Abstract: Machines, systems and methods for I/O monitoring in a plurality of compute nodes and a plurality of service nodes utilizing a Peripheral Component Interconnect express (PCIe) are provided. In one embodiment, the method comprises assigning at least one virtual function to a services node and a plurality of compute nodes by the PCIe interconnect and a multi-root I/O virtualization (MR-IOV) adapter. The MR-IOV adapter enables bridging of a plurality of compute node virtual functions with corresponding services node virtual functions. A front-end driver on the compute node requests the services node virtual function to send data and the data is transferred to the services node virtual function by the MR-IOV adapter. A back-end driver running in the services node receives and passes the data to a software service to modify/monitor the data. The back-end driver sends the data to another virtual function or an external entity.

Abstract: For block level backup, a full image level backup is created of a file system from production storage at a production server to random access storage at a backup server. A mounted image is created from the full image that exposes a block level representation of the file system of the production server for read operations in a file level format. File enumeration is performed on the mounted image for files. Block level read operations are monitored while the files are being enumerated. During monitoring, blocks that are read for the read operations are marked as meta-data blocks. Blocks that are not marked as meta-data blocks are identified as data blocks. The data blocks are migrated to sequential storage, while leaving the meta-data blocks on the random access storage, wherein the data blocks form a backup image.

Abstract: A mechanism is provided for efficiently searching a database for data associated with a key value that is unidentifiable in a set of nodes of the database. A request is received to retrieve data from the database, where the request specifies a searched-for key value that is unidentifiable in the set of nodes of the database. The data associated with the searched-for key value is identified in a lowest level node of the database between a current leftmost node and a current rightmost node. A key index is updated in at least one node of the database with the key index value associated with the data and identified from the lowest level node thereby making the searched-for key value identifiable within the at least one node. The data associated with the searched-for key value is then returned to a user.

Abstract: For block level backup, a full image level backup is created of a file system from production storage at a production server to random access storage at a backup server. A mounted image is created from the full image that exposes a block level representation of the file system of the production server for read operations in a file level format. File enumeration is performed on the mounted image for files. Block level read operations are monitored while the files are being enumerated. During monitoring, blocks that are read for the read operations are marked as meta-data blocks. Blocks that are not marked as meta-data blocks are identified as data blocks. The data blocks are migrated to sequential storage, while leaving the meta-data blocks on the random access storage, wherein the data blocks form a backup image.

Abstract: A mechanism is provided for increasing the efficiency of block-level processes. Responsive to detecting an I/O write to a storage volume, a determination is made as to whether the I/O write matches one or more suspicious I/O reads in a plurality of suspicious I/O reads previously recorded in a memory. Responsive to the I/O write matching one or more suspicious I/O reads in the plurality of suspicious I/O reads, an identification is made that there is a data relocation associated with the sector associated with the I/O write. A data relocation application is informed of the data relocation of the sector associated with the one or more suspicious I/O reads associated with the sector indicated by the I/O write. Each sector associated with the one or more suspicious I/O reads is then released in a file system for reuse.

Abstract: A method according to one embodiment includes receiving, at an I/O Handler, an instruction to initiate a backup operation on data associated with an application running on multiple servers; and stretching communication between instances of the application and data storage volumes associated therewith during initiating the backup operation. Additional systems, methods, and computer program products are also disclosed.

Abstract: Embodiments of the present invention provide a scalable, efficient way to backup data in a block-level incremental-forever backup system such that backup and expiration of data can be achieved at the granularity of a single backup version, without having to read or move data that is stored in backup storage.

Abstract: A mechanism is provided for efficiently searching a database for data associated with a key value that is unidentifiable in a set of nodes of the database. A request is received to retrieve data from the database, where the request specifies a searched-for key value that is unidentifiable in the set of nodes of the database. The data associated with the searched-for key value is identified in a lowest level node of the database between a current leftmost node and a current rightmost node. A key index is updated in at least one node of the database with the key index value associated with the data and identified from the lowest level node thereby making the searched-for key value identifiable within the at least one node. The data associated with the searched-for key value is then returned to a user.

Abstract: For block level backup, a full image level backup is created of a file system from production storage at a production server to random access storage at a backup server. A mounted image is created from the full image that exposes a block level representation of the file system of the production server for read operations in a file level format. File enumeration is performed on the mounted image for files. Block level read operations are monitored while the files are being enumerated. During monitoring, blocks that are read for the read operations are marked as meta-data blocks. Blocks that are not marked as meta-data blocks are identified as data blocks. The data blocks are migrated to sequential storage, while leaving the meta-data blocks on the random access storage, wherein the data blocks form a backup image.

