, 532, and 542 to VF 1 506).

Furthermore, each SSD can implement its own block column pool, and global flash translation layer 512 can collectively manage all of block column pools 528, 538, and 548 from, respectively, each of SSDs 520, 530, and 540. In some embodiments, global flash translation layer 512 can implement the block column pools for each SSD.

Exemplary Erasure Code Encoding

The system can provide a further improvement to the efficiency of the utilization of the physical storage capacity and the reliability of the data by protecting the blocks of data in a physical storage space with erasure coding. If a single storage drive is defective, the number of NAND blocks in that single storage drive should be less than the recovery strength of the erasure keyword. In other words, the constraint is that the number of NAND blocks in a single SSD which belong to a single EC group is less than the maximum recovery capability of the EC codeword. Otherwise, if that single SSD fails, the system cannot recover the data on that single SSD.

For example, given one (n,k) erasure coding scheme, the system can maximally tolerate the failure of n-k blocks. That is, the number of NAND blocks in a single storage drive which belong to the same allocated physical storage space must be less than n-k. If the entire storage drive fails, the system can still perform an EC-based recovery of the data. This can result in a more powerful and flexible protection for data recovery, as compared to the traditional or conventional RAID process. In the embodiments described herein, each EC group can allow the maximum n-k defective blocks at random locations within the respective EC group. This can result in an extension of the capacity, and can further spread the I/O access across the multiple drives while constructing a powerful EC protection, as described below in relation to FIG. 6.

FIG. 6 illustrates an exemplary environment 600 for erasure code (EC) encoding in a physical storage space, in accordance with an embodiment of the present application. Environment 600 can include a storage controller 602 with a global flash translation layer module (not shown) which maps virtual functions 1 606, 2 608, and N 610 to block columns from at least two storage devices, such as SSDs 520, 530, and 540, in accordance with an embodiment of the present application. Storage controller 602 can also include a physical function (PF) 604 which may be associated with VFs 1 506-N 510. As in FIG. 5, each storage device can include a block column pool and a plurality of allocated physical storages spaces which correspond to a VF.

Storage controller 602 can also include an erasure coding (EC) encoder module 612 and an erasure coding (EC) decoder module 614. During operation, the system can allocate block columns from at least two of SSDs 520-540 to each VF. EC encoder 612 can perform EC encoding on data received from or associated with each VF, e.g., VF 1 606. The system can write the EC-encoded data to block columns which belong to the physical storage spaces allocated to VF 1 (i.e., physical storage spaces 522, 532, and 542), which results in the EC-encoded data spread out in multiple groups across the SSDs. That is, the system can write the EC-encoded data as EC groups across the SSDs. For example, an EC group 1 622 can be indicated with right-slanting diagonal lines, and can be stored across SSDs 520, 530, and 540. Similarly, an EC group 2 624 can be indicated with vertical lines, and can be stored similarly in SSDs 520, 530, and 540. Finally, an EC group 3 626 can be indicated with a cross-hatch pattern, and can be stored similarly in SSDs 520, 530, and 540. Thus, distributing the EC groups across the plurality of SSDs in the corresponding physical storage spaces allocated to a certain VF can result in an improvement in capacity extension and data protection.

Exemplary Method for Facilitating Organization of Data

FIG. 7A presents a flowchart 700 illustrating a method for facilitating organization of data, in accordance with an embodiment of the present application. During operation, the system divides a physical storage capacity of a non-volatile storage device into a plurality of block groups, wherein a block group comprises a plurality of block columns, and wherein a block column corresponds to a block from each of a plurality of dies of the non-volatile storage device (operation 702). The system allocates, to a function associated with a host, a number of block columns to obtain a physical storage space for the function (operation 704). In some embodiments, the system can pre-configure the physical storage capacity of the non-volatile storage device by dividing the physical storage capacity into the block groups. That is, allocating the number of block columns to the host-associated function may or may not be immediately responsive to (i.e., may or may not occur immediately after) the configuration of the physical storage media of the non-volatile storage device. In response to processing an incoming host write instruction and an internal background write instruction, the system allocates a first block column to the incoming host write instruction and a second block column to the internal background write instruction, thereby extending a lifespan of the non-volatile storage device by recycling the first block column when deleting a corresponding namespace for the function (operation 706), and the operation returns.

FIG. 7B presents a flowchart 720 illustrating a method for facilitating organization of data, in accordance with an embodiment of the present application. During operation, the system divides a physical storage capacity of a non-volatile storage device into a plurality of block groups, wherein a block group comprises a plurality of block columns, and wherein a block column corresponds to a block from each of a plurality of dies of the non-volatile storage device (operation 722). The system allocates, to a function associated with a host, a number of block columns to obtain a physical storage space for the function (operation 724). The system processes an incoming host write instruction and an internal background write instruction (operation 726).

