KR102663422B1 - A system and method for hybrid data reliability for object storage devices - Google Patents

Abstract

According to an embodiment of the present invention, the data storage method of the key-value reliability system includes N storage devices (N is an integer) grouped into a reliability group as a single logical unit and managed by a virtual device management layer. It includes determining whether a threshold corresponding to the reliability mechanism for storing is satisfied, selecting a reliability mechanism if the threshold is satisfied, and storing data according to the selected reliability mechanism.

Description

System and method for hybrid data reliability for object storage devices {A SYSTEM AND METHOD FOR HYBRID DATA RELIABILITY FOR OBJECT STORAGE DEVICES}

One or more aspects of the embodiments of the present invention relate generally to data storage systems, and more specifically, to key-value reliability systems that include a plurality of key-value storage devices. It is about how to select a reliable mechanism for storing data.

Data reliability mechanisms, such as erase coding, can be used to address data loss due to storage device failures and data corruptions in a variety of facilities containing multiple storage devices. .

Conventional solid state drives (SSDs) generally use only a block interface and use RAID (redundant array of independent disks) to perform erasure coding or replication. Data reliability can be provided through . As object formats become more diverse and disorganized, efficient data conversion between object and block level interfaces is required. Moreover, it is necessary to ensure data reliability while maintaining space efficiency and high-speed access time characteristics.

Techniques such as RAID are well studied for conventional block storage devices. However, relatively new key-value storage devices may have different interfaces and different storage semantics than traditional block devices. Accordingly, a variety of new key-value storage devices may potentially benefit from new data reliability techniques suitable for or tailored to key-value data and key-value storage devices.

Because the reliability mechanisms of the embodiments can each perform a single key recovery procedure that allows the virtual device management layer to recover all keys present in a failed memory device and copy them to a new memory device, the embodiments described herein It can provide improvements in the storage area.

According to an embodiment of the present invention, a data storage method of a key-value reliability system including N (where N is an integer) storage devices grouped into a reliability group as a single logical unit and managed by a virtual device management layer is provided. provided. The method includes determining whether the data satisfies a threshold corresponding to a reliability mechanism for storing the data, if the threshold is satisfied, selecting the reliability mechanism, and storing the data according to the selected reliability mechanism. Includes a saving step.

The threshold depends on one or more of the object size of the data, throughput consideration of the data, read/write temperature of the data, and basic erase coding capabilities of the N storage devices. It is based on

The method further includes testing the data for the reliability mechanism using one or more Bloom filters or caches.

The method includes the selected reliability mechanism, one or more checksums for each of the N storage devices storing the data, and an object size of values of the data stored in each of the N storage devices storing the data. , and inserting metadata for recording the locations of parity group members of the N storage devices indicating which of the N storage devices the data is stored in, together with a key corresponding to the data. Includes.

The selected trust mechanism includes object replication, and storing the data includes selecting a KV value, computing a hash to hash a key corresponding to the selected KV value, and a key corresponding to the KV value. Identifying some storage devices among the N storage devices for storing replicas of objects, and writing updated values corresponding to the KV value to each of the determined partial storage devices under the same user key name. Includes steps.

The selected reliability mechanism includes packing, and the step of storing the data includes k key objects stored in k (where k is an integer) storage devices among the N storage devices in the reliability group. selecting, retrieving k value objects corresponding to the k key objects, not having the largest value size among the k value objects so that the virtual sizes of all the k value objects are the same. Padding virtual zeros to the ends of value objects, generating r (where r is an integer) parity objects from the k key objects, writing the k key objects to the k storage devices. and writing the r parity objects to r storage devices among the N storage devices, wherein each of the r storage devices is distinct from the k storage devices, provided that: k+r=N.

The selected reliability mechanism includes packing using typical erasure coding, and the N storage devices are configured with typical (k,r) maximum distance separable (MDS) erasure coding.

The selected reliability mechanism includes packing using regenerative erasure coding, and the N storage devices are configured with (k,r,d) regenerative erasure coding.

The selected reliability mechanism includes splitting, and storing the data includes selecting a KV value, splitting the KV value into k (where k is an integer) equal-sized objects, and k Creating r parity objects (where r is an integer) from objects of the same size, calculating a hash for hashing the key corresponding to the selected KV value, and storing the N storage devices based on the hash. determining a main device in which the KV value is to be located among the main devices, and each of the k objects of the same size and the r parity objects in successive order starting from the main device to the N storage devices. It includes a writing step, provided that k+r=N.

The selected reliability mechanism includes splitting using typical erasure coding, and the N storage devices are configured with typical (k,r) maximum distance separable (MDS) erasure coding.

The selected reliability mechanism includes packing using regenerative erasure coding, the N storage devices are configured with (k,r,d) regenerative erasure coding, and storing the data uses the regenerative erasure coding. It further includes dividing the k objects of the same size into m (where m is an integer) subpackets, and dividing each of the r parity objects into m parity subpackets.

According to another embodiment of the present invention, a data reliability system is provided that stores data based on a selected reliability mechanism. The data reliability system stores data based on a selected reliability mechanism, comprising N (where N is an integer) storage devices configured as virtual devices using stateless data protection, and the selected reliability system. According to the mechanism, a virtual device management layer is configured to manage the N storage devices as the virtual devices to store data in selected storage devices among the N storage devices, wherein the virtual device management layer stores the data. and determine whether a threshold corresponding to a reliability mechanism for storing the data is satisfied, and if the threshold is satisfied, select the reliability mechanism, and store the data according to the selected reliability mechanism.

The selected trust mechanism includes object replication, wherein the virtual device management layer selects a KV value, computes a hash to hash the key corresponding to the selected KV value, and creates replicas of key objects corresponding to the KV value. To store the data by determining some storage devices among the N storage devices and writing updated values corresponding to the KV value to each of the determined storage devices under the same user key name. It is composed.

The selected trust mechanism includes packing, and the virtual device management layer selects k key objects stored in k (where k is an integer) storage devices among the N storage devices, and selects the k keys. Retrieve the k value objects corresponding to the objects, and add virtual value to the ends of the value objects that do not have the largest value size among the k value objects so that the virtual value sizes of all the k value objects are the same. Padding zeros, creating r parity objects (where r is an integer) from the k key objects, writing the k key objects to the k storage devices, and storing the r parity objects in the k storage devices. It is configured to store the data by writing to r storage devices among the N storage devices, where each of the r storage devices is distinct from the k storage devices, and k+r=N.

