Indirect Dataset Replication To Cloud-Based Targets
Systems, methods, and computer readable storage mediums for performing remote replication including receiving, by a target storage system, an indication that a dataset stored on a source storage system will be replicated to the target storage system; identifying, by the target storage system, portions of the dataset that are not already stored on the target storage system; and initiating, by the target storage system, replication of one or more of the portions of the dataset that are not already stored on the target storage system from a source other than the source storage system.
This application is a continuation application for patent entitled to a filing date and claiming the benefit of U.S. Pat. No. 11,775,392, issued Oct. 3, 2023, herein incorporated by reference in its entirety, which is a continuation of U.S. Pat. No. 11,061,786, issued Jul. 13, 2021, which is a continuation application of and claims priority from U.S. Pat. No. 10,235,065, issued Mar. 19, 2019, which is a continuation application of and claims priority from U.S. Pat. No. 9,552,248, issued Jan. 24, 2017.
BACKGROUND Technical FieldEmbodiments described herein relate to storage systems, and more particularly, to techniques for monitoring storage subsystems in a replication environment.
Description of the Related ArtBusiness continuity and remote data protection are of paramount concern for enterprises. Unfortunately, most data protection and disaster recovery (DR) solutions require complex planning and significant investment to achieve the desired recovery point objective (RPO). More importantly, even if desired low RPO is achieved, the recovery times are much higher without making significant investments in DR solutions. Also, many DR solutions require DR site infrastructure to be identical to the production site, making the DR configurations restrictive. As a result of complexity, inflexibility, and economics, DR solutions are applied to only select few mission-critical applications or forgone completely.
SUMMARYVarious embodiments of systems and methods for generating an alert on failure to phone home in a replication environment are contemplated.
A storage system may include a plurality of storage subsystems, with each storage subsystem including a storage controller and one or more storage devices. In one embodiment, a first storage subsystem may replicate a dataset to a second storage subsystem. The dataset may include any type of data, such as one or more volumes, virtual machines, files, protection groups, disk images, databases, applications, and/or other collections of data. The first storage subsystem may be configured to phone home log data to a cloud-based service on a regularly scheduled basis. The log data may include performance data, capacity data, system health data, diagnostics, logs, and/or other data. If the cloud service detects that the first storage subsystem has not phoned home as expected, the cloud service may generate a first alert to notify the second storage subsystem. In response to receiving the first alert, the second storage subsystem may begin disaster recovery operations for the dataset.
In another embodiment, the cloud service may analyze the log data generated by the first storage subsystem to determine a health rating for the first storage subsystem. If the cloud service detects that the health rating of the first storage subsystem has declined below a programmable threshold, then the cloud service may generate a second alert to notify the second storage subsystem. In further embodiments, the cloud service may generate other types of alerts for notifying the second storage subsystem based on analysis of the first storage subsystem's log data.
In one embodiment, the cloud service may be configured to select a third storage subsystem to take over as the new secondary storage subsystem when the current primary storage subsystem fails and the current secondary storage subsystem takes over as the new primary storage subsystem for a first dataset. The cloud service may be configured to generate a medium graph for the first dataset to identify all of the mediums which underlie the first dataset. The cloud service may translate local medium IDs from the medium graph to global medium IDs using a global to local medium ID mapping table. The cloud service may determine which storage subsystem stores the highest medium ID of the first dataset's medium graph, and then the cloud service may select this storage subsystem to be the new secondary storage subsystem. The cloud service may then determine which mediums of the first dataset are missing on the new secondary storage subsystem. The cloud service may be configured to cause these missing mediums to be replicated to the new secondary storage subsystem.
These and other embodiments will become apparent upon consideration of the following description and accompanying drawings.
While the methods and mechanisms described herein are susceptible to various modifications and alternative forms, specific embodiments are shown by way of example in the drawings and are herein described in detail. It should be understood, however, that drawings and detailed description thereto are not intended to limit the methods and mechanisms to the particular form disclosed, but on the contrary, are intended to cover all modifications, equivalents and alternatives apparent to those skilled in the art once the disclosure is fully appreciated.
DETAILED DESCRIPTIONIn the following description, numerous specific details are set forth to provide a thorough understanding of the methods and mechanisms presented herein. However, one having ordinary skill in the art should recognize that the various embodiments may be practiced without these specific details. In some instances, well-known structures, components, signals, computer program instructions, and techniques have not been shown in detail to avoid obscuring the approaches described herein. It will be appreciated that for simplicity and clarity of illustration, elements shown in the figures have not necessarily been drawn to scale. For example, the dimensions of some of the elements may be exaggerated relative to other elements.
This specification includes references to “one embodiment”. The appearance of the phrase “in one embodiment” in different contexts does not necessarily refer to the same embodiment. Particular features, structures, or characteristics may be combined in any suitable manner consistent with this disclosure. Furthermore, as used throughout this application, the word “may” is used in a permissive sense (i.e., meaning having the potential to), rather than the mandatory sense (i.e., meaning must). Similarly, the words “include”, “including”, and “includes” mean including, but not limited to.
Terminology. The following paragraphs provide definitions and/or context for terms found in this disclosure (including the appended claims):
“Comprising.” This term is open-ended. As used in the appended claims, this term does not foreclose additional structure or steps. Consider a claim that recites: “A system comprising a storage subsystem. . . .” Such a claim does not foreclose the system from including additional components (e.g., a network, a server, a display device).
“Configured To.” Various units, circuits, or other components may be described or claimed as “configured to” perform a task or tasks. In such contexts, “configured to” is used to connote structure by indicating that the units/circuits/components include structure (e.g., circuitry) that performs the task or tasks during operation. As such, the unit/circuit/component can be said to be configured to perform the task even when the specified unit/circuit/component is not currently operational (e.g., is not on). The units/circuits/components used with the “configured to” language include hardware—for example, circuits, memory storing program instructions executable to implement the operation, etc. Reciting that a unit/circuit/component is “configured to” perform one or more tasks is expressly intended not to invoke 35 U.S.C. § 112, paragraph (f), for that unit/circuit/component. Additionally, “configured to” can include generic structure (e.g., generic circuitry) that is manipulated by software and/or firmware (e.g., an FPGA or a general-purpose processor executing software) to operate in a manner that is capable of performing the task(s) at issue. “Configured to” may also include adapting a manufacturing process (e.g., a semiconductor fabrication facility) to fabricate devices (e.g., integrated circuits) that are adapted to implement or perform one or more tasks.
“Based On.” As used herein, this term is used to describe one or more factors that affect a determination. This term does not foreclose additional factors that may affect a determination. That is, a determination may be solely based on those factors or based, at least in part, on those factors. Consider the phrase “determine A based on B.” While B may be a factor that affects the determination of A, such a phrase does not foreclose the determination of A from also being based on C. In other instances, A may be determined based solely on B.
Referring now to
For the purposes of this discussion, original storage array 105 represents the array on which a given dataset is being utilized and/or modified by a client application. Replica storage array 160 may represent the array to which the given dataset is being replicated. Source storage array 170 may represent an array containing portions of the given dataset to be replicated and from which replica storage array 160 is pulling missing data necessary for the given dataset. It is noted that these designations of the various storage arrays are used in the context of a given replication operation. For subsequent replication operations, these designations may change. For example, a first dataset may be replicated from original storage array 105 to replica storage array 160 at a particular point in time. At a later point in time, a second dataset may be replicated from replica storage array 160 to original storage array 105. For the replication of the second dataset, replica storage array 160 may be referred to as an “original” storage array while original storage array 105 may be referred to as a “replica” storage array. Also, the source storage system and the original storage system may be the same for a given replication event. In other words, replica storage array 160 could pull data to replicate a dataset from original storage array 105 directly if it chooses.
Storage system 100 also includes networks 120 and 150, cloud service 180, server 115, and clients 125 and 175. Server 115 is representative of any number and type (e.g., file server, application server, block server, database server) of servers which may be coupled to original storage array 105. Server 115 may be configured to enable storage and retrieval of data from original storage array 105 by client 125 and one or more other clients (not shown). Additionally, any number and type of virtual servers may be hosted by server 115, depending on the embodiment. Although not shown in
As shown, storage device group 130 includes storage devices 135A-N, which are representative of any number and type of storage devices (e.g., solid-state drives (SSDs), Peripheral Component Interconnect Express (PCIe) cards). Storage controller 110 may be coupled to client computer system 125 via server 115, and storage controller 110 may be coupled remotely over network 120 to client computer system 175. Clients 125 and 175 are representative of any number of clients which may utilize storage controller 110 for storing and accessing data in system 100. It is noted that some systems may include only a single client, connected directly or remotely to storage controller 110. It is also noted that original storage array 105 may include more than one storage controller in some embodiments.
Storage controller 110 may include software and/or hardware configured to provide access to storage devices 135A-N. Although storage controller 110 is shown as being separate from storage device groups 130 and 140, in some embodiments, storage controller 110 may be located within one or each of storage device groups 130 and 140. Storage controller 110 may include or be coupled to a base operating system (OS), a volume manager, and additional control logic for implementing the various techniques disclosed herein.
Storage controller 110 may include and/or execute on any number of processors and may include and/or execute on a single host computing device or be spread across multiple host computing devices, depending on the embodiment. In some embodiments, storage controller 110 may generally include or execute on one or more file servers and/or block servers. Storage controller 110 may use any of various techniques for storing data across devices 135A-N to prevent loss of data due to the failure of a device or the failure of storage locations within a device. Storage controller 110 may also utilize any of various deduplication and compression techniques for reducing the amount of data stored in devices 135A-N.
In various embodiments, cloud service 180 may include program instructions which when executed by a processor are configured to perform a variety of tasks related to the replication of data within storage system 100. Cloud service 180 may be configured to execute on a server, computer, or other computing device to perform the functions described herein. In some embodiments, cloud service 180 may include hardware and/or control logic configured to perform the functions and tasks described herein. For example, cloud service 180 may be implemented using any combination of dedicated hardware (e.g., application specific integrated circuit (ASIC)), configurable hardware (e.g., field programmable gate array (FPGA)), and/or software (e.g., program instructions) executing on one or more processors. It is noted that cloud service 180 may also be referred to as cloud-based service 180 or cloud assist service 180.
In one embodiment, cloud service 180 may execute within a cloud computing platform provided by a web services provider (e.g., Amazon). The cloud computing platform may provide large amounts of computing assets and storage availability to cloud service 180. In another embodiment, cloud service 180 may execute on a separate system or network external to the local network of original storage array 105, wherein cloud service 180 may be described as executing on or residing in a private cloud.
Each of original storage array 105, replica storage array 160, and source storage array 170 may be configured to generate and send log data and performance-related data to cloud service 180 for analysis. Cloud service 180 may analyze the log data and performance data and generate alerts to send to the arrays based on the analysis. In various embodiments, the log data and performance data may include capacity data, system health data, logs, diagnostics, past alerts, data associated with replication events, latency data for one or more storage devices of the array, storage capacity utilization of the one or more storage devices, a number of program-erase cycles for the one or more storage devices, an age of the one or more storage devices, and/or other metrics.
