Oracle International Corporation

Samarjeet Singh Tomar

Prasad V Bagal

Harish Nandyala

Saurabh Manchanda

Amit Roy

Sameer Arun Joshi

Ankur Kemkar

James Aubrey Williams

Ravi Ranjan

Somya Gupta

Mounica Prodduturu

Harendra Kumar Mishra

Shrihari Amarendra Bhat

'Semi-supervised' machine learning relies on less human input than a supervised algorithm to train a machine learning algorithm to perform entity recognition (NER). Starting with a known entity value or known pattern value for a specific entity type, phrases in a training data corpus are identified that include the known entity value. Context-value patterns are generated to match selected phrases that include the known entity value. One or more context-value patterns may be validated based on human input. The validated patterns identify additional entity values. A subset of the additional entity values may also be validated based on human input. Occurrences of validated entity values may be labeled in the training corpus. Sample phrases from the labeled training dataset may be extracted to form a reduced-size training set for a supervised machine learning model which may be further used in production to label data for any named entity recognition application.

System and method to generate a labeled dataset for training an entity detection system

Disclosed is an approach to implement an architecture and methodology that allows rapid access to provisioned software without requiring the entirety of the software distribution to be transferred to the target system in an upfront manner. Instead, a multi-tier architecture is used that allows the software to be provisioned and accessed with efficient access and reads of distribution materials from one or more remote storage locations.

Rapid software provisioning and patching

Techniques are provided for transitioning from a hierarchical file system to an object store. A request is received to change file metadata of a hierarchical file system. In response to the request: in-cache file metadata of the hierarchical file system is modified. The in-cache file metadata includes a directory structure of the hierarchical file system and file metadata of the hierarchical file system. Additionally, an in-store metadata journal entry is generated. The in-store metadata journal entry indicates an update to in-store file metadata. The in-store file metadata includes the directory structure of the hierarchical file system and file metadata of the hierarchical file system and is stored in an object store. The in-store metadata journal entry is stored in the object store. Furthermore, in-store file metadata and in-store metadata journal entry may be used to reconstruct the in-cache file metadata on another node.

System for transition from a hierarchical file system to an object store

A shared storage architecture persistently stores database files in non-volatile memories (NVMs) of a plurality of computing nodes of a multi-node DBMS. The computing nodes of the multi-node DBMS store data blocks in NVM and each computing node of the DBMS stores copies of each data block stored on the plurality of computing nodes. A computing node that disconnects and subsequently rejoins the DBMS employs an on-demand approach to resilvering stale data blocks that have been updated in other computing nodes in the DBMS while the computing node was offline. A data block may be resilvered on-demand based on an I/O request for a specific data block from a workload running on the reconnected computing node. Stale data blocks on the reconnected computing node are not resilvered unless they are accessed by the workload.

Stale block resynchronization in NVM based systems

Techniques are described herein for re-syncing a snapshot database with a base database. Data units of each database are associated with version numbers. When a data unit is modified, the data unit is associated with the current version number. When a snapshot database is re-synced with a base database, particular data units are copied from the base database to the snapshot database, based on the version number associated with the particular data units.

Efficient mirror data re-sync

Systems, methods, and other embodiments associated with separating executables and working data of an application between separate images are described. In one embodiment, a method includes mounting an application image together with a working image to create an application directory for an application. The application image is read-only and includes at least an executable of the application and the working image includes working data modified by the application. The application directory provides files and paths of the files for the application together within a single unified directory while separating the executable in the application image from the working data in the working image. The method includes in response to receiving a write request to the application directory, arbitrating access to the application directory by controlling the processor to write data for the write request to the working image to maintain the working data separate from the executable of the application.

Computerized system and method for patching an application by separating executables and working data using different images

Techniques are described for logically isolating data I/O requests from different operating systems (OSes) for a same multi-tenant storage system (MTSS). Techniques provide for OSes and the MTSS to obtain security tokens associated with the OSes. In an embodiment, an OS uses a security token to generate an authentication token based on the contents of a data input/output (I/O) request and sends the authentication token to the MTSS along with the data I/O request. When an MTSS receives such data I/O request, MTSS retrieves its own copy of the security token associated with the OS and generates its own authentication token based on the contents of the received data I/O request. If the authentication token generated by the MTSS matches the authentication token generated by the OS, then the data I/O request is successfully authenticated. Otherwise, if the authorization tokens fail to match, then the data I/O request has been compromised. For example, either the contents of data I/O request has been tampered with, or an entity other than the OS, has sent the data I/O request in the first place. Accordingly, the data I/O request may not be serviced by the MTSS.

Storage isolation using I/O authentication

A method, apparatus, and system for interposed file system driver is provided, which provides a logical file system on top of an existing base file system. One such interposed file system driver is a compression and deduplication layered driver ('COLD driver'). File system operations are intercepted from the operating system through the COLD driver, which is provided as an upper-level operating system driver that operates on top of an existing base file system. By processing file data through various modules, the existing base file system can be extended as a logical file system with compression, deduplication, indexing, and other functionality. The COLD driver can be implemented without requiring modifications to existing base file system structures or base file system drivers. Server deployments may thus leverage the additional file system functionality provided by the COLD driver without having to migrate to another file system.

Growing community of inventors

Bengaluru, India

Samarjeet Singh Tomar