Tenant-level sharding of disks with tenant-specific storage modules to enable policies per tenant in a distributed storage system

Abstract

Embodiments include receiving an indication of a data storage module to be associated with a tenant of a distributed storage system, allocating a partition of a disk for data of the tenant, creating a first association between the data storage module and the disk partition, creating a second association between the data storage module and the tenant, and creating rules for the data storage module based on one or more policies configured for the tenant. Embodiments further include receiving an indication of a type of subscription model selected for the tenant, and selecting the disk partition to be allocated based, at least in part, on the subscription model selected for the tenant. More specific embodiments include generating a storage map indicating the first association between the data storage module and the disk partition and indicating the second association between the data storage module and the tenant.

Description

TECHNICAL FIELD

This disclosure relates in general to the field of data storage and, in particular, to tenant-level sharding of disks with tenant-specific storage modules to enable policies per tenant in a distributed storage system.

BACKGROUND

In recent years, cloud-based storage has emerged to offer a solution for storing, accessing, and protecting electronic data owned or controlled by all types of private and public entities. Distributed storage systems may offer a storage platform designed to provide object based, block based, and file based storage from a single distributed storage cluster in a cloud. A distributed storage cluster may contain numerous nodes for storing objects and other data. Generally, a single storage cluster of a distributed storage system, such as Ceph, is designed to accommodate data from multiple tenants, where the same set of rules and weights apply to all of the tenants. Typically, data belonging to the multiple tenants share the same storage device daemons or other software and disk partitions. Tenants, however, sometimes prefer to receive particular types and levels of service.

BRIEF DESCRIPTION OF THE DRAWINGS

To provide a more complete understanding of the present disclosure and features and advantages thereof, reference is made to the following description, taken in conjunction with the accompanying figures, wherein like reference numerals represent like parts, in which:

FIG. 1 is a simplified block diagram of a distributed storage system according to at least one embodiment of the present disclosure;

FIG. 2 is a simplified block diagram showing additional possible details of a storage node in the distributed storage system according to at least one embodiment;

FIG. 3 is a simplified flowchart illustrating possible activities associated with the distributed storage system according to at least one embodiment; and

FIG. 4 is a simplified flowchart illustrating other possible activities associated with the distributed storage system according to at least one embodiment.

DETAILED DESCRIPTION

Overview

The present disclosure describes a recovery system for a distributed storage system. A method is provided in one example of the present disclosure and includes receiving an indication of a data storage module to be associated with a tenant of a distributed storage system, allocating a partition of a disk for data of the tenant, creating a first association between the data storage module and the disk partition, creating a second association between the data storage module and the tenant, and creating rules for the data storage module based on one or more policies configured for the tenant.

In specific embodiments, the method further includes receiving an indication of a type of subscription model selected for the tenant, and selecting the disk partition to be allocated based, at least in part, on the subscription model selected for the tenant. In further specific embodiments, the data storage module can store at least some data of the tenant in the disk partition based, at least in part, on the rules. In some embodiments, the one or more policies are related to at least one of a performance requirement of disks to store the data, a distribution requirement for the data, and a replication requirement for the data.

In at least some embodiments, the method includes generating a storage map indicating the first association between the data storage module and the disk partition and indicating the second association between the data storage module and the tenant. The method may also include generating a unique identifier of the data storage module, wherein the unique identifier of the data storage module is mapped to the tenant in the storage map, and wherein the unique identifier is not mapped to any other tenants in the storage map. In specific implementations, the storage map includes a mapping of the unique identifier of the data storage module to the tenant and a mapping of the unique identifier of the data storage module to the disk partition.

In at least some embodiments, one or more other partitions of the disk are associated with one or more other data storage modules, respectively, and the one or more other data storage modules are associated with one or more other tenants, respectively, of the distributed storage system. The disk partition can include a portion of the disk or all of the disk, and only data of the tenant is to be stored in the disk partition.

In at least some embodiments, the method can further include receiving an indication of a second data storage module to be associated with the tenant of the distributed storage system, allocating a second disk partition for other data of the tenant, creating a third association between the second data storage module and the second disk partition, creating a fourth association between the second data storage module and the tenant, and creating other rules for the second data storage module based on one or more other policies configured for storing other data of the tenant. The data may include one of objects, files or blocks. In at least one embodiment, the distributed storage system is a Ceph storage system.

Some or all of the elements, operations, and features may be included in respective systems, apparatuses, and devices for performing the described functionality. Furthermore, some or all of the features may be implemented in at least one machine readable storage medium.

DESCRIPTION OF EMBODIMENTS

FIG. 1 is a simplified block diagram of an example distributed storage system 100 with an administration host 10 configured to enable tenant-level sharding of disks with one or more unique data storage modules per tenant. Administration host 10 may communicate with storage nodes 60(1)-60(X) of storage cluster 50 over one or more networks such as network 5. In at least one example, administration host 10 can include a policy module 12, a storage map generator 14, and a configuration module 16 to facilitate tenant-level configuration of distributed storage system 100. Administration host 10 also includes at least one processor 19, at least one memory element 17, and other suitable hardware (not shown) to facilitate the operations thereof. A policies repository 34 may be associated with administration host 10, to store tenant-level policies including subscription models associated with tenants. Administration host 10 may include a user interface to enable communication with a user via a user device 40. When storage cluster 50 is configured to enable tenants to store data in the cluster, administration host 10 can generate a storage map 32 including rules and settings based, at least in part, on tenant policies. Storage map 32 can indicate how data is to be stored in and retrieved from storage nodes in storage cluster 50 and can include tenant-specific rules related to storing and retrieving the data. A gateway 80 may receive tenant data 24(1)-24(Y) for distribution and storage across storage nodes 60(1)-60(X) based on storage map 32. In this example, each instance of tenant data, indicated at 24(1)-24(Y), is associated with a single tenant. Some instances of the tenant data 24(1)-24(Y) may be associated with the same tenant, but stored in a different disk partition in storage cluster 50.

