Oracle International Corporation

Devrev, Inc.

Nimar Singh Arora

Lik Wong

Thuvan Hoang

Byron Wang

Lei Gao

James William Stamos

Edwina Lu

Anand Lakshminath

Edwina Ming-Yue Lu

Jun Yuan

Jingwei Wu

Haobo Xu

Ravi Murthy

Zhen Hua Liu

Muralidhar Krishnaprasad

Dieter Gawlick

Dheeraj Pandey

James W Warner

Alan Robert Downing

Benny Souder

Geeta Arora

Sundeep Abraham

Hung Viet Tran

Ramkumar Venkatesan

Sean Lehouillier

Vasanth Rajamani

Patricia H McElroy

Cristina Schmidt

Amit Ganesh

Sivasankaran Chandrasekar

Lakshminarayanan Luxi Chidambaran

Kapil Laxmikant Surlaker

Alan J Demers

Jan Olderdissen

Anindya Misra

Mahesh Subramaniam

Brent Chun

Ken Chen

Jim Stamos

Lakshminaray Chidambaran

Lewis S Kaplan

Brian Byrne

Anthony Giardullo

Neeraj Shadhan

Tianshu Li

Pratyush Goel

Meher Ritesh Kumar Goru

Steven Poitras

Bhavana Thudi

Manoj Agarwal

Thuyan Hoang

Techniques for implementing a vector database in a multi-tenant environment are described. A system creates an index of a tenant that scales efficiently in a multi-tenant environment. The index is created by clustering the plurality of vectors into a set of clusters. The created index forms a hierarchical index including plurality of layers and is stored in a primary data storage unit. The system includes an intermediate data storage unit to store new vectors and to avoid re-indexing every time a new vector with an associated operation such as insert, update, and delete, is added. Further, the system provides reliable nearest neighbor vectors from the created index of the tenant. Read operation is performed over the quick-retrieval data, primary data, and intermediate to determine the nearest neighbor vectors.

System and method to implement a scalable vector database

Disclosed is an improved approach to process data and events from disparate ecosystems pertaining to products. An approach is provided to automatically cluster events from various ecosystems into noteworthy incidents and to correlate them with entities extracted from each system. Incidents are correlated between ecosystems to classify the type of incidents and to give a coherent converged picture of the event streams coming from the various ecosystems. Noteworthy incidents are automatically converted into tickets and their severity is ascertained from the associated incidents. Tickets that reference underlying defects with the product or service are converted into issues.

Method and system for integrated analysis of data from multiple source environments

Techniques are provided for eager replication of uncommitted transactions. In embodiments, a replication client receives, in a data stream, change records corresponding to database changes applied to a source database in a transaction. The change records does not include a commit record that indicates that the transaction is committed on the source database. Before receiving the commit record, the replication client computes transaction dependency data based on the change records and detects, based on the transaction dependency data, that the transaction can be at least partially applied to a target database. Also before receiving the commit record, the replication client applies, to a target database and based on the detecting, at least some of the change records. Upon receiving the commit record of the transaction, the replication client completes applying the change records and commits the transaction on the target database.

Eager replication of uncommitted transactions

Techniques are provided for dependency-aware transaction batching for data replication. A plurality of change records corresponding to a plurality of transactions is read. Inter-transaction dependency data is generated, the inter-transaction dependency data including at least one inter-transaction dependency relationship between a plurality of pending transactions. Each inter-transaction dependency relationship indicates that a first transaction is dependent on a second transaction. A batch transaction is generated based on the inter-transaction dependency data. The batch transaction includes at least one pending transaction of the plurality of pending transactions. The batch transaction is assigned to an apply process of a plurality of apply processes configured to apply batch transactions in parallel.

Dependency-aware transaction batching for data replication

Parallel logical replication involves multiple apply threads running on a destination database server applying, in parallel, changes made by source transactions, where the changes of a single source transaction may be applied in parallel by multiple apply threads. An apply transaction for a source transaction may be committed by an apply thread independently of the commitment of any other apply transaction of the source transaction, that is, without coordinating the committing of another apply transaction executed by another apply thread for the source transaction. A configuration language is used to configure parallel logical replication. The language facilitates the configuration of various aspects of parallel logical replication, including the number of apply threads, partitioning schemes for the apply threads for partitioning change records between the apply threads, and various other aspects of parallel logical replication.

Scalable coordination aware static partitioning for database replication

Techniques are provided for eager replication of uncommitted transactions. A first plurality of change records is received, corresponding to database changes applied to a source database in a first transaction. First transaction dependency data is computed based on the first transaction. At least a portion of the first plurality of change records is applied to the target database before processing a commit record indicating that has been committed on the source database. Target dependency data is updated after processing the first commit record to reflect completion of the first transaction, the target dependency data including dependency data for a plurality of transactions applied or scheduled to be applied on a target database.