The selected reliability mechanism includes splitting, wherein the virtual device management layer selects a KV value, splits the KV value into k (where k is an integer) equally sized objects, and Generate r parity objects from objects, where r is an integer, calculate a hash for hashing the key corresponding to the selected KV value, and calculate the KV value among the N storage devices based on the hash. Storing the data by determining the main device on which to locate the main device and writing each of the k objects of the same size and the r parity objects to the N storage devices in successive order starting from the main device. It is configured to do so, provided that k+r=N.

The selected reliability mechanism includes splitting using regenerative erasure coding, the N storage devices are configured with (k,r,d) regenerative erasure coding, and the virtual device management layer uses the regenerative erasure coding. It is further configured to store the data by dividing the k objects of the same size into m (where m is an integer) subpackets and dividing each of the r parity objects into m parity subpackets. .

According to another embodiment of the invention, when executed by a processor, storing data in a key-value reliability system comprising N storage devices grouped into reliability groups as a single logical unit and managed by a virtual device management layer. A non-transitory computer-readable medium containing computer code for performing a method is provided. The method includes determining whether the data satisfies a threshold corresponding to a reliability mechanism for storing the data, if the threshold is satisfied, selecting the reliability mechanism, and storing the data according to the selected reliability mechanism. It includes steps to:

The selected trust mechanism includes object replication, and storing the data includes selecting a KV value, computing a hash to hash a key corresponding to the selected KV value, and a key corresponding to the KV value. selecting some of the N storage devices to store replicas of objects, and writing updated values corresponding to the KV value to each of the identified storage devices under the same user key name. Includes steps.

The selected trust mechanism includes packing, and the step of storing the data includes selecting k key objects stored in k (where k is an integer) storage devices among the N storage devices in the trust group. , Retrieving k value objects corresponding to the k key objects, a value object that does not have the largest value size among the k value objects so that the virtual value sizes of all the k value objects are the same. padding virtual zeros at the ends of the k key objects, generating r (where r is an integer) parity objects from the k key objects, writing the k key objects to the k storage devices. , and writing the r parity objects to r storage devices among the N storage devices, wherein each of the r storage devices is distinct from the k storage devices, and k+r= It's N.

The selected reliability mechanism includes splitting, and storing the data includes selecting a KV value, splitting the KV value into k (where k is an integer) equal-sized objects, and k Creating r (where r is an integer) parity objects from objects of the same size, calculating a hash for hashing the key corresponding to the selected KV value, and based on the hash, the N Selecting a main device among storage devices where the KV value is to be located, and starting from the main device, each of the k objects of the same size and the r parity objects are stored in the N storage devices. It includes the step of writing to the devices, and k+r=N.

The above-described and/or other ideas will become clearer from the detailed description of the embodiments below in conjunction with the accompanying drawings.
1 is a block diagram showing a key-value reliability system that stores key-value data based on a selected reliability mechanism, according to an embodiment of the present invention.
Figure 2 is a flowchart showing the selection of a reliability mechanism to be used by a key-value reliability system based on a size threshold corresponding to the size of the data of a key-value pair, according to an embodiment of the present invention.
3 shows a schematic diagram configured to store key-value data according to the reliability mechanism of K-object (k,r) erasure coding, or multi-object “Packing,” using typical erasure coding, according to an embodiment of the present invention. This is a block diagram showing a group of KV storage devices.
4 shows storage of value objects and parity objects according to the reliability mechanism of K-object (k,r) erasure coding, or multi-object “packing,” using typical erasure coding, according to an embodiment of the present invention. This is a block diagram showing.
5 shows a KV configured to store key-value data according to the reliability mechanism of single object (k,r) erasure coding, or “Splitting,” using typical erasure coding, according to an embodiment of the present invention. This is a block diagram showing a group of storage devices.
6 shows a schematic diagram configured to store key-value data according to the reliability mechanism of single object (k,r,d) erasure coding using regenerative erasure coding, or “Splitting,” according to an embodiment of the present invention. This is a block diagram showing a group of KV storage devices.

Various features of the present invention and methods of achieving the present invention can be understood in more detail with reference to the accompanying drawings and the following detailed description of the embodiments. Hereinafter, embodiments will be described in more detail with reference to the attached drawings. Like reference numerals refer to like elements throughout. However, the present invention may be implemented in various other forms, and should not be understood as being limited to the described embodiments. Rather, these embodiments are provided as examples to aid the overall understanding of the present invention, and will sufficiently convey the technical features and aspects of the present invention to those skilled in the art. Accordingly, processes, elements, and techniques that are unnecessary for a person skilled in the art to fully understand the features and aspects of the invention may not be described. Unless otherwise noted, throughout the accompanying drawings and the detailed description, like reference numerals refer to like elements and their descriptions will not be repeated. In the drawings, the relative sizes of elements, layers, and regions may be exaggerated for clarity.

Various embodiments are described in the text with reference to cross-sectional views, which are schematic diagrams of the embodiments and/or intermediate structures. As such, variations from the depicted forms should be expected, for example as a result of manufacturing techniques and/or tolerances. Moreover, specific structural or functional descriptions described in the text are mere examples for explaining embodiments according to the spirit of the present invention. That is, the embodiments described in the text should not be construed as limited to the specifically illustrated shapes of the regions, but should include, for example, deviations in shape as a result of improvement. For example, implant areas depicted as rectangular will typically have rounded or curved features at their edges and/or a gradient of implant concentration, rather than a binary transition from implant to non-implant area. . Likewise, the buried area formed by the implant may give rise to some implants in the area between the buried area and the surface from which the implant arises. That is, the areas shown in the drawings are schematic in nature, and their shapes are not intended to depict or limit the actual shape of the area of the device. Additionally, as those skilled in the art will recognize, the described embodiments may be modified in various other ways without departing from the spirit or scope of the invention.

In the following detailed description, numerous specific descriptions are provided for convenience of explanation and to aid understanding of various embodiments. However, various embodiments may be implemented without the detailed description or with one or more equivalent substitutes. In other instances, well-known structures and devices are shown in block diagram form in order to avoid unnecessarily obscuring the various embodiments.

Terms such as “first,” “second,” “third,” etc. are used throughout the text to describe various elements, components, areas, layers, and/or zones. Although used, it will be understood that these elements, configurations, regions, layers, and/or zones are not limited to these terms. These terms are used only to distinguish one element, configuration, region, layer, or section from another element, configuration, region, layer, or section. That is, a first element, configuration, region, layer, or section described below may be called a second element, configuration, region, layer, or section without departing from the spirit and scope of the present invention.