In processing the incoming host write instruction, the system allocates a first block column to the incoming host write instruction (operation 732). In response to receiving the incoming host write instruction, the system writes data associated with the host write instruction to at least the first block column allocated to the function (operation 734). In processing the internal background write instruction, in response to receiving the internal background write operation, the system identifies a sealed block column which is filled with data (operation 742). The system allocates a second block column to the internal background write instruction (operation 744). The system executes the internal background write instruction as a garbage collection process based on the second block column (operation 746). To execute the garbage collection process, the system can copy valid data from blocks of the sealed block column to blocks of the second block column, erase data stored in the blocks of the sealed block column, and return the sealed block column to the block column pool (not shown).

FIG. 7C presents a flowchart 750 illustrating a method for facilitating organization of data, in accordance with an embodiment of the present application. The function is a virtual function, and during operation, the system receives a command to delete a namespace or a virtual machine associated with the virtual function (operation 752). The system can erase the number of block columns of the physical storage space allocated for the virtual function (operation 754), and return the number of block columns to the block column pool (operation 756).

Exemplary Computer System and Apparatus

FIG. 8 illustrates an exemplary computer system 800 that facilitates organization of data, in accordance with an embodiment of the present application. Computer system 800 includes a processor 802, a volatile memory 806, and a storage device 808. In some embodiments, computer system 800 can include a controller 804 (indicated by the dashed lines). Volatile memory 806 can include, e.g., random access memory (RAM), that serves as a managed memory, and can be used to store one or more memory pools. Storage device 808 can include persistent storage which can be managed or accessed via processor 802 (or controller 804). Furthermore, computer system 800 can be coupled to peripheral input/output (I/O) user devices 810, e.g., a display device 811, a keyboard 812, and a pointing device 814. Storage device 808 can store an operating system 816, a content-processing system 818, and data 834.

Content-processing system 818 can include instructions, which when executed by computer system 800, can cause computer system 800 or processor 802 to perform methods and/or processes described in this disclosure. Specifically, content-processing system 818 can include instructions for receiving and transmitting data packets, including data to be read or written, an input/output (I/O) request (e.g., a read request or a write request), metadata, a logical block address (LBA), a physical block address (PBA), and an indicator of a VF, a VM, a block group, a block column, or a block (communication module 820).

Content-processing system 818 can include instructions for dividing a physical storage capacity of a non-volatile storage device into a plurality of block groups, wherein a block group comprises a plurality of block columns, and wherein a block column corresponds to a block from each of a plurality of dies of the non-volatile storage device (physical capacity-dividing module 822). Content-processing system 818 can include instructions for allocating, to a function associated with a host, a number of block columns to obtain a physical storage space for the function (block column-allocating module 824). Content-processing system 818 can include instructions for, in response to processing an incoming host write instruction and an internal background write instruction (host write-processing module 826 and background write-processing module 828), allocating a first block column to the incoming host write instruction and a second block column to the internal background write instruction (block column-allocating module 824).

Content-processing system 818 can include instructions for, in response to receiving a command to delete a namespace or virtual machine associated with the virtual function (block column-recycling module 830 and block column-erasing module 832): erasing the number of block columns of the physical storage space allocated for the virtual function (block column-erasing module 832); and returning the number of block columns to the block column pool (block column-allocating module 824).

Data 834 can include any data that is required as input or generated as output by the methods and/or processes described in this disclosure. Specifically, data 834 can store at least: data; a request; a logical block address (LBA); a physical block address (PBA); a mapping between a virtual machine (VM) and a virtual function (VF); a mapping between a VM and one or more physical storage spaces; an indicator of a physical storage space which includes a number of block columns; an indicator of a block group(s) or block column(s) which have been allocated to a given VM or corresponding VF; a write instruction; an incoming host write instruction; an internal background write instruction; a lifespan of a non-volatile storage device; a block column; a block of data; a page of data; an indicator of whether data is valid or invalid; an indicator of whether a block column is sealed or open, and whether the block column is associated with or assigned to a host write or a background write; a namespace corresponding to a function or a virtual function; a command to delete a VM associated with a VF; a block column pool; an indicator of a global flash translation layer; a global block column pool; a block column pool associated with or implemented by a single storage device; erasure code (EC) encoded or decoded data; an EC codeword; and a distributed EC codeword.