Abstract: For block level backup, a full image level backup is created of a file system from production storage at a production server to random access storage at a backup server. A mounted image is created from the full image that exposes a block level representation of the file system of the production server for read operations in a file level format. File enumeration is performed on the mounted image for files. Block level read operations are monitored while the files are being enumerated. During monitoring, blocks that are read for the read operations are marked as meta-data blocks. Blocks that are not marked as meta-data blocks are identified as data blocks. The data blocks are migrated to sequential storage, while leaving the meta-data blocks on the random access storage, wherein the data blocks form a backup image.

Abstract: Read-ahead of data blocks in a storage system is performed based on a policy. The policy is stochastically selected from a plurality of policies in respect to probabilities. The probabilities are calculated based on past performances, also referred to as rewards. Policies which induce better performance may be given precedence over other policies. However, the other policies may be also utilized to reevaluate them. A balance between exploration of different policies and exploitation of previously discovered good policies may be achieved.

Abstract: Providing shared access to data storage resources in a cluster computing environment, including detecting when a computer node requests access to a data storage device in a manner that excludes a computing device from accessing the data storage device, where the computer node is configured to communicate with the data storage device along a first communications channel, and where the computing device is configured to communicate with the data storage device along a second communications channel that is exclusive of the first communications channel, and responsive to the detecting, causing the data storage device to allow the computer node and the computing device to share access to the data storage device.

Abstract: A method according to one embodiment includes receiving, at an I/O Handler, an instruction to initiate a backup operation on data associated with an application running on multiple servers; and stretching communication between instances of the application and data storage volumes associated therewith during initiating the backup operation. Additional systems, methods, and computer program products are also disclosed.

Abstract: Multiple target blocks are allocated on a first storage device to store a target object. The target blocks are arranged to be in a one-to-one correspondence with multiple source blocks of a source object. The target blocks are set to a non-populated state, and target blocks in the non-populated state are populated with data from corresponding source blocks. While the target blocks are being populated, if a request is received to retrieve data from one of the target blocks that is in the non-populated state, then the one of the target blocks is populated with the data from the corresponding source block and set to a populated state.

Abstract: Multiple target blocks on a first storage device are allocated to store a target object. The target blocks are arranged to be in a one-to-one correspondence with multiple source blocks of a source object. The target blocks are set to a non-populated state, and target blocks in the non-populated state are populated with data from corresponding source blocks. While the target blocks are being populated, if a request is received to retrieve data from one of the target blocks that is in the non-populated state, then the one of the target blocks is populated with the data from the corresponding source block and set to a populated state.

Abstract: A method according to one embodiment includes receiving a request to perform a backup of data associated with an application running on multiple servers; communicating with I/O Handlers on the servers for initiating a coordinated backup operation on the data at about a same start time; and instructing the I/O Handlers to stretch communication between instances of the application and data storage volumes associated therewith during initiating the backup operation. Additional systems, methods, and computer program products are also disclosed.

Abstract: Read-ahead of data blocks in a storage system is performed based on a policy. The policy is stochastically selected from a plurality of policies in respect to probabilities. The probabilities are calculated based on past performances, also referred to as rewards. Policies which induce better performance may be given precedence over other policies. However, the other policies may be also utilized to reevaluate them. A balance between exploration of different policies and exploitation of previously discovered good policies may be achieved.

Abstract: An exposed Internet Small Computer System Interface (iSCSI) target of a backup repository is defined as a source of, and a storage volume to be recovered as a target of, a mirrored pair of storage resources. The iSCSI target represents a point-in-time backup snapshot. Data from the iSCSI target is synchronized to the storage volume. The mechanisms of the illustrated embodiments leverage iSCSI with Instant Restore to reduce a Recovery Time Objective (RTO) and provide a secure transport when recovering a volume over a WAN.

Abstract: A system according to one embodiment includes at least one processor configured to: receive a request to perform a snapshot of data associated with an application running on multiple servers; instruct I/O Freezers on different servers to freeze data requests from instances of the application running on the servers; and instruct the I/O Freezers to initiate snapshot processing on the data associated with the application so that the I/O Freezers perform the snapshot processing within the same window of synchronization.

Abstract: Read-ahead of data blocks in a storage system is performed based on a policy. The policy is stochastically selected from a plurality of policies in respect to probabilities. The probabilities are calculated based on past performances, also referred to as rewards. Policies which induce better performance may be given precedence over other policies. However, the other policies may be also utilized to reevaluate them. A balance between exploration of different policies and exploitation of previously discovered good policies may be achieved.