Spatially relative terms such as "beneath, below, lower, under", "above, upper", etc. refer to one element(s) shown in the drawings. Alternatively, it can be used in the text to easily explain the relationship between feature(s) and one element or features. It will be understood that spatially relative terms are intended to include other orientations of the device in operation or use in addition to the orientation shown in the figures. For example, if the device in the drawings is turned over, elements described as “below or beneath or under” other elements or features will be oriented “above” the other elements or features. That is, example terms “below, under” may include both up and down directions. The device may be oriented in other directions (e.g., rotated 90 degrees or in other directions), and spatially relative descriptions used in the text should be interpreted accordingly. Similarly, when a first portion is described as being aligned “on” a second portion, this means that the first portion is aligned with the top surface of the second portion, or without limitation to its top surface, based on the direction of gravity. Indicates alignment on the lower surface.

When an element, layer, region, or configuration is referred to as “on, connected to, or coupled to” another element, layer, region, or configuration, it refers to the other element, layer, region, or configuration ( There may be one or more intermediate elements, layers, regions, or configurations that are directly connected (directly). However, the term "directly connected" refers to one component being directly connected to another component without any intermediate components. On the other hand, other expressions that describe relationships between constructs, such as "between, immediately between" or "adjacent to or directly adjacent to" can be interpreted similarly. Additionally, if an element or hierarchy consists of two elements When referred to as between elements or layers, it can be understood that there may be only an element or layer between elements or configurations, or one or more intermediate elements or layers may further exist.

The terms used in the text are only illustrative to describe specific embodiments, and the present invention is not limited thereto. As used in the text, singular terms are intended to include plural forms unless the context clearly dictates otherwise. When the term "comprises" is used in the detailed description, it defines the presence of listed features, integers, steps, operations, elements, and/or configurations, but also one or more other features, integers, and/or elements. It does not exclude the presence or addition of elements, steps, operations, elements, elements, and/or groups thereof. As used in the text, the term “and/or” includes any combination or portion of one or more of the associated listed items.

As used in the text, the terms “substantially,” “about, approximately,” and similar terms are used as terms of approximation and not as terms of degree. It is intended to account for inherent deviations from measured or calculated values that may be recognized by those skilled in the art. As used herein, the term "about, approximately" includes the stated value and is determined by a person skilled in the art, taking into account the errors and doubtful measurements associated with the measurement of a particular quantity (e.g., limitations of the measurement system). It means that it is within the allowable deviation from a certain determined value. For example, “about” can mean within one or more standard deviations or within ±30%, 20%, 10%, 5% of a stated value. As used in the text, the terms “use, using, and used” may be considered synonymous with “utilize, utilizing, and utilized.” Additionally, the term “exemplary” is intended to refer to “an example or illustration.”

If a particular embodiment is implemented differently, a specific process sequence may be performed differently from the described sequence. For example, two sequentially described processes may be performed substantially simultaneously or in the reverse order from that described.

Electrical or electronic devices and/or other related devices or configurations according to embodiments of the present invention described in the text may include appropriate hardware, firmware (e.g., application-specific integrated circuit (ASIC), For example, various configurations of these devices may be formed on a single integrated circuit (IC) chip or on separate IC chips. Moreover, various configurations of these devices can be implemented on a flexible printed circuit film, a tape carrier package (TCP), a printed circuit board (PCB), or on a single substrate. Moreover, various configurations of such devices may be one or more computing devices that communicate with other system components and execute computer program instructions to perform various functions described herein. In the computer program, which may be a process or thread running on one or more processors, the instructions are stored in a computer memory that can be implemented in a computing device using a standard memory device such as random access memory (RAM). Program instructions may also be stored on other non-transitory computer-readable media, such as, for example, a CD-ROM, a flash drive, etc., without departing from the spirit and aspect of the exemplary embodiments of the invention. It is to be understood that the functionality of various computing devices may be combined or integrated into a single computing device, or the functionality of a particular computing device may be distributed to one or more other computing devices.

Unless otherwise defined, all terms, including technical/scientific terms used in the text, have the same meaning as commonly understood by those skilled in the art in the technical field to which the present invention pertains. Terms such as those defined in the public dictionary shall be construed to have a meaning consistent with their meaning in the context of the relevant art and/or the detailed description of the invention, and unless specifically defined in the text, are ideal or It should not be interpreted in an overly formal sense.

As described below, embodiments of the present invention provide a key-value reliability system consisting of a plurality of key-value (KV) storage devices grouped into a single logical unit. -Provides a reliable way to store value data. Moreover, embodiments of the present invention provide a stateless hybrid reliability manager that manages drives and controls storage of key-value (KV) pairs. Stateless Hybris Reliability Manager supports Object Replication; K-Object (k,r) erasure coding - Packing; Single Object (k,r) erasure coding - Splitting; K-Object (k,r,d) regeneration coding - Packing; It relies on multiple pluggable reliability mechanisms/techniques/implementations including Single Object (k,r,d) regeneration coding - Splitting.

A stateless hybrid reliability manager relying on multiple pluggable reliability mechanisms can manage devices and control storage of KV pairs, and the methods described for selection of reliability mechanisms allow efficient storage of KV pairs of different sizes. The described embodiments may enhance memory storage (e.g., storage of key-value data within key-value storage devices) by ensuring recovery, retrieval, and recovery.

1 is a block diagram showing a key-value reliability system that stores key-value data based on a selected reliability mechanism, according to an embodiment of the present invention.

1 , as described above, various new key-value (KV) storage devices/memory devices/drives/KV-SSDs 130 provide key-value data and KV storage devices 130. can potentially benefit from new data reliability mechanisms employed or tailored to . Accordingly, the hybrid key-value reliability system for these KV storage devices 130 manages the KV storage devices 130 using a hybrid reliability mechanism according to one or more pluggable reliability mechanisms. , they may include a stateless hybrid reliability manager/virtual device manager layer/virtual device management layer 120 that controls storage of KV pairs. Although solid-state drives (SSDs) are commonly used to refer to the KV storage devices described in the text, other storage devices may be used according to embodiments of the present invention. The design and operations of the virtual device management layer 120 according to embodiments of the present invention will be described below.

In a current embodiment, virtual device management layer 120 may enable a method of reliably storing key-value data/KV pairs 170 in a key-value reliability system. The key-value reliability system may include a plurality of KV storage devices 130 grouped into one logical unit. A logical unit may be referred to as a reliability group 140.

KV storage devices 130 of reliability group 140 may store respective chunks of erasure coded data and/or replicated data that may correspond to key-value data 170 . The KV storage devices 130 of the reliability group 140 appear as a single virtual device 110 whose key-value operations are directed to via the virtual device management layer 120.