FIG. 9 illustrates an exemplary apparatus 900 that facilitates organization of data, in accordance with an embodiment of the present application. Apparatus 900 can comprise a plurality of units or apparatuses which may communicate with one another via a wired, wireless, quantum light, or electrical communication channel. Apparatus 900 may be realized using one or more integrated circuits, and may include fewer or more units or apparatuses than those shown in FIG. 9. Furthermore, apparatus 900 may be integrated in a computer system, or realized as a separate device or devices capable of communicating with other computer systems and/or devices.

Apparatus 900 can comprise modules or units 902-914 which are configured to perform functions or operations similar to modules 820-832 of computer system 800 of FIG. 8, including: a communication unit 902; a physical capacity-dividing unit 904; a block column-allocating unit 906; a host write-processing unit 908; a background write-processing unit 910; a block column-recycling unit 912; and a block column-erasing unit 914.

The data structures and code described in this detailed description are typically stored on a computer-readable storage medium, which may be any device or medium that can store code and/or data for use by a computer system. The computer-readable storage medium includes, but is not limited to, volatile memory, non-volatile memory, magnetic and optical storage devices such as disk drives, magnetic tape, CDs (compact discs), DVDs (digital versatile discs or digital video discs), or other media capable of storing computer-readable media now known or later developed.

The methods and processes described in the detailed description section can be embodied as code and/or data, which can be stored in a computer-readable storage medium as described above. When a computer system reads and executes the code and/or data stored on the computer-readable storage medium, the computer system performs the methods and processes embodied as data structures and code and stored within the computer-readable storage medium.

Furthermore, the methods and processes described above can be included in hardware modules. For example, the hardware modules can include, but are not limited to, application-specific integrated circuit (ASIC) chips, field-programmable gate arrays (FPGAs), and other programmable-logic devices now known or later developed. When the hardware modules are activated, the hardware modules perform the methods and processes included within the hardware modules.

The foregoing embodiments described herein have been presented for purposes of illustration and description only. They are not intended to be exhaustive or to limit the embodiments described herein to the forms disclosed. Accordingly, many modifications and variations will be apparent to practitioners skilled in the art. Additionally, the above disclosure is not intended to limit the embodiments described herein. The scope of the embodiments described herein is defined by the appended claims.

Claims

1. A computer-implemented method, comprising: allocating, to a function associated with a host, a number of block columns to obtain a physical storage space for the function, wherein a block column corresponds to a block from each of a plurality of dies of a non-volatile storage device; andin response to processing an incoming host write instruction and an internal background write instruction, allocating a first block column to the incoming host write instruction and a second block column to the internal background write instruction.

2. The method of claim 1, wherein the function is a virtual function, and wherein the method further comprises: in response to receiving a command to delete a namespace or a virtual machine associated with the virtual function: erasing the number of block columns of the physical storage space allocated for the virtual function; andreturning the number of block columns to a block column pool.

3. The method of claim 1, wherein allocating the number of block columns comprises: obtaining the number of block columns from a block column pool.

4. The method of claim 1, further comprising: in response to receiving the incoming host write instruction, writing data associated with the host write to at least the first block column allocated to the function.

5. The method of claim 1, further comprising: identifying a sealed block column which is filled with data; andexecuting the internal background write instruction as a garbage collection process based on the second block column, by: copying valid data from blocks of the sealed block column to blocks of the second block column;erasing data stored in the blocks of the sealed block column; andreturning the sealed block column to a block column pool.

6. The method of claim 1, wherein the non-volatile storage device is one of a plurality of non-volatile storage devices which communicate with a storage controller which includes a global flash translation layer,wherein the global flash translation layer allocates the number of block columns to the function,wherein the allocated block columns correspond to at least two of the non-volatile storage devices, andwherein the function is one of a plurality of virtual functions to which the global flash translation layer allocates block columns.

7. The method of claim 6, wherein the global flash translation layer maps each virtual function to an allocated physical storage space, andwherein each physical storage space includes block columns corresponding to the at least two of the non-volatile storage devices.

8. The method of claim 6, wherein the storage controller further includes an erasure coding (EC) encoder module which performs EC encoding for the functions and an EC decoder module which performs EC decoding for the functions,wherein data associated with the function is stored in the allocated number of block columns across the at least two non-volatile storage devices, and wherein the method further comprises: performing, by the EC encoder module, EC encoding on the data prior to the data being stored in the allocated number of block columns to obtain an EC codeword;distributing the EC codeword to be stored in block columns in the allocated number of block columns across the at least two non-volatile storage devices.

9. The method of claim 1, further comprising: dividing a physical storage capacity of the non-volatile storage device into a plurality of block groups, wherein a block group comprises a plurality of block columns.