Techniques are provided for client and server integration for scalable replication. A replication client transmits change records to a database server over a stream. The database server determines at least one batch comprising change records for at least one transaction. The database server generates dependency data for at least one change record in a batch based on at least one constraint identifier for at least one column. The database server determines an ordered grouping of the change records based on an operation type of each change record and the dependency data of each change record, wherein change records sharing operation types are grouped together unless a division based on the dependency data is determined. The database server generates a reordered transaction comprising a plurality of reordered operations based on the ordered grouping of the change records of the particular batch.

Client and server integration for replicating data

Techniques for replicating data in database systems are described. In an example embodiment, a set of changes is received at a destination database, where the set of changes has been applied at a source database and is being replicated from the source database to the destination database. The set of changes is analyzed and it is determined that the set of changes includes two or more of: a subset of row-level changes, a subset of statement-level changes, and a subset of procedure-level changes. A set of dependencies is determined at least between the changes that are included in the subsets of changes. The changes, in the subsets of changes, are assigned to two or more processing elements. The set of changes is applied to the destination database by executing the two or more processing elements in parallel to each other and based on the set of dependencies.

Statement-level and procedural-level replication

A method for sharing information between a publisher and multiple subscribers is provided. The publisher uses a latch-free, single publisher, multiple subscriber shared queue to share information. Logical change records representing changes made to a database are enqueued in the shared queue as messages in a stream of messages, and subscribers read the logical change records. Subscribers may filter logical change records before sending to apply processes for processing. An identifying property of the source instance of a change encapsulated in a logical change record may be included with each message enqueued.

High performant information sharing and replication for single-publisher and multiple-subscriber configuration

Techniques are provided for automatic parallelism tuning. At least one batch of change records is assigned to one or more apply processes in a set of active apply processes. A first throughput value is periodically determined based on a number of processed change records in a first time interval. An increment adjustment is periodically performed, including adding an additional apply process, determining a second throughput value, and removing the additional apply process from the set of active apply processes if the second throughput value is not greater than a previous first throughput value by at least an increment threshold. A decrement adjustment is periodically performed, including removing an apply process, determining a third throughput value, and replacing the removed apply process in the set of active apply processes if the third throughput value is not greater than the previous first throughput value by at least a decrement threshold.

Automatic parallelism tuning for apply processes

Techniques for replicating data between database systems without taking checkpoints are provided. In an embodiment, a capture process restarts. Upon restarting, the capture process reestablishes an association with an apply process. A particular logical time maintained by the apply process is then communicated to the capture process. Upon receiving the particular logical time, the capture process restarts mining from this particular logical time.

Checkpoint-free in log mining for distributed information sharing

In accordance with an embodiment of the present invention, an information sharing mechanism comprising a memory structure may be provided in a database system. In an embodiment, mined information transferred by a capture process to an associated apply process can be written into the memory structure without taking any latch. Similarly, the mined information can be read by the apply process from the memory structure without taking any latch. The capture and apply processes may work cooperatively to establish a safe point in log mining under various circumstances such as in an initial startup state, in a steady state, in a process restart scenario in the middle of information sharing. In some embodiments, the information sharing mechanism supports both checkpoint-free and checkpoint modes of log mining by the capture process. In addition, both the capture process and the apply process may employ an eager apply approach to increase processing capacity.

Combining capture and apply in a distributed information sharing system

Techniques for combining statement level, procedural, and row level replication

A computer is programmed to parse a network that interconnects databases in a distributed database system, into a number of paths, wherein each path starts in a source database that is a source of changes and ends in a destination database that uses the changes to replicate the source database. The computer identifies databases that occur within each path in the network, and queries the identified databases to obtain statistics for each path. Statistics along each path are analyzed automatically, to identify a component in each path as a candidate for further investigation. The identified component may be marked as a bottleneck e.g. if there are no other paths connected to it. Alternatively the identified component may be a source of changes for another path that has a bottleneck or that has another candidate for further investigation. The identified bottleneck is displayed to the user with recommendations for improvement.

Automated topology-based statistics monitoring and performance analysis

Techniques are disclosed for sharing information in a wide variety of contexts. An information sharing system is described that allows both an explicit capture process and an implicit capture process to add information items to a staging area. Further, the information sharing system supports both implicit and explicit consumption of information items that are stored in said staging area. A rules engine is provided to allow users to create and register rules that customize the behavior of the capture processes, the consuming processes, and propagation processes that propagate information from the staging areas to designated destinations. Techniques are also described for achieving exactly-once handling of sequence of items, where the items are maintained in volatile memory. Techniques are also provided for recording DDL operations, and for asynchronously performing operations based on the previously-performed DDL operations.

In memory streaming with disk backup and recovery of messages captured from a database redo stream

Growing community of inventors

Union City, CA, United States of America

Nimar Singh Arora