Patent Application
20170242799
The disclosed embodiments relate generally to memory systems, and in particular, to memory-efficient block/object address mapping in a storage device (e.g., comprising one or more flash memory devices).
Semiconductor memory devices, including flash memory, typically utilize memory cells to store data as an electrical value, such as an electrical charge or voltage. A flash memory cell, for example, includes a single transistor with a floating gate that is used to store a charge representative of a data value. Flash memory is a non-volatile data storage device that can be electrically erased and reprogrammed. More generally, non-volatile memory (e.g., flash memory, as well as other types of non-volatile memory implemented using any of a variety of technologies) retains stored information even when not powered, as opposed to volatile memory, which requires power to maintain the stored information. Increases in storage density have been facilitated in various ways, including increasing the density of memory cells on a chip enabled by manufacturing developments, and transitioning from single-level flash memory cells to multi-level flash memory cells, so that two or more bits can be stored by each flash memory cell.
The speed of many computer operations is frequently constrained by the speed and efficiency with which data can be stored and retrieved from data structures associated with a device. Many conventional data structures take a long time to store and retrieve data. However, tiered data structures can be used to dramatically improve the speed and efficiency of data storage. Some tiered data structures enable data searches, data insertions, data deletions, and sequential data access to be performed in logarithmic time. However, further improvements to tiered data structures can further increase the speed, efficiency, and reliability with which data can be stored and retrieved, thereby improving the performance of devices relying on such tiered data structures.
Various embodiments of systems, methods and devices within the scope of the appended claims each have several aspects, no single one of which is solely responsible for the attributes described herein. Without limiting the scope of the appended claims, after considering this disclosure, and particularly after considering the section entitled “Detailed Description” one will understand how the aspects of various embodiments are used to perform memory-efficient mapping of block/object addresses. In one aspect, a method of managing a storage system having one or more storage devices includes a tiered data structure in which each node has a logical ID and entries in the nodes reference other nodes in the tiered data structure using the logical IDs. As a result, when a child node is updated and stored to a new location, but retains its logical ID, its parent node does not need to be updated, because the logical ID in the entry referencing the child node remains unchanged. Further, the storage system uses a secondary mapping table to translate the logical IDs to the corresponding physical locations of the corresponding nodes.
So that the present disclosure can be understood in greater detail, a more particular description may be had by reference to the features of various embodiments, some of which are illustrated in the appended drawings. The appended drawings, however, merely illustrate pertinent features of the present disclosure and are therefore not to be considered limiting, for the description may admit to other effective features.
In accordance with common practice the various features illustrated in the drawings may not be drawn to scale. Accordingly, the dimensions of the various features may be arbitrarily expanded or reduced for clarity. In addition, some of the drawings may not depict all of the components of a given system, method or device. Finally, like reference numerals may be used to denote like features throughout the specification and figures.
Non-volatile memory systems are often used to store a database of information, such as a database in which data objects and their attributes are stored. The database is typically organized and stored as a tiered data structure, for example a tree data structure (e.g., a B-tree) having nodes in which data objects, pointers to the data objects, and attributes (sometimes called metadata) are stored. Tiered data structures can be used to dramatically improve the speed and efficiency of data storage. However, further improvements to tiered data structures can further increase the speed, efficiency, and reliability with which data can be stored and retrieved, thereby improving the performance of devices relying on such tiered data structures. The various embodiments described herein include systems, methods and/or devices used to improve performance of devices relying on tiered data structures by memory-efficient mapping of block/object addresses.
(A1) More specifically, some embodiments include a method of managing a storage system. In some embodiments, the method includes: (1) detecting a first request to perform a read operation of a first data object stored in a storage device of the storage system, wherein the first request includes first key information corresponding to the first data object, (2) locating a first entry for the first key information in a tiered data structure, wherein the first entry includes a logical ID for a leaf node corresponding to the first key information, (3) determining a first physical location of the leaf node based on the logical ID for the leaf node using a secondary mapping table, wherein the secondary mapping table is used to translate logical IDs for leaf nodes to physical locations of leaf node, (4) reading the leaf node using the first physical location to obtain a leaf node map entry, wherein the leaf node map entry includes size of the first data object and a second physical location of the first data object, and (5) reading from the second physical location to obtain the first data object.
(A2) In some embodiments of the method of A1, the tiered data structure includes a plurality of internal nodes and a plurality of leaf nodes.
(A3) In some embodiments of the method of A2, the plurality of internal nodes are cached in a volatile memory cache.
(A4) In some embodiments of the method of A3, locating the first entry for the first key information in the tiered data structure includes obtaining from a respective internal node of the tiered data structure a respective logical ID for another internal node of the tiered data structure that includes the first entry, and searching the volatile memory cache for the another internal node of the tiered data structure using the respective logical ID
(A5) In some embodiments of the method of A3, determining the first physical location of the leaf node includes searching the volatile memory cache for the logical ID for the leaf node, and upon confirming that the volatile memory cache does not have a cached node corresponding to the logical ID for the leaf node, using the secondary mapping table to determine the first physical location of the leaf node based on the logical ID for the leaf node.
(A6) In some embodiments of the method of A2, each leaf node of the plurality of leaf nodes corresponds to at least one data object.
(A7) In some embodiments of the method of A6, a respective leaf node of the plurality of leaf nodes includes metadata for a corresponding data object.
(A8) In some embodiments of the method of any of A1 to A7, the secondary mapping table is stored in a dynamic random-access memory (DRAM).
