Abstract: A data warehouse executes as at least one of a plurality of processes on a primary cluster. The data warehouse issues executable steps to processes executing on a compute cluster. The data warehouse receives a performance monitoring (PM) query from a client requesting data warehouse performance monitoring information. A single operational view performance monitoring application program interface (SOVPMAPI) executing on the primary cluster converts the performance monitoring query into a call to a federator application program interface (FAPI) to gather data warehouse performance monitoring information requested in the performance monitoring query.

Abstract: A system includes a storage device configured to store at least one database. The system includes a plurality of processing nodes in communication with the storage device. The plurality of processing nodes collects a subset of the data. The plurality of processing nodes generates a plurality of vectors using a vectorizing large language model (“LLM”) based on the subset of related data. The plurality of processing nodes stores the plurality of vectors. The plurality of processing nodes receives a natural language query from a LLM agent. The plurality of processing nodes vectorizes the natural language query. The plurality of processing nodes compares the vectorized natural language query to the plurality of vectors. The plurality of processing nodes generates a database-language query based on the comparison. The plurality of processing nodes executes the database language query. A method and computer-readable medium are also disclosed.

Abstract: A system and methods for performing approximate search to find approximate matches or similarities between a short character string or token included in a database query and strings/tokens contained within a target database table. A novel Locality Sensitivity Hashing function utilizing a table of common bigrams and accompanying values associated with frequency of occurrence of the bigrams in the English language, compares bigrams contained in search and data strings/tokens to the bigrams contained in the table of common bigrams, generates hash values for search and data string/tokens from the values associated with matching bigrams, and creates a hash table mapping similar strings or tokens together through hash values.

Abstract: A system includes a storage device and a plurality of processing nodes in communication with the storage device. At least one processing node of the plurality of processing nodes receives query history. The query history contains information on a plurality of executed queries. The at least one processing node generates, using a large language model (“LLM”), a document for each query describing the information associated with a respective query. The at least one processing node divides generated documents into a plurality of groups. The at least one processing node compresses the documents to reduce a number of documents in each group until a predetermined threshold is reached. The at least one processing node stores a final compressed set of workload documents. The at least one processing node performs workload analysis using workloads described in the final compressed set of workload documents. A method and computer-readable medium are also disclosed.

Abstract: A system may include a storage device. The system further includes a plurality of processing nodes in communication with the storage device. Each processing node executes at least one unit of parallelism. At least one processing node generates at least one large object (“LOB”). The at least one processing node identifies a multi-thread storage system. The at least one processing node stores the LOB in the multi-thread storage system. The at least one processing node generates a unique identifier that indicates the location of the LOB in the multi-thread storage system. The at least one processing node stores unique identifier. The unique identifier is accessible by all units of parallelism. A method and computer-readable medium are also disclosed.

Abstract: A system may include a storage device. The system may further include a plurality of processing nodes. At least one processing node of the plurality of processing nodes may receive a natural language query. The at least one processing node may convert the natural language query to a database language syntax representation using a large language model. The at least one processing node may identify query history based on content of the database language syntax representation using the large language model. The at least one processing node may generate database syntax to execute the natural language query based on the identified query history using the large language model. The at least one processing node may execute the generated database syntax. A method and computer-readable medium are also disclosed.

Abstract: A system includes a storage device and a plurality of processing nodes. At least one of the plurality of processing nodes receives a natural language query. The at least one processing node converts the natural language query to a database language syntax representation using a large language model (“LLM”) and converts the natural language query to a vectorized natural language representation using a vectorizing LLM. The at least one processing node converts query history to a vectorized query history representation using the vectorizing LLM. The at least one processing node compares the vectorized natural language representation to the vectorized query history representation and generates database syntax in response to the natural language query based on the comparison of the vectorized natural language representation to the vectorized query history representation. The at least one processing node executes the generated database syntax. A method and computer-readable medium are also disclosed.

Abstract: A system may include a storage device and at least one processor in communication with the storage device. The at least one processor may receive a query that comprises a reference to at least one data object. The at least one processor may identify a locally-stored version of the at least one data object. The at least one processor may retrieve a current version of the at least one data object. The at least one processor may compare the locally-stored version and the current version of the at least one data object. The at least one processor may, in response to the locally-stored version being different than the current version, update the locally-stored version to the current version. The at least one processor may execute the query based on the current version of the at least one data object. A method and computer-readable medium are also disclosed.

