Abstract: Encryption of data occurs before it is written to the storage platform; decryption occurs after it is read from the storage platform on a computer separate from the storage platform. By encrypting data before it travels over a wide-area network to a storage platform (and by only decrypting that data once it has arrived at an enterprise from the storage platform), we address data security over the network. Application data is encrypted at the virtual disk level before it leaves a controller virtual machine, and is only decrypted at that controller virtual machine after being received from the storage platform. Encryption and decryption of data is compatible with other services of the storage system such as de-duplication. Any number of key management services can be used in a transparent manner.

Abstract: A client machine writes to and reads from a virtual disk on a remote storage platform. Metadata is generated and stored in replicas on different metadata nodes of the storage platform. A modified log-structured merge tree is used to store and compact string-sorted tables of metadata. During file storage and compaction, a consistent file identification scheme is used across all metadata nodes. A fingerprint file is calculated for each SST (metadata) file on disk that includes hash values corresponding to regions of the SST file. To synchronize, the fingerprint files of two SST files are compared, and if any hash values are missing from a fingerprint file then the key-value-timestamp triplets corresponding to these missing hash values are sent to the SST file that is missing them. The SST file is compacted with the missing triplets to create a new version of the SST file. The synchronization is bi-directional.

Abstract: A client machine writes to and reads from a virtual disk on a remote storage platform. Metadata is generated and stored in replicas on different metadata nodes of the storage platform. A modified log-structured merge tree is used to store and compact string-sorted tables of metadata. During file storage and compaction, a consistent file identification scheme is used across all metadata nodes. A fingerprint file is calculated for each SST (metadata) file on disk that includes hash values corresponding to regions of the SST file. To synchronize, the fingerprint files of two SST files are compared, and if any hash values are missing from a fingerprint file then the key-value-timestamp triples corresponding to these missing hash values are sent to the SST file that is missing them. The SST file is compacted with the missing triples to create a new version of the SST file. The synchronization is bi-directional.

Abstract: Disclosed deduplication techniques at a distributed data storage system guarantee that space reclamation will not affect deduplicated data integrity even without perfect synchronization between components. By understanding certain “behavioral” characteristics and schedule cadences of backup operations that generate backup copies received at the distributed data storage system, data blocks that are not re-written by subsequent backup copies are pro-actively aged, while promoting continued retention of data blocks that are re-written. An expiry scheme operates with block-level granularity. Each unique deduplicated data block is given an expiry timeframe based on the block's arrival time at the distributed data storage system (i.e., when a backup copy supplies the block) and further based on backup frequencies of the various virtual disks referencing a unique system-wide identifier of the block, which is based on the block's hash value. Communications between components are kept to an as-needed basis.

Abstract: A client machine writes to and reads from a virtual disk on a remote storage platform. Metadata is generated and stored in replicas on different metadata nodes of the storage platform. A modified log-structured merge tree is used to store and compact string-sorted tables of metadata. During file storage and compaction, a consistent file identification scheme is used across all metadata nodes. A fingerprint file is calculated for each SST (metadata) file on disk that includes hash values corresponding to regions of the SST file. To synchronize, the fingerprint files of two SST files are compared, and if any hash values are missing from a fingerprint file, then the key-value-timestamp triples corresponding to these missing hash values are sent to the SST file that is missing them. The SST file is compacted with the missing triples to create a new version of the SST file. The synchronization is bi-directional.

Abstract: In order to avoid writing duplicates of blocks of data into a storage platform, any virtual disk within the storage platform may have a de-duplication feature enabled. Or, all virtual disks have this feature enabled. For virtual disks with de-duplication enabled, a unique message digest is calculated for every block of data written to that virtual disk. Upon a write, these message digests are consulted in order to determine if a particular block of data has already been written, if so, it is not written again, and if not, it is written. All de-duplication virtual disks are written to a single system virtual disk within the storage platform. De-duplication occurs over the entire storage platform and over all its virtual disks because all message digests are consulted before a write is performed for any virtual disk. A read for a de-duplication virtual desk reads from the system virtual disk.

Abstract: A distributed data storage system comprises features for integration with application orchestrators such as Kubernetes, and includes a proprietary Container Storage Interface (CSI) driver. Features include setting snapshot scheduling and retention policies, and a “container data mover” that replicates data from a source to a distinct destination distributed data storage system. The migration may be configured one-to-one, one-to-many, unidirectional, and/or bi-directional. Metadata-based snapshots and metadata-based changed block tracking identify payload data to move from source to destination within application orchestrator frameworks at both ends. Payload data is migrated from source to destination using different techniques than those used for migrating metadata, e.g., kernel-to-kernel for copying payload data versus ordinary metadata writes. An illustrative barrier logic ensures that the migration follows a controlled progression of operations.