(A9) In some embodiments of the method of any of A1 to A8, the method further includes: (1) detecting a second request to perform a write operation for a second data object to the storage device of the storage system, wherein the second request includes data to be written for the second data object and second key information corresponding to the second data object; and (2) determining whether a second entry for the second key information is in the tiered data structure. The method also includes, in accordance with a determination that the second entry for the second key information is not in the tiered data structure: (1) allocating space at a third physical location in the storage device for the data to be written for the second data object; (2) writing the data to be written for the second data object to the third physical location; and (3) inserting the second entry for the second key information in the tiered data structure, wherein the second entry includes size of the second data object and the third physical location. The method also includes, in accordance with a determination that the second entry for the second key information is in the tiered data structure: (1) invalidating data previously associated with the second data object; (2) allocating space at the third physical location in the storage device for the data to be written for the second data object; and (3) updating the second entry for the second key information in the tiered data structure to include the size of the second data object and the third physical location.
(A10) In some embodiments of the method of A9, the method further includes, concurrently with writing the data to be written for the second data object to the third physical location, updating a leaf node, in the tiered data structure, corresponding to the second data object to include metadata for the second data object.
(A11) In some embodiments of the method of any of A1 to A10, the method is controlled by a host that includes a client on behalf of which data is stored in the storage system.
(A12) In some embodiments of the method of any of A1 to A10, the method is controlled by a host that includes a storage system controller of the storage system.
(A13) In some embodiments of the method of any of A1 to A10, the method is controlled by a host that includes a cluster controller of the storage system.
(A14) In some embodiments of the method of any of A1 to A13, the storage device comprises one or more flash memory devices.
(A15) In another aspect, a host system includes an interface for operatively coupling to a storage system, one or more processors, and controller memory (e.g., non-volatile memory or volatile memory in or coupled to the controller) storing one or more programs. The one or more programs including instructions that when executed by the one or more processors cause the host system to perform operations including: (1) detecting a first request to perform a read operation of a first data object stored in a storage device of the storage system, wherein the first request includes first key information corresponding to the first data object, (2) locating a first entry for the first key information in a tiered data structure, wherein the first entry includes a logical ID for a leaf node corresponding to the first key information, (3) determining a first physical location of the leaf node based on the logical ID for the leaf node using a secondary mapping table, wherein the secondary mapping table is used to translate logical IDs for leaf nodes to physical locations of leaf nodes, (4) reading the leaf node using the first physical location to obtain a leaf node map entry, wherein the leaf node map entry includes size of the first data object and a second physical location of the first data object, and (5) reading from the second physical location to obtain the first data object.
(A16) In some embodiments of the host system of A15, the one or more programs include instructions that when executed by the one or more processors cause the host system to perform or control performance of any of the methods A2 to A14 described herein.
(A17) In yet another aspect, any of the methods A1 to A14 described above are performed by a host system including means for performing any of the methods described herein.
(A18) In yet another aspect, a storage system includes a storage medium (e.g., comprising one or more non-volatile storage devices, such as flash memory devices), one or more processors, and memory (e.g., non-volatile memory or volatile memory in the storage system) storing one or more programs, which when executed by the one or more processors cause the storage system to perform or control performance of any of the methods A1 to A14 described herein.
(A19) In yet another aspect, some embodiments include a non-transitory computer readable storage medium, storing one or more programs configured for execution by one or more processors of a storage device, the one or more programs including instructions for performing any of the methods described herein.
(A20) In yet another aspect, a storage system includes one or more storage devices, one or more subsystems having one or more processors, and memory storing one or more programs. The one or more programs including instructions that when executed by the one or more processors cause the storage system to perform operations including: (1) detecting a first request to perform a read operation of a first data object stored in a storage device of the storage system, wherein the first request includes first key information corresponding to the first data object, (2) locating a first entry for the first key information in a tiered data structure, wherein the first entry includes a logical ID for a leaf node corresponding to the first key information, (3) determining a first physical location of the leaf node based on the logical ID for the leaf node using a secondary mapping table, wherein the secondary mapping table is used to translate logical IDs for leaf nodes to physical locations of leaf nodes, (4) reading the leaf node using the first physical location to obtain a leaf node map entry, wherein the leaf node map entry includes size of the first data object and a second physical location of the first data object, and (5) reading from the second physical location to obtain the first data object.
(A21) In some embodiments of the storage system of A20, the one or more programs include instructions that when executed by the one or more processors cause the storage system to perform or control performance of any of the methods A2 to A14 described herein.
Numerous details are described herein in order to provide a thorough understanding of the example implementations illustrated in the accompanying drawings. However, some embodiments may be practiced without many of the specific details, and the scope of the claims is only limited by those features and aspects specifically recited in the claims. Furthermore, well-known methods, components, and circuits have not been described in exhaustive detail so as not to unnecessarily obscure more pertinent aspects of the implementations described herein.
Computer system 110 is coupled to storage controller 124 through data connections 101. However, in some embodiments computer system 110 includes storage controller 124, or a portion of storage controller 124, as a component and/or as a subsystem. For example, in some embodiments, some or all of the functionality of storage controller 124 is implemented by software executed on computer system 110. Computer system 110 may be any suitable computer device, such as a computer, a laptop computer, a tablet device, a netbook, an internet kiosk, a personal digital assistant, a mobile phone, a smart phone, a gaming device, a computer server, or any other computing device. Computer system 110 is sometimes called a host, host system, client, or client system. In some embodiments, computer system 110 is a server system, such as a server system in a data center. In some embodiments, computer system 110 includes one or more processors, one or more types of memory, a display and/or other user interface components such as a keyboard, a touch-screen display, a mouse, a track-pad, a digital camera, and/or any number of supplemental I/O devices to add functionality to computer system 110. In some embodiments, computer system 110 does not have a display and other user interface components.