The virtual device 110 may include a stateless hybrid reliability manager as the virtual device management layer 120. That is, the virtual device management layer 120 may operate in a stateless manner (i.e., does not require maintaining mapping between key-values and devices).

Accordingly, the virtual device 110 includes N KV storage devices 130 (where N is an integer) (e.g., KV-SSDs 130-1, 130-2, 130-3, 130-4, . . . The key-value data 170 can be stored through 130-N)), and the key-value data 170 can be stored in KV storage devices 130 through the virtual device management layer 120. That is, the virtual device management layer 120 can manage the KV storage devices 130 and control storing KV pairs therein.

In other embodiments, the key-value reliability system may also include a cache to selectively store metadata and/or data associated with the keys of key-value data 170 to improve operation speed. Reliability mechanisms can attach values to KV pairs to include metadata corresponding to the KV pairs. That is, both the key and the value can be appended with a metadata identifier “MetaID” and corresponding information to store additional metadata specific to the KV pair. Metadata may include a checksum, a reliability mechanism identifier to identify the reliability mechanisms used to store the data, an erasure code identifier, object sizes, location of parity group numbers, etc.

In other embodiments, the key-value reliability system may further include bloom filters corresponding to the reliability mechanisms described below. Bloom filters may store stored keys using a corresponding reliability mechanism, thereby assisting the key-value reliability system in read operations. Accordingly, one or more Bloom filters or caches in a key-value reliability system may make it possible to quickly test keys against existing reliability mechanisms.

Each of the reliability mechanisms described in the text uses the same hash function for the key modulo of a corresponding number of KV storage devices 130 to store key-value data 170 and a corresponding KV pair. The first copy or chunk of can be stored first. That is, for each of the pluggable trust mechanisms, the trust mechanism may store at least the first copy/chunk using the same key as the user key.

As described above, embodiments of the present invention provide multiple pluggable reliability mechanisms to ensure reliable storage of key-value data 170 in multiple KV storage devices 130. do. Accordingly, virtual device management layer 120 may require reliability mechanisms and may determine which of the reliability mechanisms is used.

Reliability mechanisms may be based on policies such as value-size thresholds, and/or object read/write frequency, set during installation of virtual device 110. Accordingly, the virtual device management layer 120 can select an appropriate reliability mechanism based on the stated policies of the system.

The five reliability mechanisms of embodiments of the present invention are described below, how the reliability mechanism operates by the virtual device management layer 120, and when the reliability mechanism can be appropriately used and selected. These reliability mechanisms include Object Replication, K-Object (k,r) erasure coding, Erasure Coding - Packing, and Single Object (k,r) Erasure Coding - Packing. Splitting (Single Object (k,r) erasure coding - Splitting), K-Object (k,r,d) regeneration coding - Packing (K-Object (k,r,d) regeneration coding - Packing), and single object It may be referred to as (k,r,d) regeneration coding - splitting (Single Object (k,r,d) regeneration coding - Splitting).

Figure 2 is a flowchart 200 showing the selection of a trust mechanism used by a key-value trust system based on a size threshold corresponding to the size of the data of a KV pair, according to an embodiment of the present invention.

2, for all of the supported reliability mechanisms based on size thresholds (e.g., the five above-described trust mechanisms), virtual device management layer 120 stores data (e.g., key-value It is possible to determine the value size of data 170), and it is possible to determine whether the value size is smaller than a given threshold (t _i ) corresponding to each reliability mechanism, and a first reliability mechanism that satisfies the value size threshold requirement. You can choose.

For example, at S210, virtual device management layer 120 may receive “n” (n is an integer) number of size threshold-based supported reliability mechanisms. At S220, virtual device management layer 120 may simply review each of the reliability mechanisms, one at a time, in order from 1 to n. At S230, in reviewing each of the supported reliability mechanisms, the virtual device management layer 120 may determine whether the value size of the data is smaller than the threshold (t _i ) corresponding to each reliability mechanism.

At S240, if a reliability mechanism is found with a threshold ti greater than or equal to the value size of the data, virtual device management layer 120 may select that reliability mechanism for use. In S250, if the reliability mechanism to be used in S240 is determined or, in the final iteration of S220, it is determined that there is no reliability mechanism with an appropriate threshold (t _i ) satisfying the value size among the n reliability mechanisms, the virtual device management layer ( 120) may end up determining the reliability mechanism to be used.

In the current embodiment, “n” may be equal to 5 according to the five different reliability mechanisms of the embodiments described in the text. For relatively very small key-values (i.e., when the value size is relatively small), the virtual device management layer 120 may select the reliability mechanism of Object Replication for use. For slightly larger key-values, the virtual device management layer 120 may select the reliability mechanisms of Packing, and then Splitting (e.g., in that order), while For , typical erasure coding can be used. However, for larger key-values, virtual device management layer 120 may choose Packing, then Splitting, while regeneration erasure coding instead of typical erasure coding. can be used.

In embodiments of the invention, the selection of a reliability mechanism to use depends on the object size of the object, the throughput requirements for the object, the read/write temperature of the corresponding key-value pair, and the plurality of KV storage devices. It may be based on one or more of their basic coding capabilities, and/or detection of whether the key is hot or cold. For example, "hot" keys can use the reliability mechanism of Object Replication, regardless of their value size, whereas "cold" keys can use the reliability mechanism of Object Replication, regardless of their value size. Accordingly, it can be applied as one of the reliability mechanisms of the erasure coding method. As another example, the decision to use the reliability mechanism of Object Replication may be based on both size and write temperature. Accordingly, the threshold corresponding to the object read/write cycle can be used to determine the reliability mechanism, instead of the threshold corresponding to the size in the flowchart 200 of FIG. 2.

The operation of each of the five reliability mechanisms is described below.

Referring back to Figure 1, as previously mentioned, when the value size of a KV pair is relatively small, the reliability mechanism of “Object Replication” may be suitable for selection by virtual device management layer 120. You can. Object replication may be applied per object (eg, per KV pair/key-value data 170). Although the reliability mechanism of Object Replication has high storage overhead, it has low read and restore costs, so it can be suitable for very small value sizes.

The reliability mechanism of Object Replication may also be suitable for key-values with frequent updates (e.g., key-value data 170) and may therefore be selected based on read and write cycles. .

During object replication, each time a write occurs, the object/key-value data 170 is replicated to one or more additional KV storage devices 130. A primary copy of key-value data 170 may be located in one of the KV storage devices 130 specified by the hash of the key modulo (N). Replicas of the primary copy of key-value data 170 may be located in consecutive KV storage devices 130, or immediately adjacent KV storage devices 130, in a circular manner. there is.

