Abstract: A system maintains a consistency database that maintains a status (current, down, stale) for copies of logical storage volumes stored on storage nodes. As failures are detected, the consistency database is updated. Copies are synchronized with one another using information in the consistency database. Write operations on a primary node for a slice of a logical storage node are assigned a virtual block address (VBA) that is mapped to a logical block address (LBA) within the slice. Consistency of the VBAs of the primary node and that of a secondary node is evaluated and used to detect currency. VBA holes are detected and corresponding write commands resent to maintain currency. Physical segments on the primary node are assigned virtual segment identifiers (VSID) that are maintained consistent with VSIDs on clone nodes so that they can be used for garbage collection and synchronization.

Abstract: The present disclosure discloses a metadata recovery method and apparatus, the method includes: searching for a target node, where the target node has been damaged or index information recorded in the target node has been lost; and recovering the index information in the target node according to index information recorded in a related node, where the related node includes a parent node and a child node that are of the target node and an adjacent node belonging to a same layer as the node, and the index information includes brother index information pointing to the adjacent node belonging to the same layer as the node, parent index information pointing to the parent node of the node, and child index information pointing to the child node of the node. The present disclosure is mainly applied to a process of virtual data recovery.

Abstract: A method and system for managing backup storage of file system entities. In an aspect, a file system catalog includes a database populator tool that generates records within a metadata table that may be maintained within a database. In response to detecting a replication cycle, the populator tool reads a stream of replication operations. For each of the replication operations, the populator tool determines the type of operation and in response to determining that a directory inode is an operand of the replication operation, the tool generates one or more catalog records. Each of the generated records includes and logically associates data entries corresponding to an inode number, a parent inode number, an entity type, a point-in-time-image (PTI) ID, an absolute path, and an operation.

Abstract: A method for execution by an integrity processing unit includes determining to reclaim storage capacity of a DSN memory based on utilization information. Slice age information for a plurality of sets of encoded data slices stored in the DSN memory is determined, and a plurality of sets of encoded data slices are selected based on the slice age information. All encoded data slices are identified for deletion when substantially each of the selected plurality of sets includes less than or equal to a decode threshold number of encoded data slices. A subset of encoded data slices is identified for deletion when substantially each of the selected plurality of sets includes more than the decode threshold number of encoded data slices, where removal of the subset of encoded data slices results in at least the decode threshold number of encoded data slices remaining for each selected set of encoded data slices.

Abstract: (A) and (B), described hereinafter, are carried out for objects which differ between an reflection source namespace and a reflection destination namespace, with respect to either: a center node in a migration, which applies an update of a first local namespace which a first edge node (any edge node) provides to a center namespace; or a second edge node (any edge node other than the first edge node) in a synchronization, which applies an update of the center namespace to a second local namespace which the second edge node provides. Specifically, either the center node or the second edge node: (A) determines whether the situation with respect to the differing objects corresponds to a conflict case among defined conflict cases; and (B) if the result of the determination in (A) is affirmative, executes a process according to a confliction resolution policy which is associated with the corresponding conflict case.

Abstract: A system and method for providing data protection services and lifecycle management in a Platform as a Service (PaaS) platform having cloud-based applications and data, by discovering an application installed on the PaaS platform using mechanisms native to the PaaS platform, associating the application and an associated data service used by the application to identify data created by the application, and performing a backup/restore operation on the application and the identified data in accordance with a defined schedule. The native mechanisms may comprise application programming interfaces (APIs) provided by the PaaS platform, or metadata associated with containers used by the PaaS platform. The schedule may be defined by RTO and RPO requirements of a user.

Abstract: A computer-implemented method for synchronizing a first data object to a second data object, wherein, each data object is assigned to a metaclass from a number of metaclasses, the metaclasses being organized in a tree structure with a root metaclass; a metaclass in the tree structure is ascertained that is situated on the path between the first data object and the root metaclass and on the path between the second data object and the root metaclass, a first and a second metaobject are produced from the ascertained metaclass; two transformations are performed to transform the first data object into the first metaobject and the second data object into the second metaobject; the first metaobject and the second metaobject are synchronized; and a third transformation is performed, the first metaobject being transformed into the first data object and the second metaobject being transformed into the second data object.