In various embodiments, multiple mapping tables may be maintained by storage controller 110. These mapping tables may include an address translation table, a deduplication table, an overlay table, and/or other tables. The address translation table may include a plurality of entries, with each entry holding a virtual-to-physical mapping for a corresponding data component. This mapping table may be used to map logical read/write requests from each of the client computer systems 125 and 175 to physical locations in storage devices 135A-N. A “physical” pointer value may be read from the mappings associated with a given dataset or snapshot during a lookup operation corresponding to a received read/write request. This physical pointer value may then be used to locate a storage location within the storage devices 135A-N. It is noted that the physical pointer value may not be direct. Rather, the pointer may point to another pointer, which in turn points to another pointer, and so on. For example, a pointer may be used to access another mapping table within a given storage device of the storage devices 135A-N that identifies another pointer. Consequently, one or more levels of indirection may exist between the physical pointer value and a target storage location.
In various embodiments, the address translation table may be accessed using a key comprising a volume, snapshot, or other dataset ID, a logical or virtual address, a sector number, and so forth. A received read/write storage access request may identify a particular volume, sector, and length. A sector may be a logical block of data stored in a volume or snapshot, with a sector being the smallest size of an atomic I/O request to the storage system. In one embodiment, a sector may have a fixed size (e.g., 512 bytes) and the mapping tables may deal with ranges of sectors. For example, the address translation table may map a volume or snapshot in sector-size units. The areas being mapped may be managed as ranges of sectors, with each range consisting of one or more consecutive sectors. In one embodiment, a range may be identified by <snapshot, start sector, length>, and this tuple may be recorded in the address translation table and one or more other tables. In one embodiment, the key value for accessing the address translation table may be the combination of the volume or snapshot ID and the received sector number. A key is an entity in a mapping table that distinguishes one row of data from another row. In other embodiments, other types of address translation tables may be utilized.
In one embodiment, the address translation table may map volumes or snapshots and block offsets to physical pointer values. Depending on the embodiment, a physical pointer value may be a physical address or a logical address which the storage device maps to a physical location within the device. In one embodiment, an index may be utilized to access the address translation table. The index may identify locations of mappings within the address translation table. The index may be queried with a key value generated from a volume ID and sector number, and the index may be searched for one or more entries which match, or otherwise correspond to, the key value. Information from a matching entry may then be used to locate and retrieve a mapping which identifies a storage location which is the target of a received read or write request. In one embodiment, a hit in the index provides a corresponding virtual page ID identifying a page within the storage devices of the storage system, with the page storing both the key value and a corresponding physical pointer value. The page may then be searched with the key value to find the physical pointer value.
The deduplication table may include information used to deduplicate data at a fine-grained level. The information stored in the deduplication table may include mappings between one or more calculated hash values for a given data component and a physical pointer to a physical location in one of the storage devices 135A-N holding the given data component. In addition, a length of the given data component and status information for a corresponding entry may be stored in the deduplication table. It is noted that in some embodiments, one or more levels of indirection may exist between the physical pointer value and the corresponding physical storage location. Accordingly, in these embodiments, the physical pointer may be used to access another mapping table within a given storage device of the storage devices 135A-N.
Networks 120 and 150 may utilize a variety of techniques including wireless connection, direct local area network (LAN) connections, wide area network (WAN) connections such as the Internet, a router, storage area network, Ethernet, and others. Networks 120 and 150 may further include remote direct memory access (RDMA) hardware and/or software, transmission control protocol/internet protocol (TCP/IP) hardware and/or software, router, repeaters, switches, grids, and/or others. Protocols such as Fibre Channel, Fibre Channel over Ethernet (FCoE), iSCSI, and so forth may be used in networks 120 and 150. The networks 120 and 150 may interface with a set of communications protocols used for the Internet such as the Transmission Control Protocol (TCP) and the Internet Protocol (IP), or TCP/IP.
Client computer systems 125 and 175 are representative of any number of stationary or mobile computers such as desktop personal computers (PCs), physical computer terminals executing thin-client software, servers, server farms, workstations, laptops, handheld computers, servers, personal digital assistants (PDAs), smart phones, and so forth. Generally speaking, client computer systems 125 and 175 include one or more processors comprising one or more processor cores. Each processor core includes circuitry for executing instructions according to a predefined general-purpose instruction set. For example, the x86 instruction set architecture may be selected. Alternatively, the ARM®, Alpha®, PowerPC®, SPARC®, or any other general-purpose instruction set architecture may be selected. The processor cores may access cache memory subsystems for data and computer program instructions. The cache subsystems may be coupled to a memory hierarchy comprising random access memory (RAM) and a storage device.
It is noted that in alternative embodiments, the number and type of storage arrays, cloud services, client computers, servers, storage controllers, networks, storage device groups, and data storage devices is not limited to those shown in
Turning now to
Replication engine 210 may be configured to select data for replication from among datasets 202A-B and protection groups 203A-B. Original storage subsystem 205 may replicate a dataset or protection group to any of a plurality of storage subsystems and/or cloud service 235. A protection group may be defined as a group of hosts, host groups, and volumes within a storage subsystem or storage system. A single protection group may consist of multiple hosts, host groups and volumes. Generally speaking, a protection group may include logical storage elements that are replicated together consistently in order to correctly describe a dataset.
Replica storage subsystems 230A-B are coupled to original storage subsystem 205 and may be the target of replication operations. In one embodiment, replica storage subsystems 230A-B may be at the same location and on the same network as original storage subsystem 205. Original storage subsystem 205 may also be coupled to cloud service 235 via network 220, and original storage subsystem 205 may utilize cloud service 235 as a target for replicating data. Original storage subsystem 205 may also be configured to phone home log data to cloud service 235 for storage and analysis. Original storage subsystem 205 may also be coupled to replica storage subsystems 250A-N via network 240, and replica storage subsystems 250A-N may be the target of replication operations.
Replication engine 210 may be configured to selectively utilize deduplication (or dedup) unit 212 and/or compression unit 213 to deduplicate and compress the data being replicated. In one embodiment, replication engine 210 may utilize deduplication unit 212 and compression unit 213 to deduplicate and compress a dataset or protection group selected for replication. Any suitable types of deduplication and compression may be utilized, depending on the embodiment. In other embodiments, replication engine 210 may bypass deduplication unit 212 and compression unit 213 when performing replication. Replication engine 210 may also be configured to selectively utilize encryption unit 215 for encrypting data being replicated to other subsystems and/or to cloud service 235. Any suitable type of encryption may be utilized, depending on the embodiment.
In one embodiment, replication engine 210 may be configured to replicate data to replica storage subsystems 230A-B without encrypting the data being replicated. In this embodiment, replication engine 210 may be configured to encrypt data being replicating using encryption unit 215 for replication events which target cloud service 235. Replication engine 210 may encrypt or not encrypt data being replicated to replica storage subsystems 250A-N, depending on the embodiment. In one embodiment, an administrator or other authorized user may be able to select when encryption is enabled depending on the type of data being replicated and/or the replication target. A user may specify that encryption should be enabled for certain replication targets regardless of the type of data being replicated.
Original storage subsystem 205 may be configured to generate and display a graphical user interface (GUI) to allow users to manage the replication environment. When a user logs into the GUI, the GUI may show which subsystems can be used as targets for replication. In one embodiment, the GUI may be populated with data stored locally on subsystem 205. In another embodiment, the GUI may be populated with data received from cloud service 235. For example, original storage subsystem 205 may be part of a first organization, and when subsystem 205 is new and first becomes operational, subsystem 205 may not yet know which other storage subsystems exist within the first organization. Subsystem 205 may query cloud service 235 and cloud service 235 may provide data on all of the storage subsystems of the first organization which are available for serving as replication targets. These storage subsystems may then appear in the GUI used for managing the replication environment.
In one embodiment, snapshots that are replicated from original storage subsystem 205 to a target subsystem may have the same global content ID but may have separate local content IDs on original storage subsystem 205 and the target subsystem. In other embodiments, global IDs may be used across multiple storage subsystems. These global IDs may be generated such that no duplicate IDs are generated. For example, in one embodiment, an ID of the device on which it (e.g., the snapshot, medium, or corresponding data) was first written may be prepended. In other embodiments, ranges of IDs may be allocated/assigned for use by different devices. These and other embodiments are possible and are contemplated. For example, the local content ID of a first snapshot on original storage subsystem 205 may map to the global content ID 290 and the local content ID of the first snapshot on the target subsystem may also map to the global content ID 290. In this way, a given storage subsystem may be able to identify which of its snapshots are also present on other storage subsystems. In one embodiment, cloud service 235 may maintain mappings of local content IDs to global content IDs for the storage subsystems of a given organization.
It is noted that the storage environment shown in
Referring now to
Original storage subsystem 305 and replica storage subsystem 310 may also be coupled to cloud service 330 via network 325, wherein network 325 is representative of any number and type of networks. Original storage subsystem 305 and replica storage subsystem 310 may each be configured to phone home log data (e.g., performance data, capacity data, system health data, diagnostics, past alerts) to cloud service 330 at programmable intervals. In one embodiment, cloud service 330 may be configured to analyze diagnostics, logs, and any additional performance data received from the storage subsystems and generate alerts based on the analysis.
In one embodiment, if cloud service 330 detects that original storage subsystem 305 has not phoned home log data at the designated time or for a threshold amount of time, cloud service 330 may generate and convey a first alert to replica storage subsystem 310 notifying replica storage subsystem 310 of the failure of original storage subsystem 305 to phone home. In response to receiving the first alert, replica storage subsystem 310 may be configured to initiate disaster recovery operations for replicated dataset 315B. In another embodiment, cloud service 330 may determine, based on an analysis of phone home data, that a health indicator associated with original storage subsystem 305 has fallen below a programmable threshold. In this embodiment, cloud service 330 may determine that the health indicator of original storage subsystem 305 has fallen below the programmable threshold based on detecting an increase in latency, increased storage capacity utilization, a given number of program-erase cycles for one or more storage devices of original storage subsystem 305, an age of the one or more storage devices, a number of failed drives, increased error rate, out of space (i.e., high storage utilization capacity), and/or one or more other conditions associated with original storage subsystem 305.
In response to determining that the health of original storage subsystem 305 has fallen below the programmable threshold, cloud service 330 may generate and convey a second alert to replica storage subsystem 310 notifying replica storage subsystem 310 of the sub-optimal health of original storage subsystem 305. In response to receiving the second alert, replica storage subsystem 310 may be configured to initiate disaster recovery operations for replicated dataset 315B.
Although not shown in
Referring now to
A cloud-based service may receive log and performance data which is phoned home from a first storage subsystem on a regularly scheduled basis (block 405). In various embodiments, the log and performance data may include at least capacity data, system health data, diagnostics, past alerts, replication event data, logs, storage device count, host count, volume count, queue depth, read bandwidth (BW), read input/output operations per second (IOPS), read latency, write BW, write TOPS, write latency, and other data. In one embodiment, the first storage subsystem may be a storage array. In other embodiments, the first storage subsystem may be any of various other types of storage systems. At a given point in time, the cloud-based service may detect that the first storage subsystem has not phoned home data for a threshold amount of time (block 410). The threshold amount of time may be programmable and may vary from embodiment to embodiment.