Elements of FIG. 1 may be coupled to one another through one or more interfaces employing any suitable connections (wired or wireless), which provide viable pathways for network communications in a network environment. Additionally, one or more of these elements of FIG. 1 may be combined, divided, or removed from the architecture based on particular configuration needs. Distributed storage system 100 may include a configuration capable of transmission control protocol/internet protocol (TCP/IP) communications for the transmission and/or reception of packets in a network. Distributed storage system 100 may also operate in conjunction with a user datagram protocol/IP (UDP/IP), any other suitable protocol, or any suitable combination thereof where appropriate and based on particular needs.

For purposes of illustrating the techniques of administration host 10, it is important to understand the activities that may be present in a distributed storage system. The following foundational information may be viewed as a basis from which the present disclosure may be properly explained. Such information is offered for purposes of explanation only and, accordingly, should not be construed in any way to limit the broad scope of the present disclosure and its potential applications.

Distributed storage systems have emerged to provide a scalable option for cloud storage with greater accessibility and protection of stored data. Object storage involves storing chunks of data in an object, with each object including metadata and a unique identifier. Distributed storage systems can also be applied to other types of data storage such as block storage and file storage, for example. In block storage, data can be stored in blocks (or volumes) where each block acts as an individual hard drive. File storage is generally a hierarchical way of organizing files containing data such that an individual file can be located by a path to that file. Certain metadata describing the file and its contents is also typically stored in the file system. In distributed storage systems, multiple replicas of data in any suitable type of structure (e.g., objects, files, blocks, etc.) can be maintained in order to provide fault tolerance and high availability. Although embodiments may be described herein with reference to objects and distributed object storage, this is done for ease of illustration and it should be understood that these embodiments may also be applicable to other types of data storage structures (e.g., file, block, etc.) and distributed storage including, but not limited to distributed file storage and distributed block storage.

An example distributed storage system that provides high fault tolerance and availability includes Ceph, which is described by Sage A. Weil in the dissertation, “Ceph: Reliable, Scalable, and High-Performance Distributed Storage,” University of California, Santa Cruz, December 2007. Ceph is open source software designed to provide object, block and file storage from a distributed storage cluster. The storage cluster can be comprised of storage nodes with one or more memory elements (e.g., disks) for storing data. Storage nodes are also referred to as object storage devices (OSDs), which can be physical or logical storage elements. In Ceph, storage nodes generally include an object storage device (OSD) software or daemon, which actually stores data as objects on the storage nodes. Ceph OSD software typically stores data on a local filesystem including, but not limited to, a B-tree file system (Btrfs). At least one Ceph metadata server can be provided for a storage cluster to store metadata associated with the objects (e.g., inodes, directories, etc.). Ceph monitors are provided for monitoring active and failed storage nodes in the cluster. It should be understood that embodiments described herein could be implemented in Ceph, or potentially in other distributed storage systems.

A distributed storage system such as Ceph, can provide storage in a storage cluster for data from multiple tenants. Generally, in Ceph, objects from the tenants are pseudo-randomly distributed across the cluster and are monitored by the same set of storage processes (e.g., OSD daemons). Thus, the same global configurations and distribution settings for dividing objects between different nodes may be used to store the objects of the tenants.

In Ceph, how and where to store data in a cluster is determined by a Controlled Replication Under Scalable Hashing (CRUSH) algorithm that computes data storage locations based on a CRUSH map. The CRUSH map identifies information about the storage cluster including the layout and capacity of storage nodes and how redundancy should be managed. More specifically, the map can contain a list of rules that tells CRUSH how it should replicate data in a Ceph cluster's pool. The rules can contain a replication factor for a particular pool of data to help determine how many times the data is to be replicated within the cluster and on which storage nodes the replicated data is to be stored. A pool can comprise a collection of data, such as objects, and a replication factor can be assigned to each pool. Typically, a pool can be shared across tenants.

In a typical Ceph configuration, when multiple tenants use a shared storage cluster, the same set of bucket weights and CRUSH rules apply to all tenants, and objects belonging to different tenants share the same object storage device daemons and disk partitions. In some scenarios, however, a tenant may have different requirements or preferences for its data than other tenants sharing the same storage cluster in a distributed storage system. Such requirements may be related to performance, distribution, and/or replication in at least some scenarios. For example, a tenant may desire certain input/output operations to be satisfied that require storage on a particular type of disk. Another tenant may not need maximum priority for its data. In another example, a tenant may prefer particular storage nodes or disk partitions for its data. In addition, a multi-tenant resource isolation problem may exist when a disk partition contains data from multiple tenants. If the partition fails in this scenario, then all of the tenants sharing the partition could be affected by the failure and subsequent recovery process.