Abstract: Computer program products, as well as corresponding systems and methods are configured for performing deduplication in conjunction with random read and write operations, and include: receiving a write request comprising data; computing a fingerprint of the data; determining whether a short term dictionary comprises an entry corresponding to the fingerprint; in response to determining the short term dictionary comprises the entry corresponding to the fingerprint, writing the data to a data store in a deduplicating manner; in response to determining the short term dictionary does not comprise the entry, determining whether a long term dictionary corresponding to the namespace comprises the entry; in response to determining the long term dictionary comprises the entry, writing the data to the data store in the deduplicating manner; and in response to determining the long term dictionary does not comprise the entry, writing the data to the data store in a non-deduplicating manner.

Abstract: A system and methods for backing up and restoring updates to metadata of files stored in a database are disclosed. One example embodiment of the method includes entering the updates to the metadata of the one or more files into a first database table; generating a backup file of the first database table; and restoring the updates to the metadata to the one or more files using the backup file. Restoring the updates may include creating a recovery table populated with data from the backup file; determining the one or more files to apply the updates to by searching the recovery table for an identifier of the one or more files; and adding the updates to the metadata from the recovery table to the corresponding one or more files that matches the identifier identified from the recovery table.

Abstract: A method, computer-readable medium, and device for processing a stream of records are disclosed. A method may receive a registration request from a data source to join in providing a plurality of records to the stream, synchronize a timing system with the data source, receive an initial timestamp from the data source, and post the initial timestamp to a plurality of operators or a plurality of subscribers before the data source is allowed to send records into the stream.

Abstract: Bypassing backup operations for specified database types is described. A system determines whether a database, of multiple databases associated with a database system, corresponds to a database type specified in a configuration file, in response to receiving a backup operation request. The system executes an incremental backup operation on changes made to the database since a selected time, for the backup operation request, if the database does not correspond to the database type specified in the configuration file. The system bypasses executing any backup operation on the database for the backup operation request if the database corresponds to the database type specified in the configuration file.

Abstract: A payment service is configured to support purchase transactions generated by merchant point-of-sale (POS) devices, wherein each purchase transaction may comprise multiple transaction requests such as authorization requests and capture requests. Each transaction request may be sent to a different one of multiple data centers of the payment service. Upon successfully receiving an acknowledgement from one of the data centers for a transaction request, the POS device designates that data center as the primary data center. Subsequently, all transaction requests for that transaction identify the primary data center, even when the transaction requests are sent to other data centers. At the payment service, the data centers synchronize requests such that the primary data center is informed of all authorizations and is made responsible for performing the capture of the transaction. This allows the primary data center to detect duplicate authorizations and to prevent duplicate captures.

Abstract: A persistent data structure store may include an architecture that utilizes dual representations of data hosted in persistent storage coupled with a faster non-persistent cache residing in volatile memory. The persistent data structure store may maintain, in a non-volatile persistent memory, a byte-addressable persistent copy of a data structure storing multiple data values. The persistent data structure store may also maintain a copy of at least some of the data values in a cache copy of the data structure in a volatile memory. When updating values in the data structure, the targeted data value may be updated in the cache copy of the data structure and a log entry specifying the data update may be added to a cross-referencing data operations log from which the data operation will be subsequently applied to the persistent copy of the data structure.

Abstract: In accordance with embodiments of the present invention, a method for associating metadata with a media object is provided. The method provides the ability to tag, or bookmark, a point in time for future use. The method includes receiving the metadata, an associated time condition, and an associated user identification. The method further includes storing at least the time condition. The at least stored time condition is used, at least in part, for associating the metadata with the media object. The media object is then provided to the user. In some embodiments the media object is not available for association with the metadata at the time the metadata is received. In other embodiments, the media object is provided by an external application.

Abstract: A first state identifier is determined according to the state of a first file at a first time point, and a second state identifier is determined according to the state of a second file at a second time point. The first and second state identifiers are compared and an alignment indication is displayed to a user of a client device thereby making the user aware of the alignment state between the files. The files may be associated with different client devices or different users thereby showing group alignment. The files may also be associated with a single user or client device thereby showing local alignment. The state identifiers may include digest hashes of content of the first and second files, metadata, or state information regarding the files. The files may be encrypted and access check keys utilized to determine whether a client device is authorized to receive an encrypted file.

Abstract: One or more techniques and/or computing devices are provided for non-disruptively establishing a synchronous replication relationship between a primary volume and a secondary volume and/or for resynchronizing the primary volume and the secondary volume. For example, a baseline snapshot and one or more incremental snapshots of the primary volume are used to construct and incrementally update the secondary volume with data from the primary volume. A dirty region log is used to track modifications to the primary volume. A splitter object is used to split client write requests to the primary volume and to the secondary volume. A synchronous transfer engine session is initiated to processing incoming client write requests using the dirty region log. A cutover scanner is used to transfer dirty data from the primary volume to the secondary volume. In this way, a synchronous replication relationship is established between the primary volume and the secondary volume.