In response to detecting that the first storage subsystem has not phoned home data for the threshold amount of time, the cloud-based service may determine if the first storage subsystem is replicating a dataset to any other storage subsystems (block 415). If the first storage subsystem is replicating a dataset to another storage subsystem, the cloud-based service may determine the target(s) of replication (block 420). The dataset may include any type of data, such as one or more volumes, virtual machines, disk images, databases, applications, protection groups, and/or other data. In one embodiment, the cloud-based service may determine if the first storage subsystem is replicating a dataset to another storage subsystem from the phone home data received from the first storage subsystem. The first storage subsystem may generate an indication or identification of any ongoing replication events and include this with the log data which is phoned home to the cloud-based service. Alternatively, the cloud-based service may receive a notification when a replication event is created, and the notification may include information (e.g., the source, the target, the dataset being replicated) associated with the replication event. The cloud-based service may store this notification and utilize it later for determining if the first storage subsystem is replicating data to another storage subsystem and for determining the target of the replication event.
After determining the target of replication, the cloud-based service may notify the target that the first storage subsystem has not phoned home data as expected (block 425). In various embodiments, the target may be a second storage subsystem, a server, a cloud-based server, a computing device coupled to one or more storage devices, or any of various other computing devices. In response to receiving this notification, the target may begin disaster recovery operations for the affected dataset(s) (block 430). After block 430, method 400 may end.
Turning now to
A first storage subsystem may receive replicated data from a second storage subsystem (block 505). In one embodiment, the first storage subsystem may receive a snapshot of the most recent changes to a dataset being replicated. The first storage subsystem may receive snapshots of the dataset on a regular schedule. In other embodiments, the first storage subsystem may receive replicated data in other types of formats. At a given point in time, the first storage subsystem may receive an alert from a cloud service that the second storage subsystem has not phoned home as expected (block 510). In response to receiving the alert, the first storage subsystem may begin disaster recovery procedures using the most recent snapshot replicated from the second storage subsystem (block 515).
Next, the first storage subsystem may reclaim, from the second storage subsystem, the node port interface virtualization (NPIV) port IDs and world wide names (WWNs) associated with the dataset (block 520). In one embodiment, the first and second storage subsystems may be coupled to one or more host servers via a Fibre Channel interface, and the first and second storage subsystems may be assigned unique NPIV port IDs and WWNs. In one embodiment, the cloud service may maintain a listing of NPIV port IDs and WWNs for a plurality of datasets on a plurality of storage subsystems, and the cloud service may provide the NPIV port IDs and WWNs of the second storage subsystem to the first storage subsystem. In other embodiments, other types of protocols and communication links (e.g., SCSI) may be utilized other than Fibre Channel, and other suitable techniques for performing disaster recovery procedures may be utilized in these embodiments. After block 520, method 500 may end.
Turning now to
A primary storage subsystem may replicate a first dataset to a secondary storage subsystem (block 605). The primary storage subsystem and the secondary storage subsystem may be owned and/or operated by a first organization. In one embodiment, the primary and secondary storage subsystems may be storage arrays. In other embodiments, the primary and secondary storage subsystems may be other types of storage systems or subsystems (e.g., servers, storage shelves, storage devices, PCIe cards, flash drives, SSDs, storage clusters, data centers). The primary storage subsystem and the secondary storage subsystem may also be coupled to a cloud-based service for phoning home log data, performance data, and other information on a periodic basis. The cloud-based service may execute on a server or other computing device located externally to the primary and secondary storage subsystems. The cloud-based service may be configured to detect one or more conditions indicative of an impending failure of the primary storage subsystem (block 610). The one or more conditions may include failure to phone home, a health rating falling below a threshold, and/or other conditions. It is noted that the cloud-based service may detect the one or more conditions in block 610 prior to an actual failure of the primary storage subsystem.
Next, the cloud-based service may be configured to select a third storage subsystem to take over as the new secondary storage subsystem in anticipation of the primary storage subsystem failing and the current secondary storage subsystem taking over as the new primary storage subsystem for the first dataset (block 615). The third storage subsystem may be selected from among a plurality of storage subsystems associated with the first organization. An example of a process for selecting a third storage subsystem to take over as the new secondary storage subsystem is described in
Next, the cloud-based service may generate a medium graph (e.g., medium graph 900 of
Next, the cloud-based service may be configured to identify a highest medium ID of the first dataset stored on the third storage subsystem (block 625). The cloud-based service may utilize a global to local medium ID mapping (e.g., table 1300 of
Next, the cloud-based service may be configured to generate and convey an alert to the current secondary storage subsystem, wherein the alert identifies the third storage subsystem as the new secondary storage subsystem for the first dataset (block 630). The alert may also specify the identity of the first medium which is stored on the third storage subsystem. In response to receiving the alert, the current secondary storage subsystem may begin disaster recovery procedures and start replicating, to the third storage subsystem, only the data which is unique to the mediums of the first dataset's medium graph which have an ID larger than the first medium (block 635). For example, if the first medium has an ID of 675, and the highest medium ID of the first dataset's medium graph is 695, the secondary storage subsystem may replicate only the new data for all of the mediums with IDs between 676 and 695 to the third storage subsystem. In other words, the secondary storage subsystem may only replicate the data unique to medium IDs 676 through 695 without replicating the data corresponding to any underlying mediums with a medium ID of less than or equal to 675. In one embodiment, the secondary storage subsystem may traverse the entirety of the range of medium ID 695, and if a lookup for a given data block lands in a medium with an ID greater than 675, then the secondary storage subsystem may replicate the given data block to the third storage subsystem. If a lookup for a given data block lands in a medium with an ID less than or equal to 675, then the secondary storage subsystem may skip replicating the given data block to the third storage subsystem since in this case, the given data block already resides on the third storage subsystem as part of medium ID 675.
Alternatively, rather than traversing the entirety of the range of medium ID 695, the secondary storage subsystem may replicate only the new data for all of the mediums with IDs between 676 and 695. Some duplicate data may be sent using this approach, but this may be faster in some cases than traversing the entirety of the range of medium ID 695. For example, medium ID 694 may have overwritten some of the changes first captured in medium ID 693. Therefore, sending these changes as part of medium ID 693 will be inefficient. However, this approach may be preferable in some cases. In some cases, the cloud-based service may be configured to select which technique is used for replicating data to the third storage subsystem based on one or more characteristics of the first dataset. For example, if the size of the first dataset is greater than a first threshold and the number of new mediums, higher than the first medium, of the first dataset is less than a second threshold, then the cloud-based service may replicate only the new data for all of the new mediums rather than traversing the entire range of the first dataset. In some embodiments, the cloud-based service may be able to detect changes captured in a second medium which were overwritten by changes captured in a third medium, wherein the third medium ID is higher than the second medium ID. In these embodiments, the cloud-based service may prevent the changes in the second medium from being sent to the third storage subsystem.
In another embodiment, the cloud-based service may notify the third storage subsystem of the medium IDs which are missing and which storage subsystems and/or cloud services store these missing medium IDs, and the third storage subsystem may retrieve these medium IDs from other storage subsystems and/or the cloud. In a further embodiment, the cloud-based service may identify which storage subsystems and cloud services store the missing medium IDs and the cloud-based service may generate requests for these storage subsystems and cloud services to replicate the mediums corresponding to the missing medium IDs to the third storage subsystem. In some embodiments, the cloud-based service and storage subsystems may utilize deduplication to reduce the amount of data stored in the cloud and on the plurality of storage subsystems. In these embodiments, the cloud-based service may first send fingerprints of the data of the missing mediums to the third storage subsystem and/or request that other storage subsystems send fingerprints of the data of the missing mediums to the third storage subsystem. The third storage subsystem may check for matches to the received fingerprints, and then the third storage subsystem may request only the data corresponding to any new fingerprints. After block 635, method 600 may end.
Referring now to
It may be assumed for the purposes of the discussion of method 700 that a primary storage subsystem storing a first dataset has failed or is predicted to fail. The primary storage subsystem may be replicating the first dataset to a secondary storage subsystem. A cloud-based service may be configured to select a new secondary storage subsystem since the current secondary storage subsystem is or soon will be taking over as the new primary storage subsystem. Method 700 may be performed within this context.
In response to an indication of a failure or a prediction of failure of the primary storage subsystem, the cloud-based service may generate a medium graph of the first dataset (block 705). In some cases, the cloud-based service may have previously generated the medium graph for the first dataset and the cloud-based service may be automatically updating the medium graph for the first dataset as changes are made. In these cases, the cloud-based service may load and/or access the medium graph in block 705. In one embodiment, the medium graph may be populated with the local medium IDs of the primary storage subsystem.
Next, the cloud-based service may translate the local medium IDs of the first dataset's medium graph into global medium IDs (block 710). In one embodiment, the cloud-based service may maintain a global to local medium ID table (e.g., table 1300 of
If the given storage subsystem meets the one or more other conditions (conditional block 720, “yes” leg), then the cloud-based service may select the given storage subsystem as the new secondary storage subsystem of the first dataset (block 725). If the given storage subsystem does not meet the one or more other conditions for taking over as the new secondary storage subsystem of the first dataset (conditional block 720, “no” leg), then method 700 may return to block 715 with the cloud-based service determining which storage subsystem of the plurality of storage subsystems of the first organization has the next highest medium ID of the first dataset's medium graph. After block 725, method 700 may end.
Turning now to
A cloud-based service may detect or receive an indication of an impending failure of a replication event of a dataset from a first to a second storage subsystem (block 805). In response to the indication, the cloud-based service may identify a third storage subsystem to take over for either the first or second storage subsystem (block 810). In one embodiment, the cloud-based service may select the third storage subsystem as described in method 700. In another embodiment, an administrator may select the third storage subsystem as a backup for the replication event of the dataset.
Next, the cloud-based service may identify which mediums of the dataset are missing from the third storage subsystem (block 815). Then, the cloud-based service may determine other storage subsystems which store these missing mediums (block 820). In one embodiment, the cloud-based service may locate the missing mediums using a global to local medium ID mapping table (e.g., (e.g., table 1300 of
Referring now to
The term “medium” as is used herein is defined as a logical grouping of data. A medium may have a corresponding identifier (ID) with which to identify the logical grouping of data. Each medium may have a unique ID that is never reused in the system or subsystem. In other words, the medium ID is non-repeating. In one embodiment, the medium ID may be a monotonically increasing number. In some embodiments, the medium ID may be incremented for each snapshot taken of the corresponding dataset, volume, or logical grouping of data. In these embodiments, the medium ID may be a sequential, non-repeating ID. Each medium may also include or be associated with mappings of logical block numbers to content location, deduplication entries, and other information. In one embodiment, medium identifiers may be used by the storage controller but medium identifiers may not be user-visible. A user (or client) may send a data request accompanied by a volume ID to specify which data is targeted by the request, and the storage controller may map the volume ID to a medium ID and then use the medium ID when processing the request.
The term “medium” is not to be confused with the terms “storage medium” or “computer readable storage medium”. A storage medium is defined as an actual physical device (e.g., SSD, HDD) that is utilized to store data. A computer readable storage medium (or non-transitory computer readable storage medium) is defined as a physical storage medium configured to store program instructions which are executable by a processor or other hardware device. Various types of program instructions that implement the methods and/or mechanisms described herein may be conveyed or stored on a computer readable medium. Numerous types of media which are configured to store program instructions are available and include hard disks, floppy disks, CD-ROM, DVD, flash memory, Programmable ROMs (PROM), random access memory (RAM), and various other forms of volatile or non-volatile storage.
It is also noted that the term “volume to medium mapping table” may refer to multiple tables rather than just a single table. Similarly, the term “medium mapping table” may also refer to multiple tables rather than just a single table. It is further noted that volume to medium mapping table 915 is only one example of a volume to medium mapping table. Other volume to medium mapping tables may have other numbers of entries for other numbers of volumes.