Storage medium 132 is coupled to storage controller 124 through connections 103. Connections 103 are sometimes called data connections, but typically convey commands in addition to data, and optionally convey metadata, error correction information and/or other information in addition to data values to be stored in storage medium 132 and data values read from storage medium 132. In some embodiments, however, storage controller 124 and storage medium 132 are included in the same device (i.e., an integrated device) as components thereof. Furthermore, in some embodiments, storage controller 124 and storage medium 132 are embedded in a host device (e.g., computer system 110), such as a mobile device, tablet, other computer or computer controlled device, and the methods described herein are performed, at least in part, by the embedded storage controller. Storage medium 132 may include any number (i.e., one or more) of memory devices (e.g., NVM 134-1, NVM 134-2 through NVM 134-n) including, without limitation, non-volatile semiconductor memory devices, such as flash memory device(s). For example, flash memory device(s) can be configured for enterprise storage suitable for applications such as cloud computing, for database applications, primary and/or secondary storage, or for caching data stored (or to be stored) in secondary storage, such as hard disk drives. Additionally and/or alternatively, flash memory device(s) can also be configured for relatively smaller-scale applications such as personal flash drives or hard-disk replacements for personal, laptop, and tablet computers.
Memory devices (NVM 134-1, NVM 134-2, etc.) of storage medium 132 include addressable and individually selectable blocks, such as selectable portion of storage medium 136 (also referred to herein as selected portion 136). In some embodiments, the individually selectable blocks (sometimes called erase blocks) are the minimum size erasable units in a flash memory device. In other words, each block contains the minimum number of memory cells that can be erased simultaneously. Each block is usually further divided into a plurality of pages and/or word lines, where each page or word line is typically an instance of the smallest individually accessible (readable and writable) portion in a block. In some embodiments (e.g., using some types of flash memory), the smallest individually accessible unit of a data set, however, is a sector, which is a subunit of a page. That is, a block includes a plurality of pages, each page contains a plurality of sectors, and each sector is the minimum unit of data for writing and reading data to and from the flash memory device.
In some embodiments, storage controller 124 includes a management module 121, a host interface 129, a storage medium interface 128, and additional module(s) 125. Storage controller 124 may include various additional features that have not been illustrated for the sake of brevity and so as not to obscure pertinent features of the example embodiments disclosed herein, and a different arrangement of features may be possible. Host interface 129 provides an interface to computer system 110 through data connections 101. Similarly, storage medium interface 128 provides an interface to storage medium 132 though connections 103. In some embodiments, storage medium interface 128 includes read and write circuitry, including circuitry capable of providing reading signals to storage medium 132 (e.g., reading threshold voltages for NAND-type flash memory).
In some embodiments, management module 121 includes one or more processing units 122 (also sometimes called processors, hardware processors, CPUs or the like) configured to execute instructions in one or more programs (e.g., in management module 121). In some embodiments, the one or more CPUs 122 are shared by one or more components within, and in some cases, beyond the function of storage controller 124. Management module 121 is coupled to host interface 129, additional module(s) 125 and storage medium interface 128 in order to coordinate the operation of these components. In some embodiments, one or more modules of management module 121 are implemented in computer system 110, as discussed in more detail below.
Additional module(s) 125 are coupled to storage medium interface 128, host interface 129, and management module 121. As an example, additional module(s) 125 may include an error control module to limit the number of uncorrectable errors inadvertently introduced into data during writes to memory or reads from memory. In some embodiments, an error control module, included in additional module(s) 125, includes an encoder and a decoder. In some embodiments, additional module(s) 125 are executed in software by the one or more CPUs 122 of management module 121, and, in other embodiments, additional module(s) 125 are implemented in whole or in part using special purpose circuitry (e.g., to perform encoding and decoding functions). In some embodiments, additional module(s) 125 are implemented in whole or in part by software executed on computer system 110.
In some embodiments, a write operation is initiated when computer system (host) 110 sends one or more host write commands (e.g., via data connections 101, or alternatively a separate control line or bus) to storage controller 124. In response, storage controller 124 sends one or more write access commands to storage medium 132, from storage medium interface 128 (e.g., via data connections 103, or alternatively a separate control line or bus), to write data to physical memory locations (addresses) within storage medium 132.
In some embodiments, during a write operation, host interface 129 receives data to be stored in storage medium 132 from computer system 110. The data received by host interface 129 is made available to an encoder (e.g., in additional module(s) 125), which encodes the data.
In some embodiments, a read operation is initiated when computer system (host) 110 sends one or more host read commands (e.g., via data connections 101, or alternatively a separate control line or bus) to storage controller 124 requesting data from storage medium 132. Storage controller 124 sends one or more read access commands to storage medium 132, from storage medium interface 128 (e.g., via data connections 103, or alternatively a separate control line or bus), to obtain raw read data in accordance with physical memory locations (addresses) within storage medium 132.
In some embodiments, storage medium interface 128 provides the raw read data to a decoder (e.g., in additional module(s) 125). If the decoding is successful, the decoded data is provided to host interface 129, where the decoded data is made available to computer system 110, or is provided to computer system 110 as a response to the host read command. In some embodiments, if the decoding is not successful, storage controller 124 may resort to a number of remedial actions or provide an indication of an irresolvable error condition.