Abstract: A deep link can have a name of a mobile app and a network address of a device on which an instance of the app is running. Deep links can be stored in a blockchain, along with related data, like the location of the device running the app. The deep links in the blockchain can be analyzed to make a social network graph. A node in the graph is a network address of a mobile device running a mobile app. An edge between 2 nodes is a mobile app, which had 2 instances of the app interacting, where each instance runs on one of the nodes. The edge could be directed, going from the first earlier instance to the second instance. A blockchain can be used by a mobile app to write a deep link referring to itself. The app instance is looking to interact with other instances of the app. Another device scans the blockchain to get the deep link. It makes an instance of the app that interacts with the instance that wrote the deep link to the blockchain.

Abstract: Techniques for processing a speech to text query are described herein. The techniques may include receiving a plurality of speech to text translation alternatives for a phrase of a natural language query, and tagging and parsing each of the translation alternatives based on a static analysis of the known domain that is at least partially structured, known tags of the known domain, and custom rules. The techniques may also include ranking the translation alternatives based on the tagging and parsing and translating the phrase based on the ranking.

Abstract: Disclosed herein are distributed ledger systems and methods for efficiently creating and updating a query optimized distributed ledger. In particular, the present disclosure introduces methods and apparatuses for efficiently updating indexes when new blocks are added to the distributed ledger by using snapshots of data and appending new snapshot tables and indexes to previous snapshot tables and indexes.

Abstract: A method for executing multiple versions of an application within a computing environment. The method includes a computer processor identifying a request to execute an instance of a first version of an application within a computing environment and determining that an instance of a second version of the application is active within the computing environment. The method further includes accessing a data structure including information associated with the application to obtain a first set of data corresponding to the first version of the application. The method further includes copying one or more executable objects associated with the first version of the application based, at least in part, on the first set of data to the computing environment. The method further includes executing the copied one or more executable objects associated with the first version that comprise the requested instance of the first version of the application within the computing environment.

Abstract: Disclosed is a source code transfer control method, a computer program therefor, and a recording medium therefor. The source code transfer control method is a method executed in a configuration management system interworking with a static analyzer server. The method includes: (a) receiving a source code transfer request; (b) transmitting file identification information on a source code that is requested for transfer, to a static analyzer server; (c) receiving a return value from the static analyzer server on the basis of the file identification information, the return value being produced from an analysis result associated with the transfer-requested source code and loaded on the static analyzer server, the return value being a value produced for each check item that is preset for transfer control; and (d) providing information on whether the transfer-requested source code can be normally transferred on the basis of the return value.

Abstract: This disclosure relates to providing information access in an enterprise network. The method includes creating automatically an ontology by analyzing at least one document comprising semantic information for roles, responsibilities, authority, and restrictions associated with a plurality of users; extracting based on the ontology a plurality of contextual terms associated with at least one of roles, responsibilities, authority, or restrictions; assigning a plurality of annotation tags to each sentence in at least one enterprise policy document based on the plurality of contextual terms; generating a plurality of information access rules based on the plurality of contextual terms and assigned plurality of annotation tags to each sentence in the at least one enterprise policy document; and creating a context similarity tree based on the assigned plurality of annotation tags and the plurality of information access rules.

Abstract: Techniques for processing a speech to text query are described herein. The techniques may include receiving a plurality of speech to text translation alternatives for a phrase of a natural language query, and tagging and parsing each of the translation alternatives based on a static analysis of the known domain that is at least partially structured, known tags of the known domain, and custom rules. The techniques may also include ranking the translation alternatives based on the tagging and parsing and translating the phrase based on the ranking.

Abstract: Methods, apparatus, and other embodiments facilitate transitioning a database between two modes. The two modes include a mode where table information for separate tables in the database is stored in a single monolithic file and a mode where the table information for separate tables is stored in separate files. Conventional attempts to make this type of transition lock the database for an unacceptable amount of time. Example methods and apparatus make a copy of a database that is to be transitioned, lock the copy, make a second instance of the copy, and then perform the work necessary for the transition in the second instance. This allows the original database to remain running. When the second instance is complete, the second instance is synchronized to the original database and then the transition is completed by copying the second instance to the original or failing the original over to the second instance.