Each medium is depicted in graph 900 as three conjoined boxes, with the leftmost box showing the medium ID, the middle box showing the underlying medium, and the rightmost box displaying the status of the medium (RO—read-only) or (RW—read-write). Read-write mediums may be referred to as active mediums, while read-only mediums may represent previously taken snapshots. Within graph 900, a medium points to its underlying medium. For example, medium 20 points to medium 12 to depict that medium 12 is the underlying medium of medium 20. Medium 12 also points to medium 10, which in turn points to medium 5, which in turn points to medium 1. Some mediums are the underlying medium for more than one higher-level medium. For example, three separate mediums (12, 17, 11) point to medium 10, two separate mediums (18, 10) point to medium 5, and two separate mediums (6, 5) point to medium 1. Each of the mediums which is an underlying medium to at least one higher-level medium has a status of read-only.
It is noted that the term “ancestor” may be used to refer to underlying mediums of a given medium. In other words, an ancestor refers to a medium which is pointed to by a first medium or which is pointed to by another ancestor of the first medium. For example, as described above and shown in
The set of mediums on the bottom left of graph 900 is an example of a linear set. As depicted in graph 900, medium 3 was created first and then a snapshot was taken resulting in medium 3 becoming stable (i.e., the result of a lookup for a given block in medium 3 will always return the same value after this point). Medium 7 was created with medium 3 as its underlying medium. Any blocks written after medium 3 became stable were labeled as being in medium 7. Lookups to medium 7 return the value from medium 7 if one is found, but will look in medium 3 if a block is not found in medium 7. At a later time, a snapshot of medium 7 is taken, medium 7 becomes stable, and medium 14 is created. Lookups for blocks in medium 14 would check medium 7 and then medium 3 to find the targeted logical block. Eventually, a snapshot of medium 14 is taken and medium 14 becomes stable while medium 15 is created. At this point in graph 900, medium 14 is stable with writes to volume 102 going to medium 15.
Volume to medium mapping table 915 maps user-visible volumes to mediums. Each volume may be mapped to a single medium, also known as the anchor medium. This anchor medium, as with all other mediums, may take care of its own lookups. A medium on which multiple volumes depend (such as medium 10) tracks its own blocks independently of the volumes which depend on it. Each medium may also be broken up into ranges of blocks, and each range may be treated separately in medium DAG 900.
Turning now to
Each medium may be identified by a medium ID, as shown in the leftmost column of table 1000. A range attribute may also be included in each entry of table 1000, and the range may be in terms of data blocks. The size of a block of data (e.g., 4 KB, 8 KB) may vary depending on the embodiment. It is noted that the terms “range” and “extent” may be used interchangeably herein. A medium may be broken up into multiple ranges, and each range of a medium may be treated as if it is an independent medium with its own attributes and mappings. For example, medium ID 2 has two separate ranges. Range 0-99 of medium ID 2 has a separate entry in table 1000 from the entry for range 100-999 of medium ID 2.
Although both of these ranges of medium ID 2 map to underlying medium ID 1, it is possible for separate ranges of the same source medium to map to different underlying mediums. For example, separate ranges from medium ID 35 map to separate underlying mediums. For example, range 0-299 of medium ID 35 maps to underlying medium ID 18 with an offset of 400. This indicates that blocks 0-299 of medium ID 35 map to blocks 400-699 of medium ID 18. Additionally, range 300-499 of medium ID 35 maps to underlying medium ID 33 with an offset of −300 and range 500-899 of medium ID 35 maps to underlying medium ID 5 with an offset of −400. These entries indicate that blocks 300-499 of medium ID 35 map to blocks 0-199 of medium ID 33, while blocks 500-899 of medium ID 35 map to blocks 100-499 of medium ID 5. It is noted that in other embodiments, mediums may be broken up into more than three ranges.
The state column of table 1000 records information that allows lookups for blocks to be performed more efficiently. A state of “Q” indicates the medium is quiescent, “R” indicates the medium is registered, and “U” indicates the medium is unmasked. In the quiescent state, a lookup is performed on exactly one or two mediums specified in table 1000. In the registered state, a lookup is performed recursively. The unmasked state determines whether a lookup should be performed in the basis medium, or whether the lookup should only be performed in the underlying medium. Although not shown in table 1000 for any of the entries, another state “X” may be used to specify that the source medium is unmapped. The unmapped state indicates that the source medium contains no reachable data and can be discarded. This unmapped state may apply to a range of a source medium. If an entire medium is unmapped, then the medium ID may be entered into a sequence invalidation table and eventually discarded.
In one embodiment, when a medium is created, the medium is in the registered state if it has an underlying medium, or the medium is in the quiescent state if it is a brand-new volume with no pre-existing state. As the medium is written to, parts of it can become unmasked, with mappings existing both in the medium itself and the underlying medium. This may be done by splitting a single range into multiple range entries, some of which retain the original masked status, and others of which are marked as unmasked.
In addition, each entry in table 1000 may include a basis attribute, which indicates the basis of the medium, which in this case points to the source medium itself. Each entry may also include an offset field, which specifies the offset that should be applied to the block address when mapping the source medium to an underlying medium. This allows mediums to map to other locations within an underlying medium rather than only being built on top of an underlying medium from the beginning block of the underlying medium. As shown in table 1000, medium 8 has an offset of 500, which indicates that block 0 of medium 8 will map to block 500 of its underlying medium (medium 1). Therefore, a lookup of medium 1 via medium 8 will add an offset of 500 to the original block number of the request. The offset column allows a medium to be composed of multiple mediums. For example, in one embodiment, a medium may be composed of a “gold master” operating system image and per-VM (virtual machine) scratch space. Other flexible mappings are also possible and contemplated.
Each entry also includes an underlying medium attribute, which indicates the underlying medium of the source medium. If the underlying medium points to the source medium (as with medium 1), then this indicates that the source medium does not have an underlying medium, and all lookups will only be performed in the source medium. Each entry may also include a stable attribute, with “Y” (yes) indicating the medium is stable (or read-only), and with “N” (no) indicating the medium is read-write. In a stable medium, the data corresponding to a given block in the medium never changes, though the mapping that produces this data may change. For example, medium 2 is stable, but block 50 in medium 2 might be recorded in medium 2 or in medium 1, which may be searched logically in that order, though the searches may be done in parallel if desired. In one embodiment, a medium will be stable if the medium is used as an underlying medium by any medium other than itself.
Turning now to
Using the level ID, page ID, and a key value generated from the medium ID and block number, the corresponding mapping table entry may be located and a pointer to the storage location may be returned from this entry. The pointer may be used to identify or locate data stored in the storage devices of the storage system. In addition to the pointer value, status information, such as a valid indicator, a data age, a data size, and so forth, may be stored in Field( ) to FieldN shown in Level N of mapping table 1120. It is noted that in various embodiments, the storage system may include storage devices (e.g., SSDs) which have internal mapping mechanisms. In such embodiments, the pointer in the mapping table entry may not be an actual physical address per se. Rather, the pointer may be a logical address which the storage device maps to a physical location within the device.
For the purposes of this discussion, the key value used to access entries in index 1110 is the medium ID and block number corresponding to the data request. However, in other embodiments, other types of key values may be utilized. In these embodiments, a key generator may generate a key from the medium ID, block number, and/or one or more other requester data inputs, and the key may be used to access index 1110 and locate a corresponding entry.
In one embodiment, index 1110 may be divided into partitions, such as partitions 1112a-1112b. In one embodiment, the size of the partitions may range from a 4 kilobyte (KB) page to 256 KB, though other sizes are possible and are contemplated. Each entry of index 1110 may store a key value, and the key value may be based on the medium ID, block number, and other values. For the purposes of this discussion, the key value in each entry is represented by the medium ID and block number. This is shown merely to aid in the discussion of mapping between mediums and entries in index 1110. In other embodiments, the key values of entries in index 1110 may vary in how they are generated.
In various embodiments, portions of index 1110 may be cached, or otherwise stored in a relatively fast access memory. In various embodiments, the entire index 1110 may be cached. In some embodiments, where the primary index has become too large to cache in its entirety, or is otherwise larger than desired, secondary, tertiary, or other index portions may be used in the cache to reduce its size. In addition to the above, in various embodiments mapping pages corresponding to recent hits may be cached for at least some period of time. In this manner, processes which exhibit accesses with temporal locality can be serviced more rapidly (i.e., recently accessed locations will have their mappings cached and readily available).
In some embodiments, index 1110 may be a secondary index which may be used to find a key value for accessing a primary index. The primary index may then be used for locating corresponding entries in address translation table 1100. It is to be understood that any number of levels of indexes may be utilized in various embodiments. In addition, any number of levels of redirection may be utilized for performing the address translation of received data requests, depending on the embodiment. In some embodiments, a corresponding index may be included in each level of mapping table 1120 for mappings which are part of the level. Such an index may include an identification of mapping table entries and where they are stored (e.g., an identification of the page) within the level. In other embodiments, the index associated with mapping table entries may be a distinct entity, or entities, which are not logically part of the levels themselves. It is noted that in other embodiments, other types of indexes and mapping tables may be utilized to map medium IDs and block numbers to physical storage locations.
Mapping table 1120 may comprise one or more levels. For example, in various embodiments, table 1120 may comprise 16 to 64 levels, although other numbers of levels supported within a mapping table are possible and contemplated. Three levels labeled Level “N”, Level “N-1” and Level “N-2” are shown for ease of illustration. Each level within table 1120 may include one or more partitions. In one embodiment, each partition is a 4 kilo-byte (KB) page. In one embodiment, a corresponding index 1110 may be included in each level of mapping table 1120. In this embodiment, each level and each corresponding index 1110 may be physically stored in a random-access manner within the storage devices.
In another embodiment, table 1100 may be a deduplication table. A deduplication table may utilize a key comprising a hash value determined from a data component associated with a storage access request. For each data component, a deduplication application may be used to calculate a corresponding hash value. In order to know if a given data component corresponding to a received write request is already stored in one of the storage devices, bits of the calculated hash value (or a subset of bits of the hash value) for the given data component may be compared to bits in the hash values of data components stored in one or more of the storage devices.
In a further embodiment, table 1100 may be an overlay table. One or more overlay tables may be used to modify or elide tuples corresponding to key values in the underlying mapping table and provided by other tables in response to a query. The overlay table(s) may be used to apply filtering conditions for use in responding to accesses to the mapping table or during flattening operations when a new level is created. Keys for the overlay table need not match the keys for the underlying mapping table. For example, an overlay table may contain a single entry stating that a particular range of data has been deleted or is otherwise inaccessible and that a response to a query corresponding to a tuple that refers to that range is invalid. In another example, an entry in the overlay table may indicate that a storage location has been freed, and that any tuple that refers to that storage location is invalid, thus invalidating the result of the lookup rather than the key used by the mapping table. In some embodiments, the overlay table may modify fields in responses to queries to the underlying mapping table. In some embodiments, a range of key values may be used to efficiently identify multiple values to which the same operation is applied. In this manner, tuples may effectively be “deleted” from the mapping table by creating an “elide” entry in the overlay table and without modifying the mapping table. The overlay table may be used to identify tuples that may be dropped from the mapping table in a relatively efficient manner. It is noted that in other embodiments, other types of mapping tables may be utilized with the replication techniques disclosed herein. For example, in another embodiment, a single log file may be utilized to map logical addresses to physical addresses. In a further embodiment, a key-value store may be utilized. Other structures of mapping tables are possible and are contemplated.