The virtual device management layer 120 or the user can decide how many copies of key-value data 170 to create. For example, a distributed system using virtual device management layer 120 can select 3-way replication and create 3-way replication by default. However, the user of the system can configure the number of replicas of an object to be more or less than the selected default.

Thus, for example, if three-way replication is used and primary KV storage device 130-2 contains a primary copy of data (e.g., key-value data 170), the virtual device management layer ( 120 may store a replica of the primary copy of the data in subsequent replica KV storage devices 130-3, 130-4, with all copies of the data being identical. That is, copies of the data are stored in two (or more) immediately subsequent KV storage devices 130-3 and 130-4 ( For example, it is stored in a circular manner).

Copies of data may be stored under the same keyname/same user key as primary KV storage device 130-2, such as in replica KV storage devices 130-3, 130-4. there is. All copies of the data may include an identifier and checksum to identify the replicated key-value data 170.

Accordingly, when a specific KV storage device 130 fails (for example, when the KV storage device 130-3 fails), the KV storage devices 130 immediately before and after the failed KV storage device 130 ( For example, by recovering the value using a recovery mechanism for key names in the KV storage devices 130-2 and 130-4 immediately before and after the KV storage device 130-3, Ensures that duplicated keys are recovered.

To summarize the reliability mechanism of object replication, the virtual device management layer 120 can receive key-value data 170, and the KV storage device to be used to hash the key object and store replicas of the key object. (130) can be determined. The virtual device management layer 120 then stores the updated values, under the same user key name (e.g., an appropriate MetaID field), in the selected KV storage devices 130 (e.g., the selected KV storage devices (e.g., It can be written as 130-2, 130-3, 130-4).

3 shows K-Object (k,r) erasure coding or multi-object “packing” using traditional erasure coding, according to an embodiment of the present invention. This is a block diagram showing a group of KV storage devices configured to store key-value data according to the reliability mechanism of “Packing” (multiple object “”).

Referring to Figure 3, the reliability mechanism of packing using typical erasure coding is to handle data with small value sizes where it is not appropriate to split into chunks (e.g., for better data throughput). can be selected for For example, the reliability mechanism of Packing using typical erasure coding results in value sizes that are larger (but still relatively may be selected by the virtual device management layer 120 for data with small).

Packing using typical erasure coding can consist of typical (k,r) maximum distance separable (MDS) erasure coding and can be used with a systematic MDS code. As an example, the elimination code could be essentially a (4,2) Reed-Solomon Code, and the (4,2) Reed-Solomon Code is relatively well-studied and has a corresponding fast implementation library. is easily available.

Using the reliability mechanism of Packing using typical erasure coding, k keys/key objects from queues of k different KV storage devices 330 that are part of the same parity group/erase code group 340. Fields 350 are selected, erase coded, and packed. (k is an integer.)

For example, virtual device management layer 120 may store recently written key objects for each KV storage device 330 (e.g., each KV storage device 130 of trust group 140 in FIG. 1). Maintains a buffer of 350 so that virtual device management layer 120 can select k key objects 350 from k different KV storage devices to be erasure coded, thereby Pack k key objects 350.

In the current embodiment, virtual device management layer 120 stores four key objects (350x, 350y, 350b) from four different KV storage devices (330-1, 330-3, 330-4, 330-N), respectively. , 350c) is selected. (In the current embodiment, k=4)

4 shows storage of value objects and parity objects according to the reliability mechanism of K-object (k,r) erasure coding, or multi-object “packing,” using typical erasure coding, according to an embodiment of the present invention. This is a block diagram showing.

3 and 4, key objects 350 (key modulo hash of n) are located in ^th KV storage device 330. That is, for each key object 350, each hash of the key modulo n that can be transmitted to the queue of the specific KV storage device 330 may be performed. In the current embodiment, the ith key (Key _i ) 350-i is hashed and located in KV-SSD1 (330-1), and the jth key (Key _j ) 350-j is hashed and located in KV-SSD1 (330-1). It is located in -SSD2 (330-2), and the k-th key (Key _k ) (350-k) is hashed and located in KV-SSD4 (330-4).

The user value length/value size 462 of each stored value object 450 is the same as written. However, for consistency to enable erasure coding, user values/value objects 450 have "0" fillings/virtual zeros/virtual zero padding 464 attached to them so that they all have the same size. It appears that That is, because the respective user value sizes 462 of different value objects 450 may vary (i.e., value objects 450 may have differential lengths or may be differential-length key values). ), by implementing a method of virtual zero padding 464 (i.e., padding the zeros of virtual zero padding 464 into value objects 450 for coding purposes, while value objects 450 containing padded zeros) ), the parity objects 460 may have the same size as the largest object(s) 470 in the parity group 340 . Accordingly, in the current embodiment, value objects 450 “Val x,” “Val y,” and “Val b” are padded with virtual zeros that are not actually stored in KV storage devices 330, thereby It appears to have the same size as the value object 470 “Val c”. Therefore, parity objects 460 can be computed.

After coding the k key objects 350, the virtual device management layer 120 generates r parity objects from the k values/k value objects 450 corresponding to the k key objects 350. 460 can be calculated. At this time, r is an integer, k+r=N, and N is the KV storage devices 330 of the parity group 340 (e.g., the N KV storage devices of the reliability group 140 in FIG. 1 (130)).

The virtual device management layer 120 divides the r parity objects 460 into the r remaining other KV storage devices 330 of the parity group 340 (i.e., the k key objects 350 are selected and erase coded). r KV storage devices are distinguished from the k KV storage devices 330 containing the queues. Accordingly, each of the k key objects 350 and r parity objects 460 may be stored in each of the N storage devices 330, and the data corresponding to them may be stored in the parity group 340. It is evenly distributed to each of the N KV storage devices 330.

Although reads and writes are relatively simple with respect to the reliability mechanism of Packing using typical erasure coding, recalculation and recovery of parity can be less simple. For parity recovery and operations (e.g., in the case of updates), which key objects 350 are grouped together in the same parity group 340 and thereby perform parity operations (e.g., which In order to recognize whether the key object 350 belongs to the erase code group 340, information about the groups of key objects 350 includes the actual value size 462 ( That is, stored as a metadata object on each of the KV storage devices 330 (e.g., KV storage devices 130 of FIG. 1), with a value size 462 without virtual zero padding 464. It can be. Accordingly, in the current embodiment, additional metadata may be stored in key objects 350 (e.g., key objects 350 located in trust group 140 in Figure 1), corresponding to key objects 350. The original length of each of the value objects 450 and the coding order of the key objects 350 may be used to store the KV storage devices 130 .