Abstract: An application within a virtual machine is an iSCSI Initiator and is allowed to use as an iSCSI Target another virtual machine within the same hypervisor in order to make a persistent reservation for a virtual disk within a remotely-located storage platform. Any number of virtual machines within different hypervisors, and perhaps on different computers, use a local controller virtual machine to make a persistent reservation for the same virtual disk. The registration list and the current reservation holder data for an iSCSI persistent reservation for a particular virtual disk are held on a storage node of the storage platform rather than within a single virtual machine of a remote computer. A metadata module on the storage platform handles the incoming requests. A coordinator module within the storage platform uses a lock mechanism to guarantee that the reserve, release, preempt and clear commands are handled properly.

Abstract: A distributed data storage system comprises features for integration with application orchestrators such as Kubernetes, and includes a proprietary Container Storage Interface (CSI) driver. Features include setting snapshot scheduling and retention policies, and a “container data mover” that replicates data from a source to a distinct destination distributed data storage system. The migration may be configured one-to-one, one-to-many, unidirectional, and/or bi-directional. Metadata-based snapshots and metadata-based changed block tracking identify payload data to move from source to destination within application orchestrator frameworks at both ends. Payload data is migrated from source to destination using different techniques than those used for migrating metadata, e.g., kernel-to-kernel for copying payload data versus ordinary metadata writes. An illustrative barrier logic ensures that the migration follows a controlled progression of operations.

Abstract: In order to avoid writing duplicates of blocks of data into a storage platform, any virtual disk within the storage platform may have a de-duplication feature enabled. Or, all virtual disks have this feature enabled. For virtual disks with de-duplication enabled, a unique message digest is calculated for every block of data written to that virtual disk. Upon a write, these message digests are consulted in order to determine if a particular block of data has already been written, if so, it is not written again, and if not, it is written. All de-duplication virtual disks are written to a single system virtual disk within the storage platform. De-duplication occurs over the entire storage platform and over all its virtual disks because all message digests are consulted before a write is performed for any virtual disk. A read for a de-duplication virtual desk reads from the system virtual disk.

Abstract: In a running distributed data storage system that actively processes I/Os, metadata nodes are commissioned and decommissioned without taking down the storage system and without introducing interruptions to metadata or payload data I/O. The inflow of reads and writes continues without interruption even while new metadata nodes are in the process of being added and/or removed and the strong consistency of the system is guaranteed. Commissioning and decommissioning nodes within the running system enables streamlined replacement of permanently failed nodes and advantageously enables the system to adapt elastically to workload changes. An illustrative distributed barrier logic (the “view change barrier”) controls a multi-state process that controls a coordinated step-wise progression of the metadata nodes from an old view to a new normal. Rules for I/O handling govern each state until the state machine loop has been traversed and the system reaches its new normal.

Abstract: A strongly consistent distributed data storage system comprises an enhanced metadata service that is capable of fully recovering all metadata that goes missing when a metadata-carrying disk, disks, and/or partition fail. An illustrative recovery service runs automatically or on demand to bring the metadata node back into full service. Advantages of the recovery service include guaranteed full recovery of all missing metadata, including metadata still residing in commit logs, without impacting strong consistency guarantees of the metadata. The recovery service is network-traffic efficient. In preferred embodiments, the recovery service avoids metadata service downtime at the metadata node, thereby reducing the impact of metadata disk failure on the availability of the system. The disclosed metadata recovery techniques are said to be “self-healing” as they do not need manual intervention and instead automatically detect failures and automatically recover from the failures in a non-disruptive manner.

Abstract: In a running distributed data storage system that actively processes I/Os, metadata nodes are commissioned and decommissioned without taking down the storage system and without introducing interruptions to metadata or payload data I/O. The inflow of reads and writes continues without interruption even while new metadata nodes are in the process of being added and/or removed and the strong consistency of the system is guaranteed. Commissioning and decommissioning nodes within the running system enables streamlined replacement of permanently failed nodes and advantageously enables the system to adapt elastically to workload changes. An illustrative distributed barrier logic (the “view change barrier”) controls a multi-state process that controls a coordinated step-wise progression of the metadata nodes from an old view to a new normal. Rules for I/O handling govern each state until the state machine loop has been traversed and the system reaches its new normal.