In accordance with at least one embodiment of the present disclosure, the aforementioned issues (and more) associated with existing distributed storage systems can be resolved. Embodiments of distributed storage system 100 enable tenant-specific sharding of storage disks in each storage node 60(1)-60(X) of storage cluster 50. Administration host 10 is configured to enable tenant-specific data storage modules to control data replication processes and primary node selection to provide unique, per tenant behaviors. In particular, one or more tenant-specific disk partitions may be allocated for each tenant sharing the storage cluster. The tenant-specific disk partitions can be controlled by individual per tenant data storage modules. This enables independent configurations per tenant for data storage in the storage cluster. A storage map can be generated with the configurations. In at least one embodiment, independent configurations for a tenant could include rules based, at least in part, on the tenant's subscription model to the distributed storage service. Other independent configurations for a tenant could include rules based on policies configured for differentiated services (e.g., performance, distribution, replication) for the tenant. In at least some embodiments, different sets of independent configurations for the same tenant may be created for different data storage modules controlling the multiple disk partitions. Accordingly, the tenant may be allowed to apply different sets of configurations to different sets of data.

Several advantages can be achieved by offering a distributed storage system that enables unique, per-tenant configurations for tenant data stored in a shared storage cluster. First, differentiated service levels can be provided to tenants based on a subscription model associated with each tenant. The differential behavior enabled for each tenant can provide the benefits of performance isolation between tenants and failure isolation between disk partitions. By enabling tenant-specific partitions per disk with a dedicated data storage module per tenant, a partition failure that occurs for a particular tenant can trigger a recovery process that impacts only that tenant during rebalancing and recreating replicas of data that is lost due to the failure. Also, by enabling differentiated settings for each data storage module per tenant, the performance impacts of each tenant can be isolated. Such performance impacts can occur during cluster rebalancing, failure recovery, and reading/writing data that relies on a replica storage node selection. Also, individual data storage modules per tenant prevent contention of a single data storage module between tenants sharing the data storage module.

It should be noted that, as used herein, ‘tenant’ is intended to refer to an entity (or an identifier or other representation of the entity) that is associated with certain data stored (or allowed to be stored) in a distributed storage system. The association between an entity and the stored data may be in the form of ownership, management, control, etc. of that data, which can include objects, files, blocks, etc. Generally, each object, block or file stored in a distributed storage system is associated with a single tenant. Multiple tenants may have data stored in the distributed storage system.

Turning to the infrastructure of FIG. 1, FIG. 1 is a simplified block diagram of distributed storage system 100, including administration host 10, storage cluster 50, and gateway 80 communicating via network 5. Network 5 represents a series of points or nodes of interconnected communication paths for receiving and transmitting packets of information that propagate through the network environment. Network 5 offers a communicative interface between nodes (e.g., storage nodes 60(1)-60(X)), administration host 10, and gateway 80, and may include any type or topology of one or more networks such as a local area network (LAN), wireless local area network (WLAN), metropolitan area network (MAN), virtual local area network (VLAN), Intranet, Extranet, wide area network (WAN) such as the Internet, virtual private network (VPN), any other appropriate network configuration, or any suitable combination thereof that facilitates communications in a network environment. In at least some embodiments, one or more other elements in the network environment may also communicate via networks such as, for example, those networks described with reference to network 5. For ease of illustration, however, not all elements of FIG. 1 are depicted with communication lines traversing network 5 (e.g., policies repository 34, user device 40, etc.).

In network 5, network traffic, which is inclusive of packets, frames, signals, cells, datagrams, protocol data units (PDUs), data, etc., can be sent and received according to any suitable communication messaging protocols. Suitable communication messaging protocols can include a multi-layered scheme such as Open Systems Interconnection (OSI) model, or any derivations or variants thereof (e.g., Transmission Control Protocol/Internet Protocol (TCP/IP), user datagram protocol/IP (UDP/IP)). A packet is a unit of data for communicating information in a network, and can be routed between a source node (e.g., administration host 10, gateway 80) and a destination node (e.g., storage nodes 60(1)-60(X)) via network 5. A packet includes, but is not limited to, a source network address, a destination network address, and a payload containing the information to be communicated. By way of example, these network addresses can be Internet Protocol (IP) addresses in a TCP/IP messaging protocol. Information is generally represented by data and, as used herein, ‘data’ refers to any type of binary, numeric, voice, video, media, textual, or script data, or any type of source or object code, or any other suitable information in any appropriate format that may be communicated from one point to another in electronic devices and/or networks.

Administration host 10 and gateway 80 can be implemented as one or more network elements in a network environment. As used herein, the term ‘network element’ is meant to encompass servers, processors, modules, routers, switches, gateways, bridges, load balancers, firewalls, inline service nodes, proxies, or any other suitable device, component, element, proprietary appliance, or object operable to exchange information in a network environment. This network element may include any suitable hardware, software, components, modules, interfaces, or objects that facilitate the operations thereof. This may be inclusive of appropriate algorithms and communication protocols that allow for the effective exchange of data or information.