As explained above, a storage medium (e.g., storage medium 132) is divided into a number of addressable and individually selectable blocks and each block is optionally (but typically) further divided into a plurality of pages and/or word lines and/or sectors. While erasure of a storage medium is performed on a block basis, in many embodiments, reading and programming of the storage medium is performed on a smaller subunit of a block (e.g., on a page basis, word line basis, or sector basis). In some embodiments, the smaller subunit of a block consists of multiple memory cells (e.g., single-level cells (SLC) or multi-level cells (MLC)). In some embodiments, programming is performed on an entire page.
As an example, if data is written to a storage medium in pages, but the storage medium is erased in blocks, pages in the storage medium may contain invalid (e.g., stale) data, but those pages cannot be overwritten until the whole block containing those pages is erased. In order to write to the pages with invalid data, the pages with valid data in that block are read and re-written to a new block and the old block is erased (or put on a queue for erasing). This process is called garbage collection. After garbage collection, the new block contains pages with valid data and free pages that are available for new data to be written, and the old block that was erased is also available for new data to be written. Since flash memory can only be programmed and erased a limited number of times, the efficiency of the algorithm used to pick the next block(s) to re-write and erase has a significant impact on the lifetime and reliability of flash-based storage systems.
Write amplification is a phenomenon where the actual amount of physical data written to a storage medium (e.g., storage medium 132) is a multiple of the logical amount of data intended to be written by a host (e.g., computer system 110, sometimes called a host). As discussed above, when a storage medium must be erased before it can be re-written, the garbage collection process to perform these operations results in re-writing data one or more times. This multiplying effect increases the number of writes required over the life of a storage medium, which shortens the time it can reliably operate. The formula to calculate the write amplification of a storage system is given by equation (1):
amount of data written to a storage medium/amount of data written by a host (1)
One of the goals of any storage system architecture is to reduce write amplification as much as possible so that available endurance is used to meet storage medium reliability and warranty specifications. Higher system endurance also results in lower cost as the storage system may need less over-provisioning. By reducing the write amplification, the endurance of the storage medium is increased and the overall cost of the storage system is decreased. Generally, garbage collection is performed on erase blocks with the fewest number of valid pages for best performance and best write amplification. However, since different erase blocks have different wear characteristics, it is important to use erase blocks based on how much life a respective erase block has left, rather than simply the number of program-erase cycles performed on the respective erase block thus far. As described below, in some implementations, garbage collection that is performed based on characteristics of erase blocks (e.g., an age metric) helps to improve wear leveling, thus extending the life of the memory device.
Memory 206-1 includes high-speed random access memory, such as DRAM, SRAM, DDR RAM or other random access solid state memory devices, and may include non-volatile memory, such as one or more magnetic disk storage devices, optical disk storage devices, flash memory devices, or other non-volatile solid state storage devices. Memory 206-1 optionally includes one or more storage devices remotely located from processor(s) 202. Memory 206-1, or alternately the non-volatile memory device(s) within memory 206-1, comprises a non-transitory computer readable storage medium. In some embodiments, memory 206-1, or the computer readable storage medium of memory 206-1 stores the following programs, modules, and data structures, or a subset thereof:
Each of the above identified elements may be stored in one or more of the previously mentioned memory devices, and corresponds to a set of instructions for performing a function described above. The above identified modules or programs (i.e., sets of instructions) need not be implemented as separate software programs, procedures or modules, and thus various subsets of these modules may be combined or otherwise re-arranged in various embodiments. In some embodiments, memory 206-1 may store a subset of the modules and data structures identified above. Furthermore, memory 206-1 may store additional modules and data structures not described above. In some embodiments, the programs, modules, and data structures stored in memory 206-1, or the computer readable storage medium of memory 206-1, provide instructions for implementing respective operations in the methods described below with reference to
Although
Each of the above identified elements may be stored in one or more of the previously mentioned memory devices, and corresponds to a set of instructions for performing a function described above. The above identified modules or programs (i.e., sets of instructions) need not be implemented as separate software programs, procedures or modules, and thus various subsets of these modules may be combined or otherwise re-arranged in various embodiments. In some embodiments, memory 206-2 may store a subset of the modules and data structures identified above. Furthermore, memory 206-2 may store additional modules and data structures not described above. In some embodiments, the programs, modules, and data structures stored in memory 206-2, or the computer readable storage medium of memory 206-2, provide instructions for implementing respective operations in the methods described below with reference to
Although
In some embodiments, non-volatile memory is organized into groups of fixed size segments (e.g., segments 324-1, 324-2 through 324-p). Each segment is further partitioned into a group of fixed size slabs. All the slabs within a particular segment have the same size (e.g., segment 324-1 is a 2 kB slab segment). Optionally, a variety of slab sizes are supported by dividing the full set of segments into groups, with a different slab size for each group. In some embodiments, data object 320 is stored in a slab within a segment. Such a situation is depicted in
In some embodiments, various portions of the tiered data structure 300 are cached in volatile memory (e.g., in DRAM). For example, in one embodiment, all of the internal nodes 304 are cached, but only a portion of the leaf nodes 306 are cached. In another embodiment, all of the internal nodes 304, and none of the leaf nodes 306 are cached. In some embodiments, the portion of the leaf nodes that is cached is determined by how frequently the corresponding data objects are accessed. For example, leaf nodes corresponding to data objects accessed more frequently than other data objects, by a predefined margin, are added to the portion of the leaf nodes that are cached, and leaf nodes corresponding to data objects that are accessed less frequently than other data objects, by a predefined margin, are removed from or not added to the cache. The caching of portions of the tiered data structure can help reduce the number of I/O operations required to perform data access operations, for example by avoiding additional I/O operations to access nodes stored in storage medium 132 in storage device 120.