Turning now to
In one embodiment, each of storage arrays 1210, 1230, and 1240 may include the components (e.g., storage controller, device groups) shown in storage array 105 (of
For the purposes of this discussion, original storage array 1240 represents the array on which a given volume and snapshot were first created. Replica storage array 1210 may represent the array to which the given snapshot is being replicated. Source storage array 1230 may represent an array containing the medium to be replicated from which replica storage array 1210 is pulling missing data necessary for the given snapshot. It is noted that these designations of the various storage arrays are used in the context of a given replication operation. For subsequent replication operations, these designations may change. For example, a first snapshot may be replicated from original storage array 1240 to replica storage array 1210 at a particular point in time. At a later point in time, a second snapshot may be replicated from replica storage array 1210 to original storage array 1240. For the replication of the second snapshot, storage array 1210 may be referred to as an “original” storage array while storage array 1240 may be referred to as a “replica” storage array. Also, the source storage system and the original storage system may be the same for a given replication event. In other words, system 1210 could pull data to replicate a medium from system 1240 directly if it chooses.
In system 1200, snapshots may be taken independently by original storage array 1240. Then, replica storage array 1210 may decide which particular snapshots to replicate when replica storage array 1210 connects to original storage array 1240. In this way, replica storage array 1210 does not need to copy a large number of snapshots if it has not connected to original storage array 1240 for a long period of time. Instead, replica storage array 1210 may only choose to replicate the most recent snapshot. Alternatively, original storage array 1240 may make a policy decision and notify replica storage array 1210 to pull a given snapshot as embodied in a given medium. Replica storage array 1210 may then choose to pull extents of the given medium from any storage array to which it has access.
In one embodiment, system 1200 may implement a replication mechanism using mediums to avoid copying data. For example, suppose that M is a medium comprising a snapshot S of volume V, and that M′ is a medium comprising a later snapshot S′ of V. If replica storage array 1210 already contains M, source storage array 1230 may transfer data in M′ but not in M to replica storage array 1210 so as to perform the replication process of medium M′ Source storage array 1230 may determine which regions fall through and which regions are actually in M′ by reading the medium map that it maintains.
In one embodiment, each storage array may utilize a local name for every medium maintained by the storage array, including mediums that originated locally and mediums that were replicated from other storage arrays. For mediums originating from other storage arrays, the local storage array may keep a table mapping original array ID and original medium ID to local medium ID. An example table for mapping original array ID and original medium ID to local medium ID is shown in
In one embodiment, to replicate a snapshot from original storage array 1240 to replica storage system 1210, the following steps may be taken: First, the anchor medium corresponding to the snapshot on original storage array 1240 may be made stable by taking a snapshot of the volume if necessary. If this anchor medium is already stable, then there is no need to take the snapshot. Next, replica storage system 1210 may initiate the replication process by querying original storage array 1240 for a list of snapshots of the volume that could be replicated. Original storage array 1240 may respond with a list of possible snapshots and corresponding mediums for each snapshot. Then, the medium corresponding to the desired snapshot may be replicated to storage array 1210. This medium may be called ‘M’. Replica storage system 1210 may then contact any source storage system 1230 in system 1200 with the medium M that it wants to replicate. Replica storage system 1210 may utilize its mapping table to identify all of the medium extents that are available for use as sources for deduplicated data, and may also optionally supply this list of medium extents that it maintains locally to source storage system 1230. Again, it is noted that source storage system 1230 may be original storage system 1240, or it may be another storage system to which original storage system 1240 has, directly or indirectly, previously replicated medium M.
Source storage array 1230 may use the list of medium extents and the medium ‘M’ selected for replication to build a list of information that needs to be sent to replica storage system 1210 to replicate medium M. Each packet of information may be referred to as a “quantum” or an “rblock”. An rblock can specify the content of a particular region of M as either medium extents that already exist on replica storage system 1210 or as data that has previously been sent from source storage array 1230 to replica storage system 1210 for M. An rblock can also contain a list of data tuples for M. A tuple may be a combination of block ID and data for the particular region of M. An rblock may also contain a combination of references and data tuples.
Replica storage system 1210 may acknowledge rblocks sent by source storage array 1230. Replica storage system 1210 may batch acknowledgements and send several at once rather than sending an acknowledgement after receiving each rblock. Acknowledgements may be sent using any suitable technique, including explicit acknowledgement by serial number of each rblock or acknowledging the latest serial number received with no gaps in serial number.
Source storage array 1230 may keep track of the latest rblock that replica storage system 1210 has acknowledged. Source storage array 1230 may discard rblocks that replica storage system 1210 has acknowledged since these will not need to be resent. Source storage array 1230 may add the extents that replica storage array 1210 acknowledges to the list of medium extents that replica storage array 1210 knows about. This list may help reduce the amount of actual data that source storage array 1230 sends to replica storage system 1210 as part of the replication process.
The above-described techniques for performing replication offer a variety of advantages. First, data that source storage array 1230 can determine already exists in a medium extent present on replica storage system 1210 is not sent; instead, source storage array 1230 sends a reference to the already-present data. Second, streamed rblocks do not overlap. Rather, each rblock specifies a disjoint range of content in M. Third, an rblock may only refer to a medium extent that source storage array 1230 knows is on replica storage system 1210, either because it was in the original list of extents sent by replica storage system 1210 to source storage array 1230, or because replica storage system 1210 has acknowledged the extent to source storage array 1230. In some embodiments, replica storage system 1210 may respond that it does not have the referenced extents. In such a case, source storage array 1230 may be requested to resend the extents.
The above-described techniques allow system 1200 to efficiently discover duplicate blocks on source storage array 1230 to produce a correct duplicate. One approach which may be used involves running a differencing algorithm on source storage array 1230 to determine which data blocks must be sent in full and which regions of M can be sent as references to already-extant extents. In one embodiment, for a given extent ‘E’, an optionally discontiguous set of rblocks with patterns may be sent first, and then a reference rblock may be sent that fully covers the extent E.
A typical medium mapping table may map extents such that <M1, offset1, length> maps to <M2, offset2>, wherein M1, and M2 are two separate mediums and offsets and offset 2 are the offsets within those mediums. It may be challenging to determine whether a particular medium is reachable multiple ways using the individual medium extent map that maps <M1, offset1, length>—><M2, offset2>. In other words, it may be challenging to determine if other medium extents also point to <M2, offset2>. To address this problem, a set D1 of medium extents that are mapped to one another may be built. Thus, this set would include all instances of <MD, offsetD> that are pointed to by more than one <M, offset>. This set may allow a merge of all references to the duplicated medium extent <MD, offsetD> by ensuring that all references to blocks in the region refer to the canonical extent MD, rather than to whatever medium they were in that points to MD.
It may also be challenging to determine whether a particular block is a duplicate by resolving it through the medium maps, since translating a given <medium, block> results in a physical address. If blocks <M1, s1> and <M2, s2> both correspond to physical address X, it may be difficult to know when we resolve <M1, S1> that there are other blocks with address X. In other words, working backwards from X to the <medium, block> addresses that refer to it may be problematic. To mitigate these challenges, a set D2 of medium extents may be built that are duplicates of other medium extents. This set may indicate what ranges in different mediums actually correspond to the same blocks, whether by entries in the medium table or by fully resolving the addresses. Any suitable method for building this set D2 of medium extents may be utilized, depending on the embodiment. The two sets of D1 and D2 may be combined into a combined set D of duplicate medium extents.
Once a set of duplicate references has been built, source storage array 1230 may determine which blocks need to be sent to replica storage array 1210. Source storage array 1230 may determine which blocks need to be sent by performing the following steps: First, the set of duplicate extents D may be provided as previously described. Next, a set of sectors Z that replica storage array 1210 already knows about are initialized by inserting all of the sector ranges covered by the medium extents that replica storage array 1210 sent to source storage array 1230.
Next, a set of mappings P from physical addresses (X) to logical addresses (<M, s>) may be initialized to be empty. Each time actual data is sent to replica storage array 1210, the corresponding mapping may be added to set P. Then, for each sector ‘s’ in M, call a function emit_sector (M, s). Once sufficient information has been emitted, the information may be packaged into an rblock and sent to replica storage array 1210. In one embodiment, the function emit_sector (M, s) may traverse the medium extent table until one of the following three cases (a, b, c) happens. Checking for these three cases may be performed in logical order. For example, the checks may be run in parallel, but case a takes precedence over case b, and case b takes precedence over case c.
The three cases (a, b, c) mentioned above are as follows: First, case a is the following: <M, s> maps to a sector in Z called <Q, t>. In this case, emit a reference <M, s>→<Q, t>. Second, case b is the following: A sector <F, t> is hit that's in D, where F M. This means that a medium extent map in the medium mapping table has been traversed to a different medium, and an entry has been hit which allows the medium map to be “flattened” to optimize transmission. Flattening the medium map means that a duplicate entry is being deleted and both entries may now point to the same extent. In this case, emit_sector(F, t) may be called, and then a reference <M, s>→<F, t> may be emitted.
Third, case c is the following: An actual physical mapping X is hit that contains the data for the sector. There are two options when this occurs. If P already contains a mapping from X <O, t>, then emit a reference from <M, s><O, t>. Otherwise, emit the logical address of the sector—<M, s>—followed by the data for the sector. Also, add the mapping from X to <M, s> to P to allow for deduplicating on the fly to save bandwidth on the network.
In one embodiment, an optimization may be utilized. This optimization includes maintaining a list of recently sent physical addresses that map physical location X to <M, s>. This list may be used to do fine-grained deduplication on the fly. In option c above, first the list of recently-sent physical addresses may be checked. If it is discovered that <M2, s2> corresponds to physical address Y, and Y was recently sent as <M1, s1>, a reference may be sent from <M2, s2> to <M1, s1>. This step is purely optional, and the size of the list of recently-sent physical addresses can be as large or as small (including zero) as desired, with larger lists resulting in potentially less data being sent. The list of recently-sent addresses may be trimmed at any time, and any mappings may be removed. The use of table P may be omitted entirely if desired, with the only drawback being that fine grained duplicates might be sent multiple times over the network.
Another optimization is that adjacent references may be merged to save space. For example, if the references <M, s>→<O, t> and <M, s+1>→<0, t+1> were going to be sent, <M, s, 2>→<O, t> could be sent instead, where the number 2 indicates the number of sectors covered by this mapping. This optimization may be used at any time. For example, if the mapping table indicates that a mapping applies for the next 16 sectors, a single mapping may be emitted that covers the next 16 sectors. This avoids having to emit 16 individual mappings and then merge them later.
It is noted that the transmission of data and mappings from source storage array 1230 to replica storage array 1210 may be performed using any suitable network mechanism. Similarly, acknowledgments may be sent using any suitable mechanism for acknowledgment, including the use of sequence numbers or implicit acknowledgment built into network protocols.