10. A computer system, comprising: a processor; anda non-transitory memory coupled to the processor and storing instructions which, when executed by the processor, cause the processor to perform a method, the method comprising: allocating, to a function associated with a host, a number of block columns to obtain a physical storage space for the function, wherein a block column corresponds to a block from each of a plurality of dies of a non-volatile storage device; andin response to processing an incoming host write instruction and an internal background write instruction, allocating a first block column to the incoming host write instruction and a second block column to the internal background write instruction.

11. The computer system of claim 10, wherein the function is a virtual function, and wherein the method further comprises: in response to receiving a command to delete a namespace or a virtual machine associated with the virtual function: erasing the number of block columns of the physical storage space allocated for the virtual function; andreturning the number of block columns to a block column pool.

12. The computer system of claim 10, wherein allocating the number of block columns comprises: obtaining the number of block columns from a block column pool.

13. The computer system of claim 10, wherein the method further comprises: in response to receiving the incoming host write instruction, writing data associated with the host write to at least the first block column allocated to the function.

14. The computer system of claim 10, wherein the method further comprises: identifying a sealed block column which is filled with data; andexecuting the internal background write instruction as a garbage collection process based on the second block column, by: copying valid data from blocks of the sealed block column to blocks of the second block column;erasing data stored in the blocks of the sealed block column; andreturning the sealed block column to a block column pool.

15. The computer system of claim 10, wherein the non-volatile storage device is one of a plurality of non-volatile storage devices which communicate with a storage controller which includes a global flash translation layer,wherein the global flash translation layer allocates the number of block columns to the function,wherein the allocated number of block columns correspond to at least two of the non-volatile storage devices, andwherein the function is one of a plurality of virtual functions to which the global flash translation layer allocates block columns.

16. The computer system of claim 15, wherein the global flash translation layer maps each virtual function to an allocated physical storage space, andwherein each physical storage space includes block columns corresponding to the at least two of the non-volatile storage devices.

17. The computer system of claim 15, wherein the storage controller further includes an erasure coding (EC) encoder module which performs EC encoding for the functions and an EC decoder module which performs EC decoding for the functions,wherein data associated with the function is stored in the allocated number of block columns across the at least two non-volatile storage devices, and wherein the method further comprises: performing, by the EC encoder module, EC encoding on the data prior to the data being stored in the allocated number of block columns to obtain an EC codeword;distributing the EC codeword to be stored in block columns in the allocated number of block columns across the at least two non-volatile storage devices.

18. The computer system of claim 10, wherein the method further comprises: dividing a physical storage capacity of the non-volatile storage device into a plurality of block groups, wherein a block group comprises a plurality of block columns.

19. An apparatus, comprising: a block column-allocating unit configured to allocate, to a function associated with a host, a number of block columns to obtain a physical storage space for the function, wherein a block column corresponds to a block from each of a plurality of dies of a non-volatile storage device;a host write-processing unit configured to process an incoming host write instruction; anda background write-processing unit configured to process an internal background write instruction;wherein the block column-allocating unit is further configured to, in response to the host write-processing unit processing the incoming host write instruction and the background write-processing unit processing the internal background write instruction, allocate a first block column to the incoming host write instruction and a second block column to the internal background write instruction, andwherein the block column-allocating unit, the host write-processing unit, and the background write-processing unit perform their functions based on instructions stored on a non-transitory computer readable medium executed by a processor.

20. The apparatus of claim 19, wherein the function is a virtual function, wherein the number of block columns are allocated from and obtained from a block column pool, and wherein the apparatus further comprises: a physical capacity-dividing unit configured to divide a physical storage capacity of the non-volatile storage device into a plurality of block groups, wherein a block group comprises a plurality of block columns;a communication unit configured to receive a command to delete a namespace or a virtual machine associated with the virtual function; anda block column-recycling unit configured to, in response to the communication unit receiving the command to delete the namespace or the virtual machine associated with the virtual function: erase the number of block columns of the physical storage space allocated for the virtual function; andreturn the number of block columns to the block column pool, wherein the non-volatile storage device is one of a plurality of non-volatile storage devices which communicate with a storage controller which includes a global flash translation layer,wherein the global flash translation layer allocates the number of block columns to the function,wherein the allocated block columns correspond to at least two of the non-volatile storage devices,wherein the function is one of a plurality of virtual functions to which the global flash translation layer allocates block columns, andwherein the physical capacity-dividing unit, the communication unit, and the block column-recycling unit perform their functions based on the instructions stored on the non-transitory computer readable medium executed by the processor.