In some embodiments, each node (i.e., root node 302, internal nodes 304, and leaf nodes 306) of the tiered data structure 300 has a corresponding logical ID, which is used to access the node using a hash table (e.g., secondary mapping table 230-1,
A leaf node 306 may store a variable number of keys and values. Often included amongst these keys and values are data object keys 310, data object pointers 312, attribute keys 314 (e.g., attribute key 314a-314b), and attribute values 316 (e.g., attribute value 316a-316b). Attribute values are sometimes herein called attributes, for ease of discussion. Furthermore, in some embodiments, an attribute, or attribute value, comprises both an attribute identifier (e.g., identifying a type of attribute) and a value (e.g., “color, red,” where “color” is the attribute identifier, and “red” is the value). However, the present discussion is applicable without regard to the exact form or content of such attributes.
Each key/value pair in a leaf node is sometimes herein called an entry or tree entry 308 (e.g., tree entry 308a-308c). In some embodiments, keys are used to uniquely identify an entity, such as a data object or attribute, and thus the key in each tree entry 308 typically has a different value from the key in every other tree entry. Data object pointers 312 point to data objects 320 that may be stored within non-volatile memory (e.g., information in data object pointers 312 includes addresses to physical locations within the non-volatile memory). In some embodiments, attributes include source information, date information, or the like for particular data objects. A data object 320 may have any number of attributes associated with the data object.
Attribute values 316 are typically much smaller than their associated data objects. In some embodiments, when an attribute value meets (e.g., is smaller than) a predefined attribute size threshold, the attribute and its associated key are stored as a tree entry in a leaf node, and otherwise the attribute value is stored in a data node pointed to by a respective tree entry 308 in a leaf node. Similarly, in some embodiments, when a data object meets (e.g., is smaller than) a predefined object size threshold, the object and its key are stored as a tree entry in a leaf node. In
As noted above, an attribute key 314 and its associated attribute value 316 (e.g., attribute one key 314a and attribute one value 316a) are called sometimes called a key/value pair 308. A data object key 310 and its associated data object pointer 312 also constitute a key/value pair. Individual key/value pairs 308 are typically stored contiguously within the leaf node into which they are placed. For example, data object key 310 and data object pointer 312 are stored as tree entry 308a, which occupies a contiguous portion or block 321 (contiguous portion 321a in this example) of a respective leaf node 306 (node 306-2 in this example). Similarly, tree entry 308b is stored in contiguous portion 321b of leaf node 306-3, and tree entry 308c is stored in contiguous portion 321c of leaf node 306-m.
In some embodiments, some of the operations (or alternatively, steps) of method 400a are performed by a storage device (e.g., storage device 120), or one or more components of the storage device (e.g., storage controller 124), that is operatively coupled with the storage device and other operations of method 400a are performed at the host system. In some of these embodiments, the method 400a is governed, at least in part, by instructions that are stored in a non-transitory computer-readable storage medium and that are executed by one or more processors of a device, such as the one or more processing units (CPUs) 122 of management module 121 (
For ease of explanation, the following describes method 400a as performed by the host system (e.g., by CPUs 202 and device driver 216 of computer system 110,
Method 400a begins, in some embodiments, when the host system (e.g., computer system 110,
After initiating 402a the write data operation, the host system (e.g., computer system 110,
In some embodiments, the determination operation 404 is performed by “traversing” the tiered data structure, as follows. The data object is typically identified by a data object key, and that data object key is used to traverse the tiered data structure until either a leaf node is located, or it is determined that no such data object is stored in the tiered data structure. Assuming the data object is stored in the tiered data structure, starting at the root node (302,
In some embodiments, if an entry for the data object already exists within the tiered data structure (i.e., 406—“yes”), the existing version of the requested data object (e.g., data object 320,
In some embodiments, invalidating 408 the existing version of the data object includes deleting the existing version of the data object, or alternatively marking the existing version of the data object as invalid, which eventually results in the existing version of the data object being erased from a data node (e.g., data node 318,
In some embodiments, method 400a continues by storing 410 the requested data object at an allocated physical location in a storage device (e.g., data object 320 is stored in a 2 kB slab of segment 324-1 in storage medium 132,
In some embodiments, a bit vector is associated with each segment, with 1 bit per slab in the segment. If a bit is set, the corresponding slab has been allocated, otherwise it is free. In some embodiments, allocating a slab of a particular size includes: (1) finding a segment that contains slabs of the desired size, (2) searching the bit vector for that segment to find a free slab, (3) setting the bit for the free slab, and (4) returning the location of the free slab. In some embodiments, the slab-based technique includes garbage collection mechanisms to redistribute the number of segments allocated per slab size. For example, allocations of 2 kB slab segments are increased while allocations of 16 kB slab segments are decreased, in response to a distribution of data objects changing to mostly smaller data objects. In some embodiments, the bit vectors are kept in DRAM, and after a crash or shutdown the bit vectors are reconstructed by scanning the tiered data structure to identify all slabs that are in use. In another embodiment, the bit vectors are written to non-volatile memory in a periodic manner, and reloaded into volatile memory during restart. Further, the most recent changes to the bit vectors are kept in a persisted log that is used to reconstruct the bit vectors during restart.