Abstract: In some examples, a database system includes processing modules with access to a remote object store and a local database storage associated with the database system. The processing modules perform a database operation that involves use of a plurality of instances of spool data. The processing modules store a first instance of spool data in the remote object store based on a first characteristic of the first instance of spool data, and the processing modules store a second instance of spool data in the local database storage based on a second characteristic of the second instance of spool data.

Abstract: A system and methods for performing read and write operations in a database system including a managed object store with an open table format. These systems and methods provide throughput/concurrency improvements for write operations employing position deletes by avoiding unnecessary transaction aborts due to possible write conflicts. A probabilistic structure, such as a bloom or cuckoo filter, is utilized to log row updates which are matched with delete file entries to identify conflicts at the row level in write operations. Similarly, improvements in the efficiency of read operations when employing an equity delete process are achieved by using bloom filters for logging object file data, which is matched with delete file entries to eliminate rows from consideration in read operations.

Abstract: A system includes a storage device and at least one processor in communication with the storage device. The at least one processor receives a query associated with a plurality of data tables stored in the storage device. The at least one processor processes the query using a large language model (“LLM”) trained on semantic mapping information that describes relationships between data elements stored within the plurality of tables. The at least one processor generates, with the LLM, a natural language response to the query based on semantic mapping data generated from the data elements stored withing the tables. A method and computer-readable medium are also disclosed.

Abstract: Hyperparameter tuning for a machine learning model is performed in a massively parallel database system. A computer system comprised of a plurality of compute units executes a relational database management system (RDBMS), wherein the RDBMS manages a relational database comprised of one or more tables storing data. One or more of the compute units perform the hyperparameter tuning for the machine learning model, wherein the hyperparameters are control parameters used in construction of the model, and the tuning of the hyperparameters is implemented as an operation in the RDBMS that accepts training and scoring data for the model, constructs the model using the hyperparameters and the training data, and generates goodness metrics for the model using the scoring data.

Abstract: A method and apparatus for optimizing a query in a relational database management system (RDBMS) when a predicate on a data column in the query has a correlation to a partitioning attribute of a partitioning column in data retrieved from a cloud-based store, wherein the optimizing uses the correlation between the data column in the query to the partitioning column in the data retrieved from the cloud-based store for data elimination when processing the query. The correlation is defined in a formula or lookup data structure that maps or range-maps from the data column to the partitioning column.

Abstract: A memory management computer system arranges a set of consumers of a computer memory resource in an inverted tree hierarchy. The memory management computer system allocates the computer memory resource among the scheduling entities. Each of the scheduling entities allocates the computer memory resource among the consumers of the computer memory resource to which it is directly coupled in the hierarchy. One of the scheduling entities in the hierarchy, a modifying scheduling entity, determines that a change is required to the allocation of the computer memory resource to one of the consumers of the computer memory resource for which the modifying scheduling entity is responsible for allocating the computer memory resource. Modifying the allocation of the computer memory resource to the one of the consumers of the computer memory resource without changing the hierarchy or creating a new hierarchy.

Abstract: A system and method for balancing database workload in a cloud database system employing multiple database instances, wherein intermediate results are shared between database instances via object store or cloud storage system accessible by the multiple database instances. During sharing of intermediate results between database instances having different node sizes or processing unit sizes, load balancing is achieved by evaluating the target and source database instance sizes to select an appropriate load balancing strategy. Data and metadata contained within the intermediate results shared between source and target database instances are organized to provide optimal load balancing on the target database instance.

Abstract: A system and method for balancing database workload in a cloud database system employing multiple database instances, wherein intermediate results are shared between database instances via object store or cloud storage system accessible by the multiple database instances. During sharing of intermediate results between database instances having different node sizes or processing unit sizes, load balancing is achieved by evaluating the target and source database instance sizes to select an appropriate load balancing strategy. Data and metadata contained within the intermediate results shared between source and target database instances are organized to provide optimal load balancing on the target database instance.

Abstract: An apparatus, method and computer program product for neural network training over very large distributed datasets, wherein a relational database management system (RDBMS) is executed in a computer system comprised of a plurality of compute units, and the RDBMS manages a relational database comprised of one or more tables storing data. One or more local neural network models are trained in the compute units using the data stored locally on the compute units. At least one global neural network model is generated in the compute units by aggregating the local neural network models after the local neural network models are trained.

Abstract: Techniques for transitioning between code-based and data-based execution forms (or models) are disclosed. The techniques can be used to improve the performance of computing systems by allowing the execution to transition from one of the execution models to another one of the execution models that may be more suitable for carrying out the execution or effective processing of information in a computing