Emc IP Holding Company LLC

Emc IP Holding Corporation LLC

Kedar Patwardhan

Rajesh K Nair

Anand Ghatnekar

Arash Sepasi Ahoei

Mathew Sprehn

Ernest Daza

Thomas Papadakis

Devon Kemp

Dale Terrien

Wallun Chan

Michael Smolenski

Peter Cheng Chen

Mangesh Sudhir Nijasure

Kiran Dudala

Maruthi Seshidhar Inukonda

Systems and methods for adding backups to de-duplicated storage and for removing backups from de-duplicated storage are disclosed. Backups can be added to the de-duplicated storage with minimal reference count updates. Backup data sets can be removed without garbage collection processes being performed on the de-duplicated storage.

Storing de-duplicated data with minimal reference counts

Backup operations on selected directories are performed in a computing system. File system operations that impact the ability to backup the selected directories are failed while other file system operations are allowed. In particular, file system operations that would require a trawl of the selected directory in order to perform the backup operation are failed.

Directory selective backups

Systems and methods for backing up data are disclosed. Changes made to a file system are tracked in a change log. When a backup operation is initiated, the changes identified in the change log are applied to the backup save sets. When a file associated with an event in the change log cannot be identified, the file is identified using the inode or other identifier. The event in the change log is then applied once the file is identified. The backup so achieved has eventual name space consistency and can be achieved without taking a snapshot of the file system.

Incremental backup with eventual name space consistency

In one example, a method managing data includes transmitting backup format information to a storage platform, breaking data targeted for backup into data pieces, computing a unique fingerprint for each of the data pieces, and transmitting metadata concerning the data pieces to the storage platform, where the metadata includes the fingerprints. The method further includes receiving from the storage platform a list of any new fingerprints that do not correspond to any of the data pieces, and grouping data pieces together to form a jumbo data piece, where the data pieces that form the jumbo data piece are data pieces to which the new fingerprints correspond.

System and method for efficient episodic replication of data to cloud storage

A first snapshot of a primary disk of a virtual machine (VM) is requested at a beginning of a current backup cycle. A first indication that the first snapshot has been created includes a first listing of changed blocks since a last snapshot was requested. The first snapshot is deleted. Blocks listed in the first listing are copied while the first snapshot is absent. After the copying, a second snapshot is requested. A second indication that the second snapshot has been created includes a second listing of changed blocks since the first snapshot. Blocks listed in both the first and second listings are copied while the second snapshot is present.

Systems and methods for snapshot-less backups

One example method includes reading a data chunk from a data stream, compressing the data chunk, and calculating a chunk delta. When the chunk delta is greater than zero, the compressed data chunk is appended to an incomplete data chunk. When the chunk delta is zero or less, the boundaries of a completed and compressed data chunk having a size at least as large as a minimum size are declared.

Method and system for compression aware chunking of backup streams

A first list is received including locations of blocks of a file system that changed between current and previous backups. A superblock of the file system is read to generate a second list including locations of metadata records representing file objects. The first and second lists are intersected to return a third list including locations listed in the first and second lists. The locations in the third list are changed metadata records. A metadata record as of the current backup and identified in the third list is read. A previous version of the metadata record as of the previous backup is read. The current and previous metadata record versions are compared to detect potential conditions occurring between the backups. Upon detection of a first type of condition, information about the condition is recorded. Upon detection of a second type of condition, information about the condition is not recorded.

Efficient content indexing of incremental block-based backups

In general, embodiments of the invention relate to a method for processing input/output (I/O requests). The method includes receiving an I/O request from a virtual machine executing on a host computing device, where the host computing device comprises a filter driver. The method further includes making a first determination that the filter driver is in a backup mode and based on the first determination, making a second determination that the I/O request intersects with a prior filter driver tracking list (FDTL). Further, in response to the second determination, the method includes pausing, by the filter driver, the processing of the I/O request, obtaining, from a primary disk, data associated with the I/O request, after obtaining the data, resuming the processing of the I/O request, where processing the I/O request results in the data on the primary disk being overwritten with new data.

Method and system for creating virtual snapshots using input/output (I/O) interception

In a cloud storage system, fast recovery from a crash of a cloud gateway or the cloud storage is afforded without the necessity of traditional garbage collection. Data objects are each broken up into a plurality of chunk objects that are prefixed with a unique non-repeating forever increasing generation number that is incremented for each different data object. The chunk objects of a particular data object are identified and listed in a manifest file that is also associated with that generation number. Upon a crash of either the cloud gateway or the cloud storage, a query identifies the cloud and deletes all orphaned chunk objects prefixed by and all manifest files associated with the prevailing generation number at the time of the crash. The deleted chunk objects and manifest files are regenerated and restored to cloud storage to return to consistency.

Fast recovery of backup cloud gateway following crash without garbage collection

A system identifies a first partition and another partition in a full backup created at a first time for a dataset. The system identifies a revision of the first partition in a first incremental backup created at a second time for the dataset, and links the first incremental backup to the full backup based on the first partition and the revision of the first partition. The system identifies a revision of the other partition in another incremental backup created at a third time for the dataset, and links the other incremental backup to the full backup based on the other partition and the revision of the other partition. The system restores the other partition in the dataset based on the link from the other incremental backup to the full backup, in response to receiving a request to restore the other partition in the dataset.

Linking backup files based on data partitions

Data management using an open standard file system interface to a storage gateway is described. A system uses an open standard file system interface to instruct a storage gateway to create a directory, create a file for the directory, write information to the file, and store, based on the information written to the file, data to storage associated with the storage gateway.

Data management using an open standard file system interface to a storage gateway

De-duplicating backup files based on data evolution is described. A system determines that a first backup file is created for a first data object, determines that the first data object is based on a second data object, and de-duplicates the first backup file based on a second backup file that was created for the second data object. The system determines that a third backup file is created for a third data object, determines that the third data object is based on the second data object, and de-duplicates the third backup file based on the second backup file that was created for the second data object.

Growing community of inventors

Irvine, CA, United States of America

Kedar Patwardhan