In some embodiments, the requested data object is stored in a non-volatile storage medium (e.g., storage medium 132,
After (or, more generally, in conjunction with) storing 410 the requested data object at an allocated physical location in a storage device, the data object attributes associated with the requested data object are stored 412 in the tiered data structure. In some embodiments, the attributes associated with the requested data object are stored in a non-volatile storage medium (e.g., storage medium 132,
After (or, more generally, in conjunction with) storing 410 the requested data object at an allocated physical location in a storage device (and, optionally, after storing 412 the requested data object attributes in the tiered data structure), a secondary mapping table is updated 414. In some embodiments, the secondary mapping table is updated for each leaf node that is modified by the write data operation and written to a new location in storage medium 132. In particular, the secondary mapping table is updated to map the logical IDs of those leaf nodes to the new locations of the modified leaf nodes.
In some embodiments, the secondary mapping table is stored in volatile memory (e.g., dynamic random-access memory (DRAM)). In some embodiments, the secondary mapping table is continuously written to non-volatile memory in a periodic manner so that there is always a complete but “stale” version of the secondary mapping table. Additionally, the most recent changes to the secondary mapping table are kept in a persisted log that is used to bring the stale copy up-to-date during recovery from a crash or shutdown.
Additional details concerning each of the processing steps for method 400a, as well as details concerning additional processing steps, are presented below with reference to
In some embodiments, some of the operations (or alternatively, steps) of method 400b are performed by a storage device (e.g., storage device 120), or one or more components of the storage device (e.g., storage controller 124), that is operatively coupled with the storage device and other operations of method 400b are performed at the host system. In some of these embodiments, the method 400b is governed, at least in part, by instructions that are stored in a non-transitory computer-readable storage medium and that are executed by one or more processors of a device, such as the one or more processing units (CPUs) 122 of management module 121 (
For ease of explanation, the following describes method 400b as performed by the host system (e.g., by CPUs 202 and device driver 216 of computer system 110,
The method 400b begins, in some embodiments, when the host system (e.g., computer system 110,
After initiating 402b the read data operation, an entry for the requested data object is located 416 in the tiered data structure. In some embodiments, the entry for the requested data object is located in cached portions of the tiered data structure (e.g., cached portions of tiered data structure 238-1,
The method 400b continues, after locating the entry for the requested data object, with determining the physical location of a leaf node corresponding to the requested data object. In some embodiments, the physical location of the leaf node is determined 420 using a secondary mapping table to translate the logical ID of the leaf node to a physical location. In some embodiments, the secondary mapping table is a hash table having an associated hash function that is used to hash the logical ID of the leaf node to locate an entry in the secondary mapping table that contains or identifies the physical location of the leaf node.
After determining 420 the physical location of a leaf node corresponding to the requested data object, the leaf node is read 422 to determine the physical location of the requested data object. In some embodiments, reading the leaf node returns a leaf node map entry which includes the size and physical location of the requested data object. In some embodiments, the leaf node map entry includes the logical ID for the leaf node. In some embodiments, the leaf node map entry includes key information for the data object to enable the leaf node map entry to be found in the leaf node. For example, in some embodiments, if a leaf node contains a plurality of leaf node map entries, the key information for the data object is used to obtain the corresponding leaf node map entry.
The method 400b continues, after the leaf node is read 420 to determine the physical location of the requested data object, with reading 424 the requested data object using the physical location determined from the leaf node. In some embodiments or in some circumstances (e.g., when the requested data object is smaller than a predefined threshold size), the requested data object is read 424 from a leaf node (e.g., leaf node 306-2,
In some embodiments a read data operation requires two I/O operations (e.g., two operations accessing nodes stored in storage medium 132 in storage device 120), while requiring less than 0.1% of volatile memory (cache) relative to non-volatile memory. For example, for 1 TB of non-volatile storage, less than 1 GB of volatile memory is required to cache all but the leaf nodes of the tiered data structure and the secondary mapping table. In that example, the first I/O operation is required to read the required leaf node from the storage device, after determining its physical location using the cached parent nodes and secondary mapping table, and the second I/O operation is required to read the requested data object from the storage device. In some embodiments, for example larger non-volatile storage systems (e.g., 1 PB), the amount of volatile memory needed to store cached parent nodes and the secondary mapping table, which enables performing just two I/O operations per read data operation, is less than 0.03% of the amount of non-volatile memory. Thus for a 1 PB storage system, the amount of volatile memory needed to store cached parent nodes and the secondary mapping table is typically less than 300 GB.
Additional details concerning each of the processing steps for method 400a, as well as details concerning additional processing steps, are presented below with reference to
In some embodiments, some of the operations (or alternatively, steps) of method 500 are performed by a storage device (e.g., storage device 120), or one or more components of the storage device (e.g., storage controller 124), that is operatively coupled with the storage device and other operations of method 500 are performed at the host system. In some of these embodiments, the method 500 is governed, at least in part, by instructions stored in a non-transitory computer-readable storage medium and that are executed by one or more processors of a device, such as the one or more processing units (CPUs) 122 of management module 121 (
For ease of explanation, the following describes method 500 as performed by the host system (e.g., by CPUs 202 and device driver 216 of computer system 110,
With reference to
In some embodiments, after detecting a first request to perform a read operation, the method includes locating 508 a first entry for the first key information in a tiered data structure, wherein the first entry includes a logical ID for a leaf node corresponding to the first key information, as explained above with reference to
In some embodiments, the plurality of internal nodes (e.g., Nodes 302, 304-1-304-L,
In some embodiments, after locating a first entry for the first key information in a tiered data structure, method 500 includes determining 522 a first physical location of the leaf node based on the logical ID for the leaf node using a secondary mapping table, wherein the secondary mapping table is used to translate logical IDs for leaf nodes to physical locations of leaf nodes, as explained above with reference to
In some embodiments, method 500 continues by reading 526 the leaf node using the first physical location to obtain a leaf node map entry, wherein the leaf node map entry includes size of the first data object and a second physical location of the first data object. Further, method 500 includes, after determining the second physical location of the first data object, reading 528 from the second physical location to obtain the first data object.