Abstract: A computer-implemented data processing system and method writes a first plurality of copies of a data set at a first plurality of hosts and reads a second plurality of copies of the data set at a second plurality of hosts. The first and second pluralities of copies may be overlapping and the first and second pluralities of hosts may be overlapping. A hashing function may be used to select the first and second pluralities of hosts. Version histories for each of the first copies of the data set may also be written at the first plurality of hosts and read at the second plurality of hosts. The version histories for the second copies of the data set may be compared and causal between the second copies of the data set may be evaluated based on the version histories for the second copies of the data set.

Abstract: A distributed data storage system comprises features for integration with application orchestrators such as Kubernetes, and includes a proprietary Container Storage Interface (CSI) driver. Features include setting snapshot scheduling and retention policies, and a “container data mover” that replicates data from a source to a distinct destination distributed data storage system. The migration may be configured one-to-one, one-to-many, unidirectional, and/or bi-directional. Metadata-based snapshots and metadata-based changed block tracking identify payload data to move from source to destination within application orchestrator frameworks at both ends. Payload data is migrated from source to destination using different techniques than those used for migrating metadata, e.g., kernel-to-kernel for copying payload data versus ordinary metadata writes. An illustrative barrier logic ensures that the migration follows a controlled progression of operations.

Abstract: Disclosed deduplication techniques at a distributed data storage system guarantee that space reclamation will not affect deduplicated data integrity even without perfect synchronization between components. By understanding certain “behavioral” characteristics and schedule cadences of backup operations that generate backup copies received at the distributed data storage system, data blocks that are not re-written by subsequent backup copies are pro-actively aged, while promoting continued retention of data blocks that are re-written. An expiry scheme operates with block-level granularity. Each unique deduplicated data block is given an expiry timeframe based on the block's arrival time at the distributed data storage system (i.e., when a backup copy supplies the block) and further based on backup frequencies of the various virtual disks referencing a unique system-wide identifier of the block, which is based on the block's hash value. Communications between components are kept to an as-needed basis.

Abstract: Disclosed deduplication techniques at a distributed data storage system guarantee that space reclamation will not affect deduplicated data integrity even without perfect synchronization between components. By understanding certain “behavioral” characteristics and schedule cadences of backup operations that generate backup copies received at the distributed data storage system, data blocks that are not re-written by subsequent backup copies are pro-actively aged, while promoting continued retention of data blocks that are re-written. An expiry scheme operates with block-level granularity. Each unique deduplicated data block is given an expiry timeframe based on the block's arrival time at the distributed data storage system (i.e., when a backup copy supplies the block) and further based on backup frequencies of the various virtual disks referencing a unique system-wide identifier of the block, which is based on the block's hash value. Communications between components are kept to an as-needed basis.

Abstract: A distributed data storage system using erasure coding (EC) provides advantages of EC data storage while retaining high resiliency for EC data storage architectures having fewer data storage nodes than the number of EC data-plus-parity fragments. To ameliorate the effects of certain storage node outages or fatal disk failures, incoming data is temporarily replicated so that read and write operations can continue from/to the storage system. The system automatically heals failed EC write attempts in a manner transparent to users and/or applications: when all storage nodes are operational, the distributed data storage system automatically converts the temporarily replicated data to EC storage and reclaims storage space previously used by the temporarily replicated data. Individual hardware failures are healed through migration techniques that reconstruct and re-fragment data blocks according to the governing EC scheme. An illustrative embodiment is a three-node data storage system using EC 4+2.

Abstract: A distributed data storage system using erasure coding (EC) provides advantages of EC data storage while retaining high resiliency for EC data storage architectures having fewer data storage nodes than the number of EC data-plus-parity fragments. An illustrative embodiment is a three-node data storage system with EC 4+2. Incoming data is temporarily replicated to ameliorate the effects of certain storage node outages or fatal disk failures, so that read and write operations can continue from/to the storage system. The system is equipped to automatically heal failed EC write attempts in a manner transparent to users and/or applications: when all storage nodes are operational, the distributed data storage system automatically converts the temporarily replicated data to EC storage and reclaims storage space previously used by the temporarily replicated data. Individual hardware failures are healed through migration techniques that reconstruct and re-fragment data blocks according to the governing EC scheme.

Abstract: Encryption of data occurs before it is written to the storage platform; decryption occurs after it is read from the storage platform on a computer separate from the storage platform. By encrypting data before it travels over a wide-area network to a storage platform (and by only decrypting that data once it has arrived at an enterprise from the storage platform), we address data security over the network. Application data is encrypted at the virtual disk level before it leaves a controller virtual machine, and is only decrypted at that controller virtual machine after being received from the storage platform. Encryption and decryption of data is compatible with other services of the storage system such as de-duplication. Any number of key management services can be used in a transparent manner.