Storage nodes 60(1)-60(X) are network elements that include physical or logical storage elements with one or more disks for storing electronic data. In embodiments disclosed herein, tenant data is stored in storage nodes 60(1)-60(X). When the data is stored as objects, each object may have a unique identifier and associated metadata. Data storage modules may be provided in each storage node to determine storage locations for the data, to store the data, and to provide access to the data over the network. Data in storage nodes 60(1)-60(X) can be accessed by clients (not shown) via gateway 80 by an application programming interface (API) or hypertext transfer protocol (HTTP), for example. Clients can enable users, including human users and/or applications, to access the stored data.

In one implementation, network elements of FIG. 1, such as administration host 10, storage nodes 60(1)-60(X), and gateway 80, include software to achieve (or to foster) tenant-level configuration operations for distributed storage system 100, as outlined herein. Note that in one example, the network elements of FIG. 1 can have internal structures (e.g., processor 19, memory element 17, network interface card, etc.) to facilitate some of the operations described herein. In other embodiments, these tenant-level configuration activities may be executed externally to these network elements, or included in some other network element to achieve this intended functionality. Alternatively, the network elements of FIG. 1 may include this software (or reciprocating software) that can coordinate with other network elements in order to achieve the operations, as outlined herein. In still other embodiments, one or several devices may include any suitable algorithms, hardware, software, firmware, components, modules, interfaces, or objects that facilitate the operations thereof.

As shown in FIG. 1, policy module 12 may be provided in administration host 10. Policy module 12 may provide or interact with a user interface to enable a user to configure, delete, update/modify, and access policies for each tenant. A user may access administration host 10 via a user device, such as user device 40. Policies that can be configured for a tenant can include a subscription model (e.g., specifying partition size for tenant, priority of tenant relative to other tenants, etc.). By way of illustration, a ‘Gold’ subscription model could allow a partition size of 1 terabyte (TB), and a ‘Silver’ subscription model could allow a partition size of 100 gigabytes (GB).

Other tenant-level policies can be configured to enable differentiated services for the tenant. For example, such policies can include, but are not limited to, performance requirements, distribution requirements, and replication requirements. Performance requirements can be based on some performance characteristic of a disk such as speed of the disk, input/output rate of the disk, etc. Distribution requirements can be based on where data is to be stored such as a particular disk, a particular rack, a particular location, etc. Distribution requirements can also include how far apart or how close together replicas can be relative to each other. Replication requirements can be based on the number of replicas desired for the data of a tenant. By way of illustration, one tenant may have strict input/output performance requirements and thus, may choose a policy to ensure that most of its data and the primary replica nodes are in solid state device (SSD) disks. A different tenant may choose a different policy if that tenant does not have the same need for SSD-like throughputs.

In at least one embodiment, policies may be stored in policies repository 34. Policies repository 34 may be provided in any suitable type of storage, which may be internal to administration host 10 or external (entirely or in part). Internal storage could include any internal memory of administration host 10, such as static storage, random access memory (RAM), or cache, for example. External storage could include a network storage technique such as network attached storage (NAS) or storage area network (SAN), or memory of another network element.

Configuration module 16 may be provided in administration host 10 to enable per tenant configurations. In at least one embodiment, configuration module 16 and a user interface can enable a user to configure storage cluster 50 with individual per tenant data storage modules for tenant-specific disk partitions in storage nodes 60(1)-60(X). When a new data storage module is added for a tenant, a unique identifier may be generated for the data storage module and associated with the tenant. Per tenant configuration of a storage node is described in more detail with reference to FIG. 2.

Turning to FIG. 2, FIG. 2 is a simplified block diagram of an example configuration of storage node 60(1). Storage node 60(1) is depicted with a single disk 20 containing partitions 22(1)-22(Y). In this example, tenant data 24(1)-24(Y) is stored in respective partitions 22(1)-22(Y). Data storage modules 70(1)-70(Y) have respective unique identifiers 75(1)-75(Y), and are associated with respective partitions 22(1)-22(Y). Each data storage module 70(1)-70(Y) and its respective partition 22(1)-22(Y) are associated with a single tenant 26(1)-26(N). Storage node 60(1) also includes at least one processor 69 and at least one memory element 67, and other suitable hardware (not shown) to facilitate the operations thereof. For ease of illustration and explanation, storage node 60(1) is shown with single disk 20 in this example configuration. It should be appreciated, however, that storage nodes may contain more than one disk, and any one or more disks in a storage node may be partitioned one or more times to achieve tenant-level sharding according to embodiments described herein.

Embodiments disclosed herein allow for multiple data storage modules per disk, as shown by data storage modules 70(1)-70(Y) of disk 20. At least one dedicated data storage module may be provided per tenant. For example, each data storage module of data storage modules 70(1)-70(Y) is associated with a single tenant. In some instances, more than one of the data storage modules 70(1)-70(Y) may be associated with the same tenant and assigned to different disk partitions within the same disk or across multiple disks. In at least one embodiment, however, none of these data storage modules is to be associated with multiple tenants.

For illustration purposes, assume N=3 such that three tenants store data in storage node 60(1), and Y=5 such that five data storage modules are configured in storage node 60(1). In this scenario, one possible result includes data storage modules 75(1) and 75(3) associated with tenant A 26(1), data storage modules 75(2) and 75(4) associated with tenant B 26(2), and data storage module 70(5) associated with tenant C 26(3). In addition, each data storage module 70(1)-70(5) could be assigned to a respective dedicated disk partition 22(1)-22(5).