For example, the metadata object value may indicate all of the key objects 350 of the trust group 140, and may also include a field indicating the value sizes 462 of the value objects 450, and the parity object. The corresponding stored keys (i.e., value objects 450 containing zeros of virtual zero padding 464), value sizes 462 of parity objects 460, and r parity objects 460 It may include another field indicating device IDs for identification of the r number of KV storage devices 330.

Data may be stored using user keys. Metadata may be stored in an internal key formed using the user key and a MetaID directive that refers to "Metadata". Moreover, for accurate reconstruction when value objects 450 are regenerated by determining where the value objects 450 end and where the zeros of the virtual zero padding 464 begin, the value sizes 462 are It may be stored in metadata to recognize the location of the virtual zero padding 464 (i.e., where the zeros were added).

If one of the KV storage devices 330 fails, data and metadata may be potentially lost because both data and metadata may be stored in the same KV storage device 330, thereby making recovery difficult. It may be impossible. However, to prevent this situation, the metadata object value is an “object replication engine” of the virtual device management layer 120, which can implement the previously mentioned reliability mechanism of Object Replication for the metadata object value. It can be replicated using the “Replication Engine”.

Additionally, since the metadata object value is the same for all objects in the trust group 140, if the KV storage device 330 supports object linking, the same metadata object value is stored on the same KV storage device ( 330) may be connected to a plurality of key names that are commonly located in 330). Moreover, if batch writing is supported, object values can be batched together for better throughput.

To summarize the reliability mechanism of packing using typical erasure coding according to the current embodiment, the virtual device management layer 120 receives k recently stored keys from k different KV storage devices 330 through a buffer. Objects 350 can be selected. The virtual device management layer 120 then retrieves the value object 450 (different from the largest value object(s) 470 of the parity group 440) corresponding to each key object 350. And, the value objects 450 can be made to have the same size (eg, the size of the largest value object(s) 470) by padding them with virtual zero padding 350. The virtual device management layer 120 may then generate r parity objects from the k key objects 350 using an MDS code process. The virtual device management layer 120 later divides the r parity objects 460 into the k KV storage devices 330 from which the key objects 350 were selected among the N KV storage devices 330. It can be written to other r KV storage devices 330. At this time, k+r is equal to N. The virtual device management layer 120 may then create a metadata object representing the above-described information. Finally, the virtual device management layer 120 stores key objects 350 and parity objects 460, along with keys formed from user keys and metadata identifiers, on N KV storage devices 330 (e.g., (similar to a replication engine).

5 illustrates a reliability mechanism of Single Object (k,r) erasure coding, or “Splitting,” using typical erasure coding, according to an embodiment of the present invention. This is a block diagram showing a group of KV storage devices configured to store key-value data.

Referring to Figure 5, for values having value sizes larger than those of values suitable for the previously mentioned reliability mechanisms of Object Replication and Packing using typical erasure coding, The virtual device management layer 120 may select a reliability mechanism of single object (k,r) erasure coding using traditional erasure coding, or “splitting.” The reliability mechanism of splitting using typical erasure coding is good if the value size is relatively large and the KV value (570) is divided into k equally sized splits/chunks/values/objects (550). It is a reliability mechanism in object/KV pair units that can be suitable for KV value/object 570 that can have throughput.

After dividing the KV value 570, according to an embodiment, the virtual device management layer 120 may calculate a checksum for each of the k objects 550. Afterwards, the virtual device management layer 120 may insert metadata before each of the k objects 550.

Splitting using typical erasure coding divides the KV value 570 into a plurality of smaller objects 550, and then divides the plurality of smaller objects 550 of the KV value 570 into k It may include distributing onto successive storage devices 530 . Accordingly, the size of k objects 550 of the same size can be supported by each of the underlying KV storage devices 530.

When using splitting using typical erasure coding, the virtual device management layer 120 uses a systematic MDS code (e.g., virtual device layer 120 as the base code, such as a (4,2) Reed-Solomon code). We can also add r parity values/objects 560 generated using typical (k,r) MDS erasure coding. Thereafter, in a manner similar to the reliability mechanism of Packing using typical erasure coding described above, the virtual device management layer 120 divides the k objects 550 and the r parity objects 560 into N It can be written as KV storage devices 530. (k+r=N)

Accordingly, the virtual device management layer 120 can divide the relatively large KV value 570 into k objects 550, calculate and add r parity objects 560, and k objects 550 and r parity objects 560 may be stored in k+r KV storage devices 530 .

When using the reliability mechanism of splitting using typical erasure coding, after hashing the key 580 corresponding to the KV value 570, the virtual device management layer 120 stores the corresponding object (e.g. For example, the first of the k objects 550 (D1 in the embodiment of FIG. 5 may be stored in hash mark zero) may be stored in the main KV storage device 530a (e.g., shown in FIG. In an embodiment, KV-SSD2) can be determined. k+r objects 550, 560 may be written to each of the primary KV storage device 530a and N-1 consecutive KV storage devices 530 under the same user key name. That is, in the embodiment shown in FIG. 5, the first of the k objects 550 may be written to the main KV storage device 530a “KV-SSD2”, and the remaining of the k objects 550 may be written to Together with the r parity objects 560, they are written in the KV storage devices 530 “KV-SSD3” to “KV-SSDN” and “KV-SSD1” in a cyclical order. (For example, in a similar way to that described for the object replication reliability mechanism described above.)

To summarize the reliability mechanism of splitting using typical erasure coding, the virtual device management layer 120 can split a relatively large KV object 570 into k objects 550 of the same size. The virtual device management layer 120 may then generate r parity objects 560 for the k objects 550 using an MDC code process. The virtual device management layer 120 may later hash the key corresponding to the KV value 570 to determine the primary KV storage device 530a in which the object will be located. The virtual device management layer 120 is subsequently created by the virtual device management layer 120 and corresponds to the main KV storage device 530a and N-1 consecutive KV storage devices 530 in a circular manner. k+r objects 550, 560 may be entered under the same user key name, which may include an appropriate MetaID field.

Referring back to Figures 3 and 4, according to another embodiment, virtual device management layer 120 may use regenerative erasure coding, K-object (k,r,d) erasure coding, or multi-object "packing". )" reliability mechanism may be selected (e.g., according to flowchart 200 of FIG. 2). The current reliability mechanism is similar to the reliability mechanism of Packing using typical erasure coding described previously in that the virtual device management layer 120 packs the k objects into the k KV storage devices. However, packing using regenerative erasure coding uses (k,r,d) regenerative codes, instead of using typical (k,r) erasure codes. Accordingly, FIGS. 3 and 4 may be referenced generally with respect to the present embodiment.