Abstract: A logical storage layer for shared storage systems interposes between address ranges of the shards and the storage devices on which the shards are stored. The shards may be logically addressed using a plurality of addressable zones, to which the storage devices are independently mapped. Data requests related to a given piece of data associated with a shard may involve multiple storage devices, and vice versa.

Abstract: A method, a master database node and a subscriber database node for optimizing index. The method for optimizing index includes: checking, by a master database node, if role information of an index matches role information of the master database node, when the index is added into the master database node; updating database information by using the index when the role information of the index matches the role information of the master database node. In this invention, unnecessary indexes in the master database node and the subscriber database node are reduced while keeping the schema the same. Furthermore, unnecessary logs transmitted from the master database node to the subscriber database node are reduced.

Abstract: Various systems and methods for modifying the contents of a snapshot after the snapshot has been created are disclosed. One such method involves attaching a volume filter driver to a snapshot volume. The method detects one or more modifications intended for data in a read-only snapshot. Upon detecting the modification(s) intended for data in the read-only snapshot, the method performs a write operation, which includes the modification(s), to a secondary storage device instead of a primary storage device.

Abstract: Embodiments of the present invention provide systems and methods for enabling cross object synchronization. An algorithm is applied in order to obtain a consistent view of multiple Kafka topics. One set of objects are associated with a first Kafka topic and another set of objects are associated with a second Kafka topic. The first Kafka topic and the second Kafka topic are processed as a single Kafka topic which leads to common transactional coordinate points associated with the first and second Kafka topics. The common transaction coordinate point is indicative of a consistent view of the multiple Kafka topics.

Abstract: A host that is coupled to a memory device is configured to read a status register of the memory device to determine if the memory device supports host controlled enablement of automatic background operations. The memory device responds to the host regarding whether the memory device supports host controlled enablement of automatic background operations. The host can enable the automatic background operations if the memory device supports this feature. The host can then set a time period in the memory device that is indicative of when the memory device can automatically perform the background operations.

Abstract: A transactional replicator applying group commit and barrier concepts is disclosed. Group commit means that the transactional replicator commits multiple transactions in a group and is not restricted to committing single transactions one-at-a-time and is not limited to operating on a single state provider. Barrier means that the transactional replicator does not move forward to commit additional transactions until the previous group of transactions are completed. All state providers must apply their transactions and update state before additional transactions will be committed. A quorum acknowledgement “unlocks” any locks that were acquired to update the state within a state provider. However, as long as there are no lock conflicts, additional transactions can continue to create new group commits as necessary.

Abstract: A computer-implemented method for operating a database management system which includes a source database and a target database. A database management system is configured for asynchronously replicating data changes of the source database to the target database. The database management system receiving a request to perform a database statement. The method can identify the most recently committed one of a plurality of transactions having been committed in the source database before the receiving of the request. Identifying repeatedly the one of the plurality of transactions having been replicated most recently to the target database. Execution of the requested statement in the target database is delayed until an evaluation of the repeatedly identified most recently replicated transaction returns that the identified most recently committed transaction has been replicated to the target database. After the delaying execution, the method executes the requested statement in the target database.

Abstract: A computer-implemented method for asynchronous duplexing between two computing devices includes obtaining a task for execution by a first computing device and executing the task. A first computing task completion identifier is assigned upon execution of the task. The task with the assigned first computing task completion identifier is transferred to a second computing device and executed. A second computing task completion identifier is assigned upon execution of the task. The first computing task completion identifier and the second computing task identifier are logged into a task management log for facilitating the recovery of the first computing device by comparing the first computing task completion identifier and the second computing task identifier for identifying tasks executed by the first computing device and requiring execution by the second computing device.

Abstract: Provided are a computer program product, system, and method for using a forward log storage and backward log storage to recover a storage to a forward or backward point-in-time. In response to receiving writes to source data after establishing point-in-time copies, point-in-time data of the source data is copied to a backward log storage storing point-in-time data for multiple of the point-in-time copies. The point-in-time data in the backward log storage is applied to a recovery source data to roll-back the source data to a backward point-in-time of one of the point-in-time copies. Before applying the point-in-time data from the backward log storage, point-in-time data in the recovery source data, is copied to a forward log storage. The point-in-time data in the forward log storage is applied to the recovery source data to roll forward the recovery source data to a forward point-in-time subsequent to the backward point-in-time.