Embodiments disclosed herein also enable tenant-level sharding. In the example shown in FIG. 2, data storage module 70(1) is associated with tenant A 26(1) and assigned to partition 22(1), data storage module 70(2) is associated with tenant B 26(2) and assigned to partition 22(2), and data storage module 70(Y) is associated with tenant C 26(N) and assigned to partition 22(Y). Thus, tenant data 24(1) belongs to tenant A 26(1) and is stored in partition 22(1), tenant data 24(2) belongs to tenant B 26(2) and is stored in partition 22(2), and tenant data 24(Y) belongs to tenant C 26(N) and is stored in partition 22(Y). In at least one embodiment, data storage modules 70(1)-70(Y) may be configured as daemons that run as background processes to determine where to store data and to interact with logical disks to store and access data.

Embodiments also allow a single disk to be shared by a fewer number of tenants. This may occur, for example, if one or more tenants require a significant amount of storage. In certain cases, an entire disk (e.g., 1 TB) may be allocated for use by a single tenant. Having a disk shared by a fewer number of tenants can minimize the risk of interruption by other tenants.

In at least one embodiment, tenant-level disk partitions may be selected manually or automatically. Configuration module 16 may be provided in administration host 10 to enable the selection. For a manual selection, a user may add a data storage module for a particular tenant and then manually select a particular disk partition to be allocated for the data storage module. Alternatively, the disk partition may be pre-determined based on policies. In this case, after adding a data storage module for a particular tenant, a disk partition may be automatically selected and allocated for the data storage module. In this scenario, the disk partition may be selected based on policies, such as the tenant's subscription model and/or tenant-specific policies to enable differentiated services for the tenant.

For illustration purposes, assume a subscription model specifies 1 TB of storage for tenant A, and a tenant-specific policy requires SSD disks for tenant A's data. In this example scenario, 1 TB of available space on a SSD disk in a storage node of the cluster may be automatically identified and allocated for the data storage module associated with tenant A. In another illustration with reference to FIG. 2, if tenant A 26(1) has a ‘Gold’ subscription model, then partition 22(1) may have a size of 1 TB. If tenants B 26(2) and C 26(N) have a ‘Silver’ subscription model, then partitions 22(2) and 22(Y) may each have a size of 100 GB. The number of data storage modules in a single disk can be dependent, at least in part, on the size of the disk and the partition size required by each tenant for its data.

In at least one embodiment, an association is created between a data storage module and a disk partition that is selected and allocated for the data storage module. An association is also created between the data storage module and the tenant for which the data storage module was created. In addition, one or more rules for the tenant may be created based on policies configured for the tenant (e.g., subscription model, performance requirements, distribution requirements, replication factor, etc.). The rules may be associated with the tenant and the data storage module associated with that tenant. These associations may be realized in any suitable manner including, but not limited to, mapping a unique identifier of the data storage module to suitable identifiers or other representations of the disk partition, the tenant, and/or the rules.

In at least one embodiment, these mappings can be provided in storage map 32. Storage map generator 14 may be provided in administration host 10 to generate storage map 32. Storage map 32 can be used by data storage modules, including data storage modules 70(1)-70(Y), to determine how to store and retrieve data in a storage cluster such as storage cluster 50. In at least one embodiment, storage map 32 is a map of storage cluster 50, including at least a list of tenant-specific data storage modules (e.g., using their unique identifiers) mapped to associated tenants, allocated disk partitions, and sets of rules generated for the associated tenants. Because each data storage module is dedicated to a single tenant, the tenant can decide what policies to configure so that the rules that are generated enable a desired data distribution in the storage cluster.

At least one embodiment allows for a user to configure multiple sets of policies for a single tenant. Thus, multiple sets of rules can be generated for different data storage modules of the same tenant. For example, assume first and second data storage modules are associated with tenant A, and first and second rule sets are also associated with tenant A. In one possible scenario, the first rule set could be associated with the first data storage module and the second rule set could be associated with the second data storage module. Accordingly, the different data storage modules can be used for different types of data of tenant A. For example, the first and second data storage modules could be assigned to different types of disks. The first rule set could include a rule requiring an SSD disk partition, and the second rule set may not specify a particular type of disk and may rely on default settings or criteria. In this example scenario, critical data could be stored using the first data storage module (i.e., on an SSD disk) and less critical data could be stored using the second data storage module.

In one example implementation using a distributed storage system such as Ceph, an embodiment as described herein can allow for pools, which are logical groups for storing data in a storage cluster, to have a one-to-one correspondence to tenants. Users associated with a particular tenant may be authorized to access only pools corresponding to that particular tenant. The tenant's data to be added to the storage cluster is to be stored in the pool corresponding to the tenant. Rules that are created from the policies can be written for the pool belonging to the tenant, which uses one or more data storage modules that are only mapped to that tenant. The rules can be provided in the storage map and can be used by the data storage modules to determine a primary storage location for the data and its replicas in the storage cluster.