In some embodiments, managing 502 a storage system having a plurality of storage devices is controlled by a host that includes 530 a client on behalf of which data is stored in the storage system (e.g., data storage system 100,
In some embodiments, the host includes 532 a storage system controller of the storage system. In some embodiments, the storage system controller controls and/or coordinates operations among one or more storage devices (e.g., data storage device 120,
In some embodiments, the host includes 534 a cluster controller of the storage system. In some embodiments, the cluster controller controls and/or coordinates operations among one or more data storage subsystems, where each of the data storage subsystems may be implemented as a data storage system having one or more storage devices (e.g., data storage device 120,
In some embodiments, method 500 includes 536 detecting a second request to perform a write operation for a second data object to the storage device of the storage system, wherein the second request includes data to be written for the second data object and second key information corresponding to the second data object. Next, the method includes determining 538 whether a second entry for the second key information is in the tiered data structure, as explained above with reference to
In some embodiments, in accordance with a determination 540 that the second entry for the second key information is not in the tiered data structure, method 500 further includes: (1) allocating 542 space at a third physical location in the storage device for the data to be written for the second data object; (2) writing 544 the data to be written for the second data object to the third physical location; and (3) inserting 548 the second entry for the second key information in the tiered data structure, wherein the second entry includes size of the second data object and the third physical location.
In some embodiments, in accordance with a determination 550 that the second entry for the second key information is in the tiered data structure, method 500 further includes: (1) invalidating 552 data previously associated with the second data object; (2) allocating 554 space at a third physical location in the storage device for the data to be written for the second data object; (3) writing 556 the data to be written for the second data object to the third physical location; and (4) updating 558 the second entry for the second key information in the tiered data structure to include the size (e.g., a new size) of the second data object and the third physical location. In some embodiments, an error is returned to the host system if the second data object cannot be overwritten. In some embodiments, the invalidated data is garbage collected after the second data object is written to the third physical location.
If an entry for the second key information is not already stored in the tiered data structure (when the write operation is not an overwrite operation), the second entry for the second key information is a new entry. On the other hand, if an entry for the second key information is already stored in the tiered data structure (when the write operation is an overwrite operation), the second entry is modified in accordance with the location (and optionally the size) of the written data.
In some embodiments, writing the data to be written for the second data object to the third physical location includes concurrently updating 546 a leaf node, in the tiered data structure, corresponding to the second data object to include metadata for the second data object.
In the case where the metadata, for example an attribute value, represented by a key/value pair, is small enough to fit in a leaf node, it is possible to store the key/value pair with a single update to the tiered data structure, by storing the key and its value in the leaf node. In the case where the metadata are too large to fit in the leaf node with the leaf node map entry for the data object, the leaf node includes pointers to the data blocks in which the metadata is stored, and thus two I/O operations are required to write the key/value pair, one for writing the associated tree entry and one for writing the data object in which the attribute value is stored.
While the above descriptions focus on the use of logical IDs and a secondary mapping table to make read and write operations efficient in terms of the number of I/O operations required, similar efficiencies are obtained when performing other types of operations, including delete operations, existence checking operations, nameless writes, range queries, enumeration queries, and the like.
With respect to storage medium 132 (
The memory devices can be formed from passive elements, active elements, or both. By way of non-limiting example, passive semiconductor memory elements include ReRAM device elements, which in some embodiments include a resistivity switching storage element, such as an anti-fuse, phase change material, etc., and optionally a steering element, such as a diode, etc. Further by way of non-limiting example, active semiconductor memory elements include EEPROM and flash memory device elements, which in some embodiments include elements containing a charge storage region, such as a floating gate, conductive nanoparticles or a charge storage dielectric material.
Multiple memory elements may be configured so that they are connected in series or such that each element is individually accessible. By way of non-limiting example, NAND devices contain memory elements (e.g., devices containing a charge storage region) connected in series. For example, a NAND memory array may be configured so that the array is composed of multiple strings of memory in which each string is composed of multiple memory elements sharing a single bit line and accessed as a group. In contrast, memory elements may be configured so that each element is individually accessible (e.g., a NOR memory array). One of skill in the art will recognize that the NAND and NOR memory configurations are exemplary, and memory elements may be otherwise configured.
The semiconductor memory elements included in a single device, such as memory elements located within and/or over the same substrate or in a single die, may be distributed in a two- or three-dimensional manner (such as a two dimensional (2D) memory array structure or a three dimensional (3D) memory array structure).
In a two dimensional memory structure, the semiconductor memory elements are arranged in a single plane or single memory device level. Typically, in a two dimensional memory structure, memory elements are located in a plane (e.g., in an x-z direction plane) which extends substantially parallel to a major surface of a substrate that supports the memory elements. The substrate may be a wafer on which the material layers of the memory elements are deposited and/or in which memory elements are formed or it may be a carrier substrate which is attached to the memory elements after they are formed.