Abstract: Systems and methods are provided for storing a first data object comprising a first set of immutable components, the first data object being associated with a corresponding second data object stored by a remote replication system. A difference is determined between the first set of immutable components of the first data object and a second set of immutable components of the corresponding second data object. A subset of immutable components is identified from the first set of immutable components based on the difference. The subset of immutable components from the first set of immutable components is provided to the remote replication system over a communication network.

Abstract: In an approach for processing data in a database management system, a processor performs a delete operation on a first data container, wherein the first data container includes one or more data records and wherein the delete operation logically deletes at least one data record of the one or more data records. A processor associates a time of the performed delete operation with the first data container. A processor performs a write operation on a first data record of the first data container. A processor associates a time of the performed write operation with the first data record of the first data container. Responsive to receiving a database query directed at the first data container, a processor grants access to the first data record of the first data container based on, at least, the time of the performed write operation being chronologically after the time of the performed delete operation.

Abstract: A system includes transmission of a first transaction identifier and a first commit identifier to a first data storage system, the first commit identifier identifying a first data snapshot of the first data storage system, transmission of a first query to the first data storage system, transmission of a first prepare instruction and the first transaction identifier to the first data storage system, determination that a first ready response has been received from the first data storage system in response to the first prepare instruction, transmission, in response to the determination, of a first commit instruction and the first transaction identifier to the first data storage system, and reception of a second commit identifier from the first data storage system, the second commit identifier identifying a second data snapshot of the first data storage system.

Abstract: A distributed file system includes a plurality of computing devices that are coupled to a plurality of storage devices. A first computing device transfers objects to an additional computing device. The first computing device determines times when multiple computing devices, including the first computing device, transferred objects to the additional computing device. The first computing device identifies the earliest of the times and may store this information on the additional computing devices. A second computing device resumes transferring data to the additional computing device when the first computing device becomes unavailable or inoperable. The second computing device may use the earliest of the times to determine which objects should be transferred to the additional computing device.

Abstract: An online system enables content providers to efficiently create and publish sponsored content in multiple phases. Each phase of the content creation is represented by a draft of the content. Multiple drafts of the content are stored in a repository of the online system for optimizing content delivery and publication, such as quick identification of conflicts in a draft based on analysis of draft fragments of the draft. Objects associated with the sponsored content are stored in an object model, including a key-value table. The connections amongst the various draft segments of a draft are stored in a hierarchy structure, e.g., a tree structure. Each user associated with a draft has a role, such as an owner or a collaborator, where all of the roles can contribute to the draft. The online system efficiently publishes the drafts of the sponsored content and presents the sponsored content to its target users.

Abstract: A computer system includes a server having a memory connected thereto. The server is adapted to be connected to a network to permit remote storage and retrieval of data files from the memory. A file identification application is operative with the server to identify errant files stored in the memory. The file identification application provides the functions of: (1) selecting a file stored in said memory; (2) generating a unique checksum corresponding to the stored fire; (3) comparing said unique checksum to each of a plurality of previously generated checksums, wherein the plurality of previously generated checksums correspond to known errant files; and (4) marking the file for deletion from the memory if the unique checksum matches one of the plurality of previously generated checksums.

Abstract: Data to be moved from a source system to a target system, for a set of tenants, is first identified. The data is enumerated by a first computing instance in the source system to obtain an enumeration list. Data is copied from the source system to the target system based on the enumeration list by a second computing instance. The data in the source and target systems is then enumerated by a third computing instance to determine whether any data is still to be moved and another enumeration list is generated. The data still to be moved is then moved based on the other enumeration list.

Abstract: The present invention is notably directed to a method for synchronizing proprietary data in an external cloud provided by a cloud service provider with data of a private storage system. The method comprises, at a synchronization system: copying outward data from a flow of outward data sent from a private cloud to the external cloud, the outward data being proprietary data of an entity that owns data of the private cloud. Next, and in parallel to copying outward data: the synchronization system compares copied outward data with data stored on the private storage system, to determine whether the compared outward data are already replicated in the private storage system. Finally, if it is determined that the compared outward data are not yet replicated in the private storage system, it instructs to store the compared outward data on the private storage system. The present invention is further directed to related systems and computer program products.