Turning to FIG. 3, FIG. 3 is a simplified flowchart 300 illustrating potential operations that may be associated with embodiments described herein. In at least one embodiment, one or more sets of operations correspond to activities of FIG. 3. In at least one embodiment, administration host 10 may comprise means such as one or more processors (e.g., processor 19), for performing the operations. In one example, at least some operations shown in flowchart 300 may be performed by at least one of policy module 12, storage map module 14, and configuration module 16 when executed by one or more processors such as processor 19. For ease of reference, operations of flowchart 300 may be described without specific reference to particular modules of administration host 10.

At 302, administration host 10 receives an indication of a type of subscription model for a tenant. The type of the subscription model may be selected for the tenant by a user via a user interface. The subscription model may specify a partition size (e.g., 100 GB, 500 GB, 1 TB, etc.) desired by the tenant for storing its data. The subscription model may also specify a priority relative to other subscription types. Priority could be used, for example, to resolve contention between data storage modules accessing the same disk.

At 304, administration host 10 receives an indication of policies to be applied to data of the tenant. The policies may be configured by the user via a user interface. Policies may include, for example, performance requirements, distribution requirements, replication factor, etc. preferred by the tenant. In some embodiments, any of the performance, distribution and/or replication requirements may be included in a subscription model rather than being configured separately. At 306, administration host 10 receives an indication of a data storage module to be assigned to the tenant. The data storage module may be assigned to the tenant by the user via a user interface, and may be dedicated to that tenant. At 308, a unique identifier (UID) may be generated for the data storage module assigned to the tenant.

A disk partition for the data storage module may be selected manually or automatically. For manual selection at 312, administration host 10 can receive an indication of a particular disk partition (or an entire disk) to be allocated for the tenant. The particular disk partition may be selected by the user via the user interface. At 314, the selected disk partition may be allocated for the tenant. For automatic selection of a disk partition (or entire disk), at 310, administration host 10 can identify and allocate a disk partition (or an entire disk) from available storage nodes in a storage cluster based on policies configured for the tenant and/or the subscription model assigned to the tenant.

At 316, an association is created between the data storage module and the allocated disk partition. In at least one embodiment, this association may be realized by mapping the UID of the data storage module to the disk partition. In an example, a suitable identifier or other representation of the disk partition (or disk) may be used for the mapping. At 318, an association is created between the data storage module and the tenant. In at least one embodiment, this association may be realized by mapping the UID of the data storage module to the tenant. In an example, a suitable identifier or other representation of the tenant may be used for the mapping. At 320, one or more rules can be created based on policies configured for the tenant (e.g., replication factor, subscription model, performance requirements, distribution requirements, etc.). The rules may be associated with the data storage module that is associated with the tenant. This association may be realized by mapping the UID of the data storage module to the rules. In an example, a suitable identifier or other representation of the rules may be used for the mapping. In at least one embodiment, these mappings can be provided in a storage map used by data storage module to determine how to store and retrieve data in the storage cluster. In addition, the rules created from the policies may also be provided in the storage map in at least one embodiment.

Turning to FIG. 4, FIG. 4 is a simplified flowchart illustrating potential operations that may be associated with embodiments described herein. In at least one embodiment, one or more sets of operations correspond to activities of FIG. 4. In at least one embodiment, gateway 80 and a storage node (e.g., storage nodes 60(1)-60(X)) may comprise means such as one or more processors (e.g., processor 69), for performing the operations.

Initially, an authorized user of a particular tenant may access gateway 80 in order to add data to the storage cluster. In at least one implementation (e.g., in Ceph), data is added to a pool corresponding to the tenant. At 404, gateway 80 may receive a request for data of the tenant to be stored in storage cluster 50 of distributed storage system 100. In at least one embodiment, the request may be an indication that the authorized user (e.g., human user or application) has stored objects or other data in a pool corresponding to the tenant. At 406, the tenant associated with the data can be identified based on the pool in which the data is stored. At 408, a data storage module associated with the tenant is identified. This identification may be made based on a mapping of a unique identifier of the data storage module to the tenant.

At 410, the identified data storage module (or modules) associated with the tenant can be run to determine how and where to store the data based on a storage map. Rules associated with the data storage module can be determined from the storage map and used to calculate how and where to store the data (e.g., which disk partition to use, how many replicas to store, where to store the replicas, etc.). Thus, tenant-specific data storage modules can control the primary node selection and the data replication process, which enables unique tenant behaviors configured by policies. Moreover, the dedicated, tenant-specific disk partitions enable failure and performance isolation relative to other tenants and their dedicated disk partitions.

Variations and Implementations

Note that, as used herein, unless expressly stated to the contrary, use of the phrase ‘at least one of’ refers to any combination of the named items, elements, conditions, activities, etc. For example, ‘at least one of X, Y, and Z’ is intended to mean any of the following: 1) one or more X's, but not Y and not Z; 2) one or more Y's, but not X and not Z; 3) one or more Z's, but not X and not Y; 4) one or more X's and one or more Y's, but not Z; 5) one or more X's and one or more Z's, but not Y; 6) one or more Y's and one or more Z's, but not X; or 7) one or more X's, one or more Y's, and one or more Z's.