The above-described mechanisms may be used to back up data to a “slower” storage device such as disk or tape. This backup can proceed at full sequential write speeds, since all of the network traffic on the backup destination (replica storage array 1210) may be recorded to keep track of the medium extents that are stored there. Resolving references to data stored on disk or tape could be slow using this approach. However, since network traffic is being recorded, data does not need to be processed on replica storage array 1210. Instead, all of the packets that source storage array 1230 sends to replica storage array 1210 may be sequentially recorded, and minimal processing of metadata from the rblocks may be performed. Then, if a restore is needed, all of the replication sessions may be replayed to original storage array 1240 or to another storage array.
Restoring data to another storage array could be achieved by replaying all of the desired replication streams from backup storage, in order. For example, suppose that daily replication of data was performed for every day of the month of August, with the initial replication of the volume being sent on August Pt. If a user wanted to restore the system as it looked on August 15, all of the stored streams for August 1-15 may be replayed.
The above-described mechanisms may be used to back up data to the cloud. Cloud storage may be used to preserve copies of all of the rblocks that would have been sent from source storage array 1230 to replica storage array 1210, and the cloud-based system may acknowledge medium extents as it receives the rblocks that contain them. A unique identifier may be assigned to each rblock, allowing a cloud-based system to efficiently store all of the rblocks, retrieving them as necessary to perform a restore from backup.
The mechanisms described herein deal may easily handle complex replication topologies. For example, suppose an original storage site is in London, with replicas in New York and Boston. The original pushes its data out to New York first. When Boston decides to replicate a snapshot, it can contact either London or New York to discover what snapshots are available for replication. Boston can then retrieve data from either London, New York, or parts from both, making the choice based on factors such as available network capacity and available system capacity (how busy the systems are). In other words, a replica storage array can pull from any source storage array that has the desired medium extents, not just the original storage array.
For example, Boston could decide to start retrieving data for snapshot S from London, but stop in the middle and switch to New York if the network connection to London became slow or the system in London became more heavily loaded. The system in New York can associate the London medium identifiers with data it has stored locally, and resume the transfer. Similarly, the system in Boston might identify the snapshot at New York initially, perhaps picking the latest snapshot stored in New York, bypassing London entirely. Boston may also contact London to identify the latest snapshot, but conduct the entire transfer with the New York replica.
Additionally, replication may also be used to preload a system with various mediums. This can be done even if it is never intended to replicate the volumes that currently use the mediums that are being preloaded. For example, mediums could be preloaded that correspond to “gold master” images of virtual machines that are commonly cloned. Then, when a new clone of the gold master is created, future replications would go very quickly because they can refer to the mediums that the replica was preloaded with. This preloading could be done with the storage arrays in close proximity, with the replica storage array then moved to a remote location. Also, coarse-grained deduplication may be performed after the fact on the preloaded data, further optimizing replication to a preloaded replica.
Turning now to
Referring now to
Once medium 1410 has been selected for replication, the replica storage array may generate a list of medium extents stored on the replica storage array that originated from the original storage array. Table 1465 is intended to represent the mapping of external storage array medium IDs to local medium IDs on the replica storage array. For the purposes of this discussion, it may be assumed that the original storage array has an ID of 1445. As shown, there is a single entry for storage array 1445 in table 1465. This entry maps original medium ID 1425 from the original storage array to local medium ID 36 on the replica storage array. It is noted that a typical table may have a large number of entries corresponding to the original storage array. However, a single entry is shown in table 1465 for ease of illustration. The medium mapping table entry for medium ID 36 is shown in table 1470, which is intended to represent the medium mapping table of the replica storage array. Alternatively, in another embodiment, each medium may have a globally unique ID, and mediums may be identified by the same globally unique ID on different storage arrays. In this embodiment, the replica storage array may simply look for entries assigned to medium ID 1410 in its medium mapping table.
List 1415A is intended to represent an example of a list which may be sent from the replica storage array to the original storage array. The replica storage array may generate list 1415A by querying table 1465 which maps external storage array medium IDs to local medium IDs and compiling a list of medium extents corresponding to snapshots that originated on the original storage array. The replica storage array may send list 1415A to the original storage array, and then the original storage array may filter out all medium extents that do not correspond to medium 1410 and keep only the medium extents which map to extents within medium 1410. Any number of entries may be included in list 1415A, depending on the embodiment.
As part of the replication process, the original storage array may determine which extents of medium ID 1410 need to be sent to the replica storage array and which extents can be sent as references to extents already stored on the replica storage array. Extents which can be sent as references to already-existent extents may be identified using any of a variety of techniques. For instance, if a first extent in table 1400 corresponds to an extent stored in list 1415A, then a reference to the extent of list 1415A may be sent to the replica storage array rather than sending the first extent. Also, if duplicate extents are discovered in table 1400, then a reference from a second extent to a third extent may be sent to replica storage array rather than sending the second extent. The original storage array may utilize any of a variety of techniques for determining if there are duplicate extents in list 1425. Additionally, if duplicate extents are discovered in table 1400, then these duplicate extents may be deduplicated as a side benefit of the replication process.
For example, in one embodiment, the original storage array may build up a list of duplicate extents that have been detected within medium 1410. In order to build list 1430 of duplicate extents, the original storage array may traverse table 1400 entry by entry to determine the underlying mappings which exist for each extent. For example, the fourth entry of table 1400 may be traversed down to its underlying medium of 650. Then, a lookup of the overall medium mapping table 1455 may be performed for the specified range of medium ID 650 to determine if medium ID 650 has an underlying medium. The second entry of medium mapping table 1455 shows the corresponding entry for this specific range of medium ID 650. In this case, the range of C to (D-1) of medium ID 650 has an underlying medium of 645 at an offset of 0 after applying the offset of —C from the entry in table 1455. Therefore, the extent corresponding to the fourth entry of table 1400 is a duplicate extent since it maps to the same extent as the third entry of table 1400. Accordingly, an entry may be recorded in duplicate extents table 1430 corresponding to the fourth and third entries of table 1400. Additionally, after detecting these duplicate extents, the medium mapping table entry for range C to (D-1) of medium ID 1410 may be collapsed. Although not shown in
Additionally, duplicate extents table 1430 may keep track of duplicate blocks within medium ID 1410 that map to the same physical address. When separate blocks that point to the same physical address are detected, an entry may be stored in duplicate extents table 1430 for the duplicate pair of blocks. Duplicate blocks may be detected by performing a lookup of the address translation table (not shown) for each block within medium 1410 and compiling a list of the physical pointer values returned from each of the lookups. For each pair of matching physical pointer values which are found, an entry may be recorded in duplicate extents table 1430. It may be assumed for the purposes of this discussion that the block corresponding to medium ID 1410 for range D to (E-1) is a duplicate block which has the same physical pointer value as the block corresponding to medium 1410 for range M to (N-1). Therefore, the second entry of duplicate extents table 1430 stores the mapping of these duplicate blocks.
Also, a physical to logical address mappings table 1460A may be created to store physical to logical mappings of data that is sent to the replica storage array. The physical to logical address mappings table 1460A may be initialized to be empty and mappings may be added after the actual data is sent to the replica storage array. Once duplicate extents table 1430 and physical to logical address mappings table 1460A have been created, the original storage array may traverse table 1400 entry by entry and determine for each entry if the actual data needs to be sent or if a reference to an already-existent extent on the replica storage array may be sent.
While traversing table 1400 for each sector of medium ID 1410, multiple conditions may be checked for each sector. First, it may be determined if the sector of medium ID 1410 maps to a sector in list 1415A. If the sector maps to one of the sectors indicated by list 1415A, then a reference to this sector from list 1415A may be sent to the replica storage array. For example, for the first entry of table 1400, a lookup of list 1415A will hit for this sector of medium ID 1425 corresponding to range 0-(A-1). As can be seen from the first entry of medium mapping table 1455, range 0 to (A-1) of medium ID 1425 maps to range 0 to (A-1) of medium ID 1410. Therefore, rather than sending the data for this sector to the replica storage array, a reference to the sector which already exists on the replica storage array may be sent.
After checking for the first condition and determining the first condition is not met, a second condition may be checked for a given sector of medium ID 1410. The second condition includes checking if the sector of medium ID 1410 maps to a sector in duplicate extents table 1430. If the sector of medium ID 1410 already maps to a sector in duplicate extents table 1430 which has already been sent to and acknowledged by the replica storage array, then a reference to the duplicate sector may be sent to the replica storage array. For example, for the fourth entry of table 1400 corresponding to range C to (D-1) of medium 1410, an entry exists in duplicate extents table 1430 for this range of medium 1410. Therefore, a reference to the range listed in the duplicate range column of table 1430, or range B-(C-1), may be sent to the replica storage array rather than sending the actual data. Similarly, for the last entry in table 1400 corresponding to range M-(N-1), a reference to range D-(E-1) (as indicated by the second entry in table 1430) may be sent to the replica storage array rather than sending the actual data of range M-(N-1).
If the second condition is not met, then the actual physical mapping that contains the data for the sector may be located by performing a lookup of the address translation table. Once the specific physical mapping has been located, then a lookup of physical to logical address mappings table 1460A may be performed to determine if the physical mapping is already stored in table 1460A. If the physical mapping is already stored in table 1460A, then a reference to the sector indicated by the corresponding entry of table 1460A may be sent to the replica storage array. In one embodiment, the reference may be in the form of <medium ID, range>. If the physical mapping is not already stored in table 1460A, then the actual data for the sector may be sent to the replica storage array and then this physical mapping may be added to table 1460A.
After the replica storage array receives a reference or data from the original storage array, the replica storage array may send an acknowledgement to the original storage array. In some cases, the replica storage array may batch acknowledgements and send multiple acknowledgements at a time rather than sending each acknowledgement individually. Alternatively, the replica storage array may send an acknowledgement in the form of “received all data up to medium X, offset Y”. When the original storage array receives an acknowledgment for a given extent, the original storage array may then add the given extent to list 1415A.
It is to be understood that only a portion of each of tables and lists 1400, 1415, 1430, and 1455 are shown, with the portion being relevant to the above discussion. It is noted that each of the tables and lists of
Turning now to
For example, the first extent of medium ID 1410 for range 0 to (A-1), corresponding to the first entry in table 1500, may be sent as a reference since this extent is already stored (as range 0 to (A-1) of medium ID 1425) on the replica storage array as indicated by the first entry of list 1415A. The second extent of medium ID 1410 may be sent as data since this extent does not map to an entry in list 1415A or duplicate extents table 1430. After the original storage array receives an acknowledgement from the replica storage array that is has received the data corresponding to the second extent of medium ID 1410, the original storage array may add this extent to list 1415 since this extent is now stored on the replica storage array. List 1415B represents list 1415 at the point in time after the original storage array receives the acknowledgement from the replica storage array regarding the second extent of medium ID 1410. Similarly, anytime an acknowledgement is sent by the replica storage array and received by the original storage array regarding a given extent, the given extent may be added to list 1415 at that time.
The third extent of medium ID 1410 may be sent as data since this extent does not map to an entry in list 1415B or duplicate extents table 1430. The fourth extent of medium ID 1410 may be sent as a reference to the third extent of medium ID 1410 since the fourth extent is the same as third extent as indicated by duplicate extents table 1430. The fifth extent of medium ID 1410 may be sent as data since this extent does not map to an entry in list 1415B or duplicate extents table 1430. Any number of extents after the fifth extent may be sent in a similar manner. Finally, the last extent of medium ID 1410 may be sent as a reference since this extent is the same as fifth extent as indicated by duplicate extents table 1430. After acknowledgements are received by the original storage array for the third and fifth extents of medium ID 1410, these extents may be added to list 1415. List 1415C represents list 1415 after these acknowledgements have been received by the original storage array.