Therefore, in cases where regenerative codes are suitable, but splitting objects is not suitable, and preserving objects is more suitable, packing using regenerative erasure coding can be used. Packing using regenerative erasure coding may be suitable for larger value sizes than those used for object replication, previously mentioned, and the reliability mechanisms of splitting and packing using traditional erasure coding. Packing using regenerative erasure coding may be used when reading multiple subpackets of an object does not result in lower performance than reading the entire object. Packing using regenerative erasure coding allows primary KV storage devices (e.g., KV storage devices 130 in FIG. 1, or KV storage devices 330 in FIG. 3) to assist during recovery/reconstruction. This may be suitable for KV storage devices that recognize the regeneration code.

6 shows a schematic diagram configured to store key-value data according to a reliable mechanism of single object (k,r,d) erasure coding using regenerative erasure coding, or “Splitting,” according to an embodiment of the present invention. This is a block diagram showing a group of KV storage devices.

Referring to Figure 6, the current reliability mechanism is that the virtual device management layer is as shown in Figure 4, except that it uses (k,r,d) regeneration codes instead of the typical (k,r) MDS erasure coding. Likewise, it can be made to behave in a similar way to splitting using typical erasure coding. The current reliability mechanism, such as packing using regenerative erasure coding, may be suitable in cases where the underlying KV storage devices 630 are KV storage devices that recognize regenerative code that can assist during recovery/reconstruction.

If object 670 has a larger value size than the corresponding objects with previously described reliability mechanisms, then reading a plurality of subpackets 690 of k splits 680 of object 670 may result in an overall split. Splitting using regenerative erasure coding may be suitable if it does not result in lower performance than reading fields 680 (e.g., performed with Split's reliability mechanism using traditional erasure coding).

The reliability mechanism of splitting using regenerative erasure coding is that objects are partitioned into k equally sized objects/splits 650 and the splits 650 are divided into a number of subpackets 690 (e.g. In an embodiment, it is virtually further split into four subpackets 690 per split 650, and reading a plurality of subpackets 690 from an object 670 results in the entire object 670 ) is an object (KV pair)-level mechanism that can be suitable for objects/KV values with very large value sizes, which can have suitable throughput in cases where the throughput is better than reading ). Value size is supported by all basic KV storage devices 630.

Similar to splitting using typical erasure coding, as shown in Figure 4, the virtual device management layer 120 of the current reliability mechanism can add r parity objects 660 using a systematic regeneration code. And, k splits 650 and r parity objects 660 can be written to N KV storage devices 630. (k+r=N) However, each of the r parity objects 660 may contain a number of parity subpackets 692 (e.g., split/number of subpackets 690 per k object 605 and can be divided into corresponding numbers). Unlike splitting using typical erasure coding, in the current embodiment the base code may be a (4,2,5) zigzag code.

To summarize the reliability mechanism of splitting using regenerative erasure coding, the virtual device layer 120 can split a large KV value 670 into k objects 650 of the same size. The virtual device management layer 120 may then divide each of the k objects 650 into m subpackets 690 of the same size. m is an integer. The virtual device management layer 120 may then use a regenerative coding process to generate r parity objects 660 for the k objects 650, each of the r parity objects 660 Can be divided into m parity subpackets 692 of the same size. The virtual device management layer 120 may later hash the key corresponding to the KV value 670 to determine the primary KV storage device 630a in which the object will be located. The virtual device management layer 120 is subsequently created by the virtual device management layer 120 and corresponds to the main KV storage device 530a and N-1 consecutive KV storage devices 530 in a circular manner. K+r objects 650, 660, each containing m subpackets 690, 692, may be entered under the same user key name, which may include an appropriate MetaID field.

As discussed above, the virtual device management layer may select an appropriate trust mechanism from a group of trust mechanisms for storage of the data based on one or more attributes of the data. Accordingly, the embodiments described herein provide improvements in the field of memory storage because the described reliability mechanisms are each capable of performing a single key recovery procedure. When all memory devices fail, the virtual device management layer of an embodiment of the present invention can recover all keys existing in the failed memory device and copy them to a new memory device. The virtual device management layer performs a repetitive operation on all keys existing in the failed memory device and adjacent memory devices in the reliability group, and performs a key-by-key recovery operation on the keys that the reliability mechanism determines to exist in the failed memory device. , recovery and copying of all keys can be achieved.

Very large KV pairs with value sizes larger than those supported by the underlying reliability mechanisms (e.g., based on underlying storage device constraints) are explicitly split into multiple KV pairs by the reliability manager, and the reliability mechanism The described embodiments provide further improvements in the area of memory storage because they store multiple splits and split number information along with metadata stored in values.

Although the embodiments described in the text use specific terms, they should be interpreted in a general and technical sense, and are not limited thereto. In some instances, features, attributes, and/or features described in connection with a particular embodiment, as would appear by a person of ordinary skill in the art to which the present invention pertains, unless otherwise stated with respect to the indicated embodiment. Elements may be used independently or in combination with features, attributes, and/or elements described in connection with other embodiments. Accordingly, it will be understood by those skilled in the art that various changes in form or description, as well as functional equivalents thereof, may be made without departing from the spirit and scope of the description and the following claims.

Claims (20)