In certain example implementations the tenant-level configuration functions for a distributed storage system outlined herein may be implemented by logic encoded in one or more machine readable storage media (e.g., embedded logic provided in an application specific integrated circuit (ASIC), digital signal processor (DSP) instructions, software (potentially inclusive of object code and source code) to be executed by a processor or other similar machine, etc.). In some of these instances, a memory element (e.g., memory elements 17, 67, a memory element of gateway 80) can store data used for the operations described herein. This includes the memory element being able to store software, logic, code, or processor instructions that are executed to carry out the activities described in this Specification. A processor can execute any type of instructions associated with the data to achieve the operations detailed herein. In one example, the processor (e.g., processors 19, 69, a processor of gateway 80) could transform an element or an article (e.g., data) from one state or thing to another state or thing. In another example, the activities outlined herein may be implemented with fixed logic or programmable logic (e.g., software/computer instructions executed by a processor) and the elements identified herein could be some type of a programmable processor, programmable digital logic (e.g., a field programmable gate array (FPGA), an erasable programmable read only memory (EPROM), an electrically erasable programmable ROM (EEPROM)) or an ASIC that includes digital logic, software, code, electronic instructions, or any suitable combination thereof.

In one example implementation, administration host 10 may include software in order to achieve at least some of the tenant-level configuration functions outlined herein. These activities can be facilitated by policy module 12, storage map generator 14, and configuration module 16 (where the functionality of these modules can be suitably combined or divided in any appropriate manner, which may be based on particular configuration and/or provisioning needs). Administration host 10 can include memory elements (e.g., memory element 17) for storing information to be used in achieving at least some of the tenant-level configuration activities, as discussed herein. Additionally, administration host 10 may include one or more processors (e.g., processor 19) that can execute software or an algorithm to perform the tenant-level configuration operations, as disclosed in this Specification. These devices may further keep information in any suitable memory elements (e.g., random access memory (RAM), ROM, EPROM, EEPROM, ASIC, etc.), software, hardware, or in any other suitable component, device, element, or object where appropriate and based on particular needs. Any of the memory items discussed herein (e.g., object, block, file, database, tables, trees, cache, repository, etc.) should be construed as being encompassed within the broad term ‘memory element.’ Similarly, any of the potential processing elements, modules, and machines described in this Specification should be construed as being encompassed within the broad term ‘processor.’ Administration host 10 can also include suitable interfaces (e.g., network interface card) for receiving, transmitting, and/or otherwise communicating data or information in distributed storage system 100.

Note that with the example provided above, as well as numerous other examples provided herein, interaction may be described in terms of two, three, or four network elements. However, this has been done for purposes of clarity and example only. In certain cases, it may be easier to describe one or more of the functionalities of a given set of operations by only referencing a limited number of network elements and nodes. It should be appreciated that distributed storage system 100 (and its teachings) is readily scalable and can accommodate a large number of components, as well as more complicated/sophisticated arrangements and configurations. Accordingly, the examples provided should not limit the scope or inhibit the broad teachings of distributed storage system 100 as potentially applied to a myriad of other architectures.

Although the present disclosure has been described in detail with reference to particular arrangements and configurations, these example configurations and arrangements may be changed significantly without departing from the scope of the present disclosure. For example, although the present disclosure has been described with reference to particular tenant-level configuration functions (e.g., applied in a Ceph storage system), these tenant-level configuration functions may be applicable in other distributed storage systems. Also, while the tenant-level configuration functions are particularly suited to distributed storage systems that store data in the form of objects, the teachings herein may also be applied to distributed storage systems that store data in various other types of structures including, but not limited to, files and blocks.

Finally, it is also important to note that the operations in the preceding flowcharts illustrate only some of the possible scenarios and patterns that may be executed in association with addressing tenant configuration operations in a distributed storage system. Some of these operations may be deleted, removed, combined, or divided where appropriate, or may be modified or changed considerably without departing from the scope of the present disclosure. In addition, a number of these operations have been described as being executed before, after, concurrently with, or in parallel to, one or more additional operations. However, the timing of these operations may be altered considerably. The preceding operational flows have been offered for purposes of example and discussion. Distributed storage system 100, including administration host 10, may provide substantial flexibility in that any suitable arrangements, chronologies, configurations, and timing mechanisms may be provided without departing from the teachings of the present disclosure.

Claims

1. A method comprising: receiving an indication of one or more data storage modules to be assigned to only one tenant of a plurality of tenants of a distributed storage system;selecting a partition of a disk to be dedicated to the tenant, the selecting based, at least in part, on one or more tenant-specific policies of the tenant, the one or more tenant-specific policies related to at least one requirement of the tenant including at least one of a performance requirement of disks that store data of the tenant, a distribution requirement for the data of the tenant, or a replication requirement for the data of the tenant;allocating the partition for the data of the tenant;assigning a first data storage module of the one or more data storage modules to the partition;assigning the first data storage module to the tenant; andcreating rules for the first data storage module based on the one or more tenant-specific policies configured for the tenant, wherein the rules indicate how the first data storage module is to store the data of the tenant in the partition and how the first data storage module is to access stored data in the partition,wherein, the one or more tenant-specific policies are related to the distribution requirement for h data and the replication requirement for the data,the replication requirement is a number of replicas desired for a portion of the data, andthe distribution requirement is how far apart or how close together the replicas can be stored relative to each other.

2. The method of claim 1, further comprising: receiving an indication of a type of subscription model selected for the tenant.