Additionally, physical to logical address mappings table 1460 may be updated after the data for the extents of the second, third, and fourth entries is sent to the replica storage array. As shown in table 1460B, the physical address of the second entry (sector <1410, 1>) is represented as 1462X, the physical address of the third entry (sector <1410, 2>) is represented as 1463X, and the physical address of the fourth entry (sector <1410, 3>) is represented as 1464X.
A lookup of physical to logical address mappings table 1460B may be performed for subsequent entries of table 1500 prior to sending data to the replica storage array. Alternatively, in another embodiment, a list of recently sent physical addresses may be maintained. The size of the list of recently sent physical addresses may be as large or as small as desired, depending on the embodiment. If it is discovered that the address for a sector is located in table 1460B (or the list of recently sent physical addresses), then a reference to the previously sent sector may be sent to the replica storage array rather than the corresponding data. Also, if an address for a sector is already stored in table 1460B, fine-grained deduplication may be performed on these two sectors since they both point to the same physical address. This allows for an additional side benefit of the replication process of enabling fine-grained deduplication to be performed on the fly.
Referring now to
An original storage array may take a snapshot ‘M’ of a volume ‘V’ (block 1605). It is noted that block 1605 may only be performed if needed. For example, if M is already stable, then a snapshot does not need to be taken. Next, the original storage array may receive a request from a replica storage array ‘It’ for a list of snapshots (block 1610). The original storage array may respond to R with a list of available snapshots including M (block 1615). The original storage array may then receive an ID of a desired snapshot from R along with a list ‘A’ of medium extents that are already stored on R (block 1620). The original storage array may then use A and M, along with the medium extent table, to build rblocks of information to send to R (block 1625).
The original storage array may check to determine if all rblocks have been received by R (conditional block 1630). If all rblocks have been received by R (conditional block 1630, “yes” leg), then method 1600 is finished. If not all of the rblocks have been received by R (conditional block 1630, “no” leg), then the original storage array may send the next rblock not yet received by R (block 1635). Then, the original storage array may update the list of rblocks acknowledged by R (block 1640). After block 1645, method 1600 may return to block 1630. It is noted that replica storage array ‘R’ may also receive rblocks from one or more source storage arrays other than the original storage array. It is noted that the original storage array may retransmit rblocks which are not acknowledged.
Turning now to
The replica storage array ‘R’ may request a list of snapshots from the original storage array ‘0’ (block 1705). After receiving the list of snapshots, R may respond to 0 with the identity of the desired medium ‘M’ to replicate (block 1710). R may also send 0 a list of available medium extents which are already stored on R (block 1715). R may receive basic information (e.g., size) about the desired medium ‘M’ from 0 (block 1720).
R may determine if it has received all rblocks of M (conditional block 1725). If R has received all rblocks of M (conditional block 1725, “yes” leg), then method 1700 may be finished (block 1720). If R has not received all rblocks of M (conditional block 1725, “no” leg), then R may receive the next rblock from 0 or from another source storage array (block 1730). Then, R may acknowledge the received rblock (block 1735). Alternatively, R may perform bulk acknowledgements. After block 1735, method 1700 may return to block 1725.
Referring now to
The original storage array ‘0’ may generate a set of extents ‘Z’ that the replica storage array ‘R’ knows about (block 1805). A set of duplicate medium extents ‘D’ of the desired medium ‘M’ may also be generated (block 1810). This set D may include pairs of extents which map to the same underlying extent as well as pairs of extents that map to the same physical pointer value. Also, a set of physical to logical mappings ‘P’ may be initialized to empty (block 1815). Next, 0 may start traversing the medium mapping table for sectors of M (block 1820). When selecting a sector ‘s’ of the medium mapping table for medium ‘M’, 0 may generate a call to emit_sector for <M, s> (block 1825). The implementation of emit_sector is described below in method 1900 (of
After block 1825, O may determine if there are more sectors in ‘M’ (conditional block 1830). If there are more sectors in ‘M’ (conditional block 1830, “yes” leg), then a call to emit_sector for <M, s> may be generated for the next sector (block 1825). If there are no more sectors in ‘M’ (conditional block 1830, “no” leg), then method 1800 may end.
Referring now to
The original storage array ‘0’ may traverse the mapping table for <M, s> (block 1905). If <M, s> maps to sector <0, t> in Z (conditional block 1910, “yes” leg), then the reference from <M, s> to <0, t> may be emitted (block 1915). It is noted that ‘Z’ is the set of extents that the replica storage array ‘R’ already stores and which originated from O, and R may send a list of the set of extents Z to O. After block 1915, method 1900 may end.
If <M, s> does not map to sector <0, t> in Z (conditional block 1910, “no” leg), then it may be determined if <M, s> maps to sector <F, t> in duplicate medium extents ‘D’ (conditional block 1920). If <M, s> maps to sector <F, t> in D (conditional block 1920, “yes” leg), then a call to emit_sector for <F, t> may be generated (block 1925). After block 1925, the reference from <M, s> to <F, t> may be emitted (block 1930). After block 1930, method 1900 may end.
If <M, s> does not map to a sector <F, t> in D (conditional block 1920, “no” leg), then the physical address ‘X’ corresponding to <M, s> may be obtained from the address translation table (block 1935). Next, it may be determined if X is in the physical to logical mappings ‘P’ (conditional block 1940). The physical to logical mappings list ‘P’ is a list of physical to logical mappings corresponding to data that has already been sent to R. If X is in the physical to logical mappings ‘P’ (conditional block 1940, “yes” leg), then the sector <E, t> in P corresponding to X may be found (block 1945). Next, the reference from <M, s> to <E, t> may be emitted (block 1950). After block 1950, method 1900 may end.
If X is not in the physical to logical mappings ‘P’ (conditional block 1940, “no” leg), then the sector data corresponding to <M, s, contents_at_X> may be emitted (block 1955). After block 1955, the correspondence between address X and <M, s> may be stored in P (block 1960). After block 1960, method 1900 may end.
Referring now to
In one embodiment, a request to replicate a first medium from a first storage array to a second storage array may be generated (block 2005). The request may be generated by the first storage array or the second storage array, depending on the embodiment. It may be assumed for the purposes of this discussion that the first medium is already read-only. If the first medium is not read-only, then a snapshot of the first medium may be taken to make the first medium stable.
Next, in response to detecting this request, the first storage array may send an identifier (ID) of the first medium to the second storage array and request that the second storage array pull the first medium (or portions thereof) from any host to which it has access (block 2010). Alternatively, the first storage array may notify the second storage array that the first storage array will push the first medium to the second storage array. In one embodiment, the first medium may be identified based only by this medium ID. In one embodiment, the ID of the first medium may be a numeric value such as an integer, although the ID may be stored as a binary number. Also, in some embodiments, the age of a given medium relative to another medium may be determined based on a comparison of the IDs of these mediums. For example, for two mediums with IDs 2017 and 2019, medium ID 2017 has a lower ID than medium ID 2019, so therefore, it may be recognized that medium ID 2017 is older (i.e., was created prior to) than medium ID 2019.
After receiving the ID of the first medium and the request to pull the first medium from any host, it may be determined which regions of the first medium are already stored on the second storage array (block 2015). In one embodiment, the second storage array may identify regions which originated from the first storage array and which are already stored on the second storage array, and then the second storage array may send a list of these regions to the first storage array. The first storage array may then use this list to determine which regions of the first medium are not already stored on the second storage array. Then, the first storage array may send a list of these regions to the second storage array. In other embodiments, other techniques for determining which regions of the first medium are not already stored on the second storage array may be utilized.
After block 2015, the second storage array may pull regions of the first medium which are not already stored on the second storage array from other hosts (block 2020). For example, the second storage array may be connected to a third storage array, and the second storage array may send a list of regions it needs to the third storage array and request that the third storage array send any regions from the list which are stored on the third storage array. It is noted that in another embodiment, the above-described steps of method 2000 may be utilized for replicating the first medium from the first storage array to a cloud service rather than to the second storage array.
Referring now to
A request to replicate a first volume from a first storage array to a second storage array may be detected (block 2105). In one embodiment, the first storage array may decide to replicate the first volume to the second storage array. Alternatively, in another embodiment, the second storage array may request for the first volume to be replicated. In response to detecting the request to replicate the first volume, the first storage array may identify a first medium that underlies the first volume and make the first medium read-only (block 2110). In one embodiment, the first medium may be made read-only by taking a snapshot of the first volume. Next, the first storage array may send an identifier (ID) of the first medium to the second storage array along with a request to replicate the first medium (block 2115). In various embodiments, the request to replicate the first medium may be implicit or it may be an actual command. In some cases, the request to replicate the first medium may indicate if the first storage array will be pushing data to the second storage array, or if the second storage array will be pulling data from the first storage array and any other storage arrays. It may be assumed for the purposes of this discussion that the first storage array will be pushing data to the second storage array during the replication process. However, in other embodiments, the second storage array may pull data from the first storage array and other storage arrays.
The first storage array may request a list of any ancestors of the first medium which are already stored on the second storage array (block 2120). Alternatively, the first storage array may request a list of any read-only mediums which are older than the first medium. In one embodiment, the second storage array may identify mediums older than the first medium by selecting medium IDs which are lower than the first medium ID. For example, if the first medium ID is 1520, then the second storage array may identify all read-only mediums with IDs lower than 1520 which are stored on the second storage array. In a further embodiment, the first storage array may request an ID of the youngest read-only medium stored on the second storage array which is older than the first medium. If the first medium ID is 1520, then the second storage array would search for the highest medium ID which is less than 1520 and then send this ID to the first storage array. This ID may be 1519, 1518, 1517, or whichever medium ID is below and closest to 1520 and is stored in a read-only state on the second storage array.
In a further embodiment, the first storage array may request for the second storage array to identify the youngest ancestor of the first medium which is stored on the second storage array. For example, if the first medium ID is 2260, and if there are four ancestors of the first medium stored on the second storage array which are medium IDs 2255, 2240, 2230, and 2225, then the second storage array may identify medium ID 2255 as the youngest ancestor of medium ID 2260. It may be assumed for the purposes of this discussion that all ancestors of the first medium are read-only. In a still further embodiment, the first storage array may request for the second storage array to identify the youngest medium stored on the second storage array. For example, in one scenario, the second storage array may only store snapshots from a single volume, and so in that scenario, the most recent snapshot stored on the second storage array will be the youngest ancestor of the first medium.
Next, in response to receiving the request for a list of ancestors of the first medium which are already stored on the second storage array, the second storage array may generate and send the list to the first storage array (block 2125). In one embodiment, the second storage array may be able to determine the ancestors of the first medium after receiving only the ID of the first medium. For example, the second storage array may already know which volume is associated with the first medium (e.g., if the second storage array generated the replication request for the first volume), and the second storage array may have received previous snapshots associated with the first volume. Therefore, the second storage array may identify all previous snapshots associated with the first volume as ancestors of the first medium. In another embodiment, the first storage array may send an ID of each ancestor of the first medium to the second storage array along with the request in block 2120. Alternatively, in a further embodiment, rather than requesting a list of ancestors, the first storage array may request a list of any read-only mediums stored on the second storage array which are older (i.e., have lower ID numbers) than the first medium. It is noted that block 2120 may be omitted in some embodiments, such that the second storage array may generate and send a list of first medium ancestors (or the other lists described above) to the first storage array automatically in response to receiving a request to replicate the first medium.