The memory elements may be arranged in the single memory device level in an ordered array, such as in a plurality of rows and/or columns. However, the memory elements may be arranged in non-regular or non-orthogonal configurations as understood by one of skill in the art. The memory elements may each have two or more electrodes or contact lines, including a bit line and a word line.
A three dimensional memory array is organized so that memory elements occupy multiple planes or multiple device levels, forming a structure in three dimensions (i.e., in the x, y and z directions, where the y direction is substantially perpendicular and the x and z directions are substantially parallel to the major surface of the substrate).
As a non-limiting example, each plane in a three dimensional memory array structure may be physically located in two dimensions (one memory level) with multiple two dimensional memory levels to form a three dimensional memory array structure. As another non-limiting example, a three dimensional memory array may be physically structured as multiple vertical columns (e.g., columns extending substantially perpendicular to the major surface of the substrate in the y direction) having multiple elements in each column and therefore having elements spanning several vertically stacked planes of memory devices. The columns may be arranged in a two dimensional configuration (e.g., in an x-z plane), thereby resulting in a three dimensional arrangement of memory elements. One of skill in the art will understand that other configurations of memory elements in three dimensions will also constitute a three dimensional memory array.
By way of non-limiting example, in a three dimensional NAND memory array, the memory elements may be connected together to form a NAND string within a single plane, sometimes called a horizontal (e.g., x-z) plane for ease of discussion. Alternatively, the memory elements may be connected together to extend through multiple parallel planes. Other three dimensional configurations can be envisioned wherein some NAND strings contain memory elements in a single plane of memory elements (sometimes called a memory level) while other strings contain memory elements which extend through multiple parallel planes (sometimes called parallel memory levels). Three dimensional memory arrays may also be designed in a NOR configuration and in a ReRAM configuration.
A monolithic three dimensional memory array is one in which multiple planes of memory elements (also called multiple memory levels) are formed above and/or within a single substrate, such as a semiconductor wafer, according to a sequence of manufacturing operations. In a monolithic 3D memory array, the material layers forming a respective memory level, such as the topmost memory level, are located on top of the material layers forming an underlying memory level, but on the same single substrate. In some implementations, adjacent memory levels of a monolithic 3D memory array optionally share at least one material layer, while in other implementations adjacent memory levels have intervening material layers separating them.
In contrast, two dimensional memory arrays may be formed separately and then integrated together to form a non-monolithic 3D memory device in a hybrid manner. For example, stacked memories have been constructed by forming 2D memory levels on separate substrates and integrating the formed 2D memory levels atop each other. The substrate of each 2D memory level may be thinned or removed prior to integrating it into a 3D memory device. As the individual memory levels are formed on separate substrates, the resulting 3D memory arrays are not monolithic three dimensional memory arrays.
Further, more than one memory array selected from 2D memory arrays and 3D memory arrays (monolithic or hybrid) may be formed separately and then packaged together to form a stacked-chip memory device. A stacked-chip memory device includes multiple planes or layers of memory devices, sometimes called memory levels.
The term “three-dimensional memory device” (or 3D memory device) is herein defined to mean a memory device having multiple layers or multiple levels (e.g., sometimes called multiple memory levels) of memory elements, including any of the following: a memory device having a monolithic or non-monolithic 3D memory array, some non-limiting examples of which are described above; or two or more 2D and/or 3D memory devices, packaged together to form a stacked-chip memory device, some non-limiting examples of which are described above.
A person skilled in the art will recognize that the invention or inventions described and claimed herein are not limited to the two dimensional and three dimensional exemplary structures described here, and instead cover all relevant memory structures suitable for implementing the invention or inventions as described herein and as understood by one skilled in the art.
It will be understood that, although the terms “first,” “second,” etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first target charge could be termed a second target charge, and, similarly, a second target charge could be termed a first target charge, without changing the meaning of the description, so long as all occurrences of the “first target charge” are renamed consistently and all occurrences of the “second target charge” are renamed consistently. The first target charge and the second target charge are both target charges, but they are not the same target charge.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the claims. As used in the description of the embodiments and the appended claims, the singular forms “a,” “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that the term “and/or” as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items. It will be further understood that the terms “comprises” and/or “comprising,” when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
As used herein, the term “if” may be construed to mean “when” or “upon” or “in response to determining” or “in accordance with a determination” or “in response to detecting,” that a stated condition precedent is true, depending on the context. Similarly, the phrase “if it is determined [that a stated condition precedent is true]” or “if [a stated condition precedent is true]” or “when [a stated condition precedent is true]” may be construed to mean “upon determining” or “in response to determining” or “in accordance with a determination” or “upon detecting” or “in response to detecting” that the stated condition precedent is true, depending on the context.
The foregoing description, for purpose of explanation, has been described with reference to specific embodiments. However, the illustrative discussions above are not intended to be exhaustive or to limit the claims to the precise forms disclosed. Many modifications and variations are possible in view of the above teachings. The embodiments were chosen and described in order to best explain principles of operation and practical applications, to thereby enable others skilled in the art.
This application claims priority to U.S. Provisional Patent Application No. 62/298,925, filed Feb. 23, 2016, which is hereby incorporated by reference in its entirety.
| Number | Date | Country | |
|---|---|---|---|
| 62298925 | Feb 2016 | US |