3. The method of claim 2, wherein the subscription model selected for the tenant includes a tenant-specific policy of the one or more tenant-specific policies used to select the partition.

4. The method of claim 1, wherein the first data storage module stores at least some data of the tenant in the partition based, at least in part, on the rules.

5. The method of claim 1, further comprising: generating a storage map indicating the assignment of the first data storage module to the partition and indicating the assignment of the data storage module to the tenant.

6. The method of claim 5, further comprising: generating a unique identifier of the data storage module, wherein the unique identifier of the data storage module is mapped to the tenant in the storage map, and wherein the unique identifier is not mapped to any other tenants in the storage map.

7. The method of claim 5, wherein the storage map includes: a mapping of a unique identifier of the first data storage module to the tenant; and a mapping of the unique identifier of the first data storage module to the partition.

8. The method of claim 1, wherein one or more other partitions of the disk are associated with one or more other data storage modules, respectively, and wherein the one or more other data storage modules are associated with one or more other tenants, respectively, of the distributed storage system.

9. The method of claim 1, wherein, the distribution requirement is how close together the replicas can be stored relative to each other.

10. The method claim 1, further comprising: selecting a second partition of the disk to be dedicated to the tenant;allocating the second partition for other data of the tenant;assigning a second data storage module of the one or more data storage modules to the second partition;assigning the second data storage module to the tenant; andcreating other rules for the second data storage module based on one or more other tenant-specific policies configured for storing the other data of the tenant.

11. At least one non-transitory machine readable storage medium comprising instructions that, when executed by at least one processor, cause the at least one processor to: receive an indication of one or more data storage modules to be assigned to only one tenant of a plurality of tenants of a distributed storage system;select a partition of a disk to be dedicated to the tenant based, at least in part, on one or more tenant-specific policies of the tenant, the one or more tenant-specific policies related to at least one requirement of the tenant including at least one of a performance requirement of disks that store data of the tenant, a distribution requirement for the data of the tenant, or a replication requirement for the data of the tenant;allocate the partition for the data of the tenant;assign a first data storage module of the one or more data storage modules to the partition;assigning the first data storage module to the tenant; andcreate rules for the first data storage module based on the one or more tenant-specific policies configured for the tenant, wherein the rules indicate how the first data storage module is to store the data of the tenant in the partition and how the first data storage module is to access stored data in the partition,wherein, the one or more tenant-specific policies are related to the distribution requirement for the data and the replication requirement for the data,the replication requirement is a number of replicas desired for a portion of the data, andthe distribution requirement is how far apart or how close together the replicas can be stored relative to each other.

12. The at least one non-transitory machine readable storage medium of claim 11, wherein the instructions when executed by the at least one processor cause the at least one processor to:receive an indication of a type of subscription model selected for the tenant.

13. The at least one non-transitory machine readable storage medium of claim 12, wherein the subscription model selected for the tenant includes a tenant-specific policy of the one or more tenant-specific policies used to select the partition.

14. The at least one non-transitory machine readable storage medium of claim 11, wherein the first data storage module stores at least some data of the tenant in the partition based, at least in part, on the rules.

15. The at least one non-transitory machine readable storage medium of claim 11, wherein the instructions when executed by the at least one processor cause the at least one processor to:generate a storage map indicating the assignment of the first data storage module to the partition and indicating the assignment of the first data storage module to the tenant.

16. An apparatus comprising: at least one processor; andat least one memory element comprising instructions that when executed by the at least one processor perform operations comprising:receiving an indication of one or more data storage modules to be assigned to only one tenant of a plurality of tenants of a distributed storage system;selecting a partition of a disk to be dedicated to the tenant, the selecting based, at least in part, on one or more tenant-specific policies of the tenant, the one or more tenant-specific policies related to at least one requirement of the tenant including at least one of a performance requirement of disks that store data of the tenant, a distribution requirement for the data of the tenant, and a replication requirement for the data of the tenant;allocating the partition of the disk for the data of the tenant;assigning a first data storage module of the one or more data storage modules to the partition;assigning the first data storage module to the tenant; andcreating rules for the first data storage module based on the one or more tenant-specific policies configured for the tenant, wherein the rules indicate how the first data storage module is to store the data of the tenant in the partition and how the first data storage module is to access stored data in the partition,wherein, the one or more tenant-specific policies are related to the distribution requirement for the data and the replication requirement for the data,the replication requirement is a number of replicas desired for a portion of the data, andthe distribution requirement is how far apart or how close together the replicas can be stored relative to each other.

17. The apparatus of claim 16, wherein the instructions when executed by the at least one processor cause the apparatus to: generating a unique identifier of the first data storage module, wherein the unique identifier of the first data storage module is mapped to the tenant in a storage map, and wherein the unique identifier is not mapped to any other tenants in the storage map.

18. The apparatus of claim 16, wherein the data includes one of objects, files or blocks.

19. The apparatus of claim 16, wherein one or more other data storage modules are assigned to one or more other partitions of the disk, respectively, and wherein the one or more other data storage modules are assigned to one or more other tenants, respectively, of the distributed storage system.

20. The apparatus of claim 16, wherein the distributed storage system includes a storage cluster with at least one metadata server configured to store metadata associated with objects.