In response to receiving the list of ancestors of the first medium which are already stored on the second storage array, the first storage array may use the list to identify regions of the first medium which are not already stored on the second storage array (block 2130). Then, the first storage array may send only these regions of the first medium to the second storage array (block 2135). It is noted that in another embodiment, the above-described steps of method 2100 may be utilized for replicating the first volume from the first storage array to a cloud service rather than to the second storage array.
It is noted that in the above description, it is assumed that when a medium ID is generated for a new medium, the most recently generated medium ID is incremented by one to generate the new medium ID. For example, medium ID 2310 will be followed by 2311, 2312, and so on for new mediums which are created. Alternatively, the medium ID may be incremented by two (or other numbers), such that medium ID 2310 will be followed by 2312, 2314, and so on. However, it is noted that in other embodiments, medium IDs may be decremented when new mediums are created. For example, the first medium which is created may get the maximum possible ID, and then for subsequent mediums, the ID may be decremented. In these other embodiments, the above described techniques may be modified to account for this by recognizing that lower IDs represent younger mediums and higher IDs represent older mediums.
Turning now to
Primary storage subsystem 2205 and secondary storage subsystem 2210 may also be coupled to cloud service 2230 via network 2225, wherein network 2225 is representative of any number and type of networks. Primary storage subsystem 2205 and secondary storage subsystem 2210 may each be configured to phone home log data (e.g., performance data, capacity data, system health data, diagnostics, past alerts) to cloud service 2230 at programmable intervals. In one embodiment, cloud service 2230 may be configured to analyze diagnostics, logs, and any additional performance data received from the storage subsystems and generate alerts based on the analysis. Cloud service 2230 may also be coupled to storage subsystems 2245A-N via network 2240, wherein network 2240 is representative of any number and type of networks. One or more of storage subsystems 2245A-N may also be configured to phone log data to cloud service 2230. Although not shown in
As shown alongside dataset 2215A, dataset 2215A may correspond to local medium ID 2272 at time N, local medium ID 2271 at time N-1, and any number of additional mediums prior to time N-1 corresponding to older snapshots of dataset 2215A. The local medium IDs 2271 and 2272 are local to primary storage subsystem 2205. Similarly, dataset 2215B may correspond to local medium ID 295 at time N, local medium ID 294 at time N-1, and any number of additional mediums prior to time N-1. The local medium IDs 294 and 295 are local to secondary storage subsystem 2210.
In one embodiment, cloud service 2230 may be configured to generate and maintain table 2255 for mapping local medium IDs to global medium IDs. Table 2255 is representative of any number and type of tables, lists, graphs, and/or other mapping structures which may be utilized to map local to global medium IDs for a plurality of storage subsystems. As shown in table 2255, global medium ID 2505 maps to local medium ID 2272 on primary storage subsystem 2205 and to local medium ID 295 on secondary storage subsystem 2210. When cloud service 2230 detects conditions indicative of an impending failure on either primary storage subsystem 2205 or secondary storage subsystem 2210, cloud service 2230 may be configured to determine which of storage subsystems 2245A-N is most suitable for taking over as a new secondary storage subsystem for dataset 2215A. In this scenario, cloud service 2230 may be configured to determine which given storage subsystem of storage subsystem 2245A-N would require the least amount of time to recreate dataset 2215A using the data already stored on the given storage subsystem. In one embodiment, cloud service 2230 may utilize table 2255 to determine which storage subsystem 2245A-N stores the most recent medium of dataset 2215A. For example, if cloud service 2230 determines that a first storage subsystem (e.g., storage subsystem 2245B) of storage subsystem 2245A-N stores a medium (e.g., local medium 547) which maps to global medium ID 2504 (corresponding to the snapshot of dataset 2215A taken at time N-1), this given storage subsystem would potentially be a good candidate for serving as a new secondary storage subsystem for dataset 2215A. However, if a second storage subsystem of storage subsystem 2245A-N only stores a medium corresponding to a snapshot of dataset 2215A taken two years ago, then the second storage subsystem would not be a good candidate for serving as a new secondary storage subsystem for dataset 2215A because of the amount of data which would need to be sent to the second storage subsystem to recreate dataset 2215A.
In one embodiment, cloud service 2230 may generate and maintain another table (not shown) tracking the mediums and their global medium IDs which correspond to dataset 2215A. Cloud service 2230 may generate and/or utilize this table when determining which of storage subsystems 2245A-N is most suitable for taking over as a new secondary storage subsystem for dataset 2215A. In one embodiment, cloud service 2230 may start with the most recent snapshot from this table and then search table 2255 for the most recent snapshot's global medium ID to determine if any of storage subsystems 2245A-N stores this snapshot. If none of storage subsystems 2245A-N store this snapshot, then cloud service 2230 may identify the next older snapshot of dataset 2215A and then search table 2255 for the global medium ID of the next older snapshot to determine if any of storage subsystems 2245A-N stores this snapshot. Cloud service 2230 may continue this process until it finds the most recent snapshot of dataset 2215A stored by one of storage subsystems 2245A-N.
It is noted that the above-described embodiments may comprise software. In such an embodiment, the program instructions that implement the methods and/or mechanisms may be conveyed or stored on a non-transitory computer readable medium. Numerous types of media which are configured to store program instructions are available and include hard disks, floppy disks, CD-ROM, DVD, flash memory, Programmable ROMs (PROM), random access memory (RAM), and various other forms of volatile or non-volatile storage.
In various embodiments, one or more portions of the methods and mechanisms described herein may form part of a cloud-computing environment. In such embodiments, resources may be provided over the Internet as services according to one or more various models. Such models may include Infrastructure as a Service (IaaS), Platform as a Service (PaaS), and Software as a Service (SaaS). In IaaS, computer infrastructure is delivered as a service. In such a case, the computing equipment is generally owned and operated by the service provider. In the PaaS model, software tools and underlying equipment used by developers to develop software solutions may be provided as a service and hosted by the service provider. SaaS typically includes a service provider licensing software as a service on demand. The service provider may host the software, or may deploy the software to a customer for a given period of time. Numerous combinations of the above models are possible and are contemplated.
Although the embodiments above have been described in considerable detail, numerous variations and modifications will become apparent to those skilled in the art once the above disclosure is fully appreciated. It is intended that the following claims be interpreted to embrace all such variations and modifications.
Claims
1. A method comprising:
- identifying portions of a dataset that are not already stored on a cloud-based target storage system, wherein the dataset is to be replicated from a source storage system; and
- replicating the identified portions of the dataset to the cloud-based storage target from an additional source storage system other than the source storage system.
2. The method of claim 1 further comprising receiving, by the cloud-based target storage system, a metadata identifier associated with the dataset.
3. The method of claim 1 further comprising sending, from the cloud-based target storage system to the source storage system, information describing portions of the dataset that are stored on the cloud-based target storage system.
4. The method of claim 3 further comprising receiving, from the source storage system and based on the information describing portions of the dataset that are stored on the cloud-based target storage system, information describing portions of the dataset that are not stored on the cloud-based target storage system.
5. The method of claim 4 further comprising sending, from the cloud-based target storage system to an additional source storage system, a request for one or more portions of the dataset, wherein the request in generated based on the information describing portions of the dataset that are not stored on the cloud-based target storage system.
6. The method of claim 5 further comprising receiving, by the cloud-based target storage system from the additional source storage systems, the one or more portions of the dataset.
7. The method of claim 1 wherein initiating replication of one or more of the portions of the dataset that are not already stored on the cloud-based target storage system from a source other than the source storage system is carried out in response to a determination that the source storage system has become unavailable.
8. The method of claim 1 wherein initiating replication of one or more of the portions of the dataset that are not already stored on the cloud-based target storage system from a source other than the source storage system is carried out in response to a detecting an impending failure of the source storage system.
9. A storage system that includes a plurality of storage devices, the storage system including a computer processor and a computer memory that includes computer program instructions which, when executed by the computer processor, cause the storage system to carry out the steps of:
- identifying portions of a dataset that are not already stored on a cloud-based target storage system, wherein the dataset is to be replicated from a source storage system; and
- replicating the identified portions of the dataset to the cloud-based storage target from an additional source storage system other than the source storage system.
10. The storage system of claim 9 further comprising computer program instructions which, when executed by the computer processor, cause the storage system to carry out the step of receiving a metadata identifier associated with the dataset.
11. The storage system of claim 9 further comprising computer program instructions which, when executed by the computer processor, cause the storage system to carry out the step of sending, from the cloud-based target storage system to the source storage system, information describing portions of the dataset that are stored on the cloud-based target storage system.
12. The storage system of claim 11 further comprising computer program instructions which, when executed by the computer processor, cause the storage system to carry out the step of receiving, from the source storage system and based on the information describing portions of the dataset that are stored on the cloud-based target storage system, information describing portions of the dataset that are not stored on the cloud-based target storage system.
13. The storage system of claim 12 further comprising computer program instructions which, when executed by the computer processor, cause the storage system to carry out the step of sending, from the cloud-based target storage system to an additional source storage system, a request for one or more portions of the dataset, wherein the request in generated based on the information describing portions of the dataset that are not stored on the cloud-based target storage system.
14. The storage system of claim 13 further comprising computer program instructions which, when executed by the computer processor, cause the storage system to carry out the steps of receiving, by the cloud-based target storage system from the additional source storage systems, the one or more portions of the dataset.
15. The storage system of claim 9 wherein initiating replication of one or more of the portions of the dataset that are not already stored on the cloud-based target storage system from a source other than the source storage system is carried out in response to a determination that the source storage system has become unavailable.
16. The storage system of claim 9 wherein initiating replication of one or more of the portions of the dataset that are not already stored on the cloud-based target storage system from a source other than the source storage system is carried out in response to a detecting an impending failure of the source storage system.
17. A non-transitory computer readable storage medium storing program instructions, wherein the program instructions are executable by a processor to:
- identifying portions of a dataset that are not already stored on a cloud-based target storage system, wherein the dataset is to be replicated from a source storage system; and
- replicating the identified portions of the dataset to the cloud-based storage target from an additional source storage system other than the source storage system.
18. The non-transitory computer readable storage medium as recited in claim 17, wherein the program instructions are further executable by a processor to send, from the cloud-based target storage system to the source storage system, information describing portions of the dataset that are stored on the cloud-based target storage system.
19. The non-transitory computer readable storage medium as recited in claim 18, wherein the program instructions are further executable by a processor to receive, from the source storage system and based on the information describing portions of the dataset that are stored on the cloud-based target storage system, information describing portions of the dataset that are not stored on the cloud-based target storage system.
20. The non-transitory computer readable storage medium as recited in claim 19, wherein the program instructions are further executable by a processor to send, from the cloud-based target storage system to an additional source storage system, a request for one or more portions of the dataset, wherein the request in generated based on the information describing portions of the dataset that are not stored on the cloud-based target storage system.
Type: Application
Filed: Sep 22, 2023
Publication Date: Jan 11, 2024
Inventors: BENJAMIN BOROWIEC (SAN JOSE, CA), STEVE HODGSON (COBHAM), ETHAN L. MILLER (SANTA CRUZ, CA)
Application Number: 18/472,817