Abstract: The disclosed embodiments generally relate to systems and methods for computerized decisioning, and more particularly, to systems and methods for standardized coding and deployment of attributes for financial decisioning systems coded once and deployed to multiple decisioning systems and environments. Consistent with disclosed embodiments, an attribute management system may include one or more memory devices storing instructions, and one or more hardware processors configured to execute the instructions to perform operations. Upon executing the instructions, the processors may receive a financial service request, and identify a decisioning system for processing the request. The processors may determine that an attribute is required for processing the request using the decisioning system. Then, the processors may initiate an attribute calculation service for determining an attribute value for the attribute.

Abstract: Techniques to provide access to repository content, via a file sharing service, are disclosed. In various embodiments, a revised content associated with a repository content object, reflecting a modification made to the repository content in the context of a file sharing service to which the repository content object was published in connection with a synchronized share of the repository object, is received. It is determined whether the revised content is to be stored in the repository and, if so, the revised content is stored in the repository.

Abstract: With respect to reconciling data between a source system and one or more destination systems, a source agent obtains data from the source system. A destination agent obtains data from the destination system. The source and destination agents generate digests of the data obtained from the source and destination systems, respectively. The digests are provided to a consistency service for storage. The digests of the data stored at the source system are stored in a first tree data structure, such as a variant of a Merckle tree. The digests of the data stored at the destination system are stored in a second tree data structure, which might also be a variant of a Merckle tree. A comparison of the trees may be made in order to determine the consistency level of the data stored at the source system and the destination systems.

Abstract: A system and method includes employing a cyclical redundancy check (CRC) code to information being transmitted on the channel, the CRC code including appending a single CRC code bit to the information, the single CRC code bit having a value of 0. The method further includes adding an error correcting code with a value of 0 to the information and to a plurality of error correcting code parity bits provided by the error correcting code. The method further includes transmitting, by the communication device, the information on the channel without the single CRC code bit, the additional information bit, and the error correcting code parity bits for being decoded and set by a receiver.

Abstract: Incrementally backing up file system hard links based on change logs is described. A system identifies a unique identifier and a file name associated with a file event in a file system. The system determines whether a change log for the file system lacks an association between the unique identifier and the file name. The system adds the file name to one of a new file name list and a deleted file name list associated with the change log in response to a determination that the change log for the file system lacks the association between the unique identifier and the file name. The system incrementally backs up the file event based on at least one of the new file name list and the deleted file name list.

Abstract: A method includes receiving a change in a primary configuration of a primary storage node while the primary storage node is operational, wherein primary data stored in the primary storage node is to be backed up as backup data in a backup storage node. The method includes locating at least one identifier in the change that identifies the primary storage node. The method includes replacing the at least one identifier in the change that identifies the primary storage node with an identification of the backup storage node to create an updated configuration change. The method includes updating a backup configuration of the backup storage node based on the updated configuration change.

Abstract: The system receives a list of one or more name prefixes associated with an original name, wherein the original name corresponds to an original root manifest that indicates a set of original content objects and includes a set of content object hash (COH) values for the indicated set of original content objects, and wherein the original root manifest is registered with a tracking service. Based on a name prefix from the list, the system receives the original root manifest and selects a peer node that stores one or more of the original content objects. The system determines which original content objects are stored at the selected peer node and generates an interest for an original content object, wherein the name for the interest includes the name prefix, and wherein the third interest includes a COH value for an original content object stored at the selected peer node.

Abstract: A storage drive and associated system and method for storing data locally at the storage drive and/or at cloud storage may buffer, in a local storage buffer, data to be stored at the cloud and upload the data to the cloud storage from the local storage buffer. Upon receipt of an acknowledgement, from the cloud storage, that the data was successfully uploaded, the storage drive may delete the data from the local storage buffer. The storage drive may provide a ghost file representation of the uploaded data through the storage drive, through which the data uploaded to the cloud storage may be accessed. The storage drive may access multiple cloud storage accounts, each corresponding to a file folder on the storage drive, allowing the user to easily access different cloud storage accounts. The storage drive may recover data that was deleted from either local persistent storage or the cloud storage.

Abstract: In general, the subject matter described in this disclosure can be embodied in methods, systems, and program products. A system includes a first data center, a second data center, and a third data center. The multiple data centers are configured to replicate a logical collection of data that comprises multiple logical partitions of data. The system comprises a first writing subsystem that is designated to write updates to a copy of a first logical partition of data that is stored by the first data center. The system comprises a second writing subsystem that is designated to write updates to a copy of a second logical partition of data that is stored by the second data center. The system comprises a third writing subsystem that is designated to write updates to a copy of a third logical partition of data that is stored by the third data center.