A data storage method in a key-value reliability system comprising N (where N is an integer) storage devices grouped into reliability groups as a single logical unit and managed by a virtual device management layer, comprising:
determining whether the data satisfies a threshold corresponding to a reliability mechanism for storing the data;
if the threshold is met, selecting the reliability mechanism; and
storing the data according to the selected reliability mechanism,
The selected reliability mechanism includes object replication,
The steps for storing the data are:
Selecting KV Value;
calculating a hash for hashing the key corresponding to the selected KV value;
determining some of the N storage devices to store replicas of key objects corresponding to the KV value; and
A data storage method in a key-value reliability system comprising the step of writing updated values corresponding to the KV value into each of the determined storage devices under the same user key name. According to claim 1,
The threshold depends on one or more of the object size of the data, throughput consideration of the data, read/write temperature of the data, and basic erase coding capabilities of the N storage devices. Data storage method of key-value reliability system based. According to claim 1,
A method of data storage in a key-value reliability system further comprising testing the data against the reliability mechanism using one or more Bloom filters or caches. According to claim 1,
the selected reliability mechanism, one or more checksums for each of the N storage devices storing the data, object sizes of values of the data stored on each of the N storage devices storing the data, and The key further comprising inserting metadata for recording the locations of parity group members of the N storage devices indicating which of the N storage devices the data is stored in, together with a key corresponding to the data. -Data storage method in value reliability system.
. delete According to claim 1,
The selected reliability mechanism includes packing,
The steps for storing the data are:
selecting k key objects stored in k (where k is an integer) storage devices among the N storage devices of the reliability group;
Retrieving k value objects corresponding to the k key objects;
padding virtual zeros at the ends of value objects that do not have the largest value size among the k value objects so that the virtual value sizes of all the k value objects are the same;
generating r parity objects (where r is an integer) from the k key objects;
writing the k key objects to the k storage devices; and
Comprising writing the r parity objects to r storage devices among the N storage devices,
Each of the r storage devices is distinct from the k storage devices, provided that k+r=N. According to claim 6,
The selected reliability mechanism includes packing using typical erasure coding,
The data storage method of the key-value reliability system in which the N storage devices are configured with typical (k,r) MDS (maximum distance separable) erasure coding. According to claim 6,
The selected reliability mechanism includes packing using regenerative erasure coding,
The data storage method of the key-value reliability system in which the N storage devices are configured with (k, r, d) regenerative erasure coding. According to claim 1,
The selected reliability mechanism includes splitting,
The steps for storing the data are:
Selecting KV Value;
Dividing the KV value into k (where k is an integer) objects of the same size;
generating r parity objects (where r is an integer) from the k objects of the same size;
calculating a hash for hashing the key corresponding to the selected KV value;
determining a main device among the N storage devices where the KV value is located based on the hash; and
Starting from the main device, writing each of the k objects of the same size and the r parity objects to the N storage devices in successive order, provided that the key where k+r=N. -Data storage method in value reliability system. According to clause 9,
The selected reliability mechanism includes splitting using typical erasure coding,
The data storage method of the key-value reliability system in which the N storage devices are configured with typical (k,r) MDS (maximum distance separable) erasure coding. According to clause 9,
The selected reliability mechanism includes packing using regenerative erasure coding,
The N storage devices are configured with (k, r, d) regenerative erasure coding,
The steps for storing the data are:
Splitting each of the k objects of the same size into m (where m is an integer) subpackets using the regenerative erasure coding; and
A data storage method in a key-value reliability system further comprising dividing each of the r parity objects into m parity subpackets. In a data reliability system that stores data based on a selected reliability mechanism,
N (where N is an integer) storage devices configured as virtual devices using stateless data protection; and
a virtual device management layer configured to manage the N storage devices as the virtual devices and store data in selected storage devices among the N storage devices according to the selected reliability mechanism;
The virtual device management layer:
determine whether the data satisfies a threshold corresponding to a reliability mechanism for storing the data;
If the threshold is met, select the reliability mechanism;
configured to store the data according to the selected reliability mechanism;
The selected reliability mechanism includes object replication,
The virtual device management layer:
Select KV Value;
Compute a hash for hashing the key corresponding to the selected KV value;
determine some of the N storage devices to store replicas of key objects corresponding to the KV value;
A data reliability system configured to store the data by writing updated values corresponding to the KV value into each of the identified storage devices under the same user key name. delete According to claim 12,
The selected reliability mechanism includes packing,
The virtual device management layer:
select k key objects stored in k (where k is an integer) storage devices among the N storage devices;
retrieve k value objects corresponding to the k key objects;
padding virtual zeros on the ends of value objects that do not have the largest value size among the k value objects so that the virtual value sizes of all of the k value objects are the same;
Generate r parity objects (where r is an integer) from the k key objects;
write the k key objects to the k storage devices;
configured to store the data by writing the r parity objects to r storage devices among the N storage devices,
A data reliability system where each of the r storage devices is distinct from the k storage devices, and k+r=N. According to claim 12,
The selected reliability mechanism includes splitting,
The virtual device management layer:
Select KV Value,
Split the KV value into k (where k is an integer) objects of equal size;
Generate r parity objects from the k objects of the same size, where r is an integer;
Compute a hash for hashing the key corresponding to the selected KV value;
Based on the hash, determine a main device among the N storage devices where the KV value is located;
configured to store the data by writing each of the k objects of the same size and the r parity objects to the N storage devices in successive order starting from the main device,
However, a data reliability system where k+r=N. According to claim 15,
The selected reliability mechanism includes splitting using regenerative erasure coding,
The N storage devices are configured with (k, r, d) regenerative erasure coding,
The virtual device management layer divides the k objects of the same size into m (where m is an integer) subpackets using the regenerative erasure coding, and divides each of the r parity objects into m parity subpackets. A data reliability system further configured to store the data by dividing into groups. Comprising computer code that, when executed by a processor, performs a method of storing data in a key-value reliability system comprising N storage devices grouped into a reliability group as a single logical unit and managed by a virtual device management layer. In a non-transitory computer-readable medium,
The above method is:
determining whether the data satisfies a threshold corresponding to a reliability mechanism for storing the data;
if the threshold is met, selecting the reliability mechanism; and
storing data according to the selected reliability mechanism,
The selected reliability mechanism includes object replication,
The steps for storing the data are:
Selecting KV Value;
calculating a hash for hashing the key corresponding to the selected KV value;
selecting some of the N storage devices to store copies of key objects corresponding to the KV value; and
and writing updated values corresponding to the KV value to each of the identified storage devices under the same user key name. delete According to claim 17,
The selected reliability mechanism includes packing,
The steps for storing the data are:
selecting k key objects stored in k (where k is an integer) storage devices among the N storage devices of the reliability group;
Retrieving k value objects corresponding to the k key objects;
padding the ends of value objects that do not have the largest value size among the k value objects with virtual zeros so that the virtual value sizes of all of the k value objects are the same;
generating r parity objects (where r is an integer) from the k key objects;
writing the k key objects to the k storage devices; and
Comprising writing the r parity objects to r storage devices among the N storage devices,
Each of the r storage devices is distinct from the k storage devices, and k+r=N. According to claim 17,
The selected reliability mechanism includes splitting,
The steps for storing the data are:
Selecting KV Value;
Dividing the KV value into k (where k is an integer) objects of the same size;
generating r parity objects (where r is an integer) from the k objects of the same size;
calculating a hash for hashing the key corresponding to the selected KV value;
Based on the hash, selecting a main device among the N storage devices where the KV value will be located; and
Writing each of the k objects of the same size and the r parity objects to the N storage devices in sequential order starting from the main device,
Non-transitory computer readable medium where k+r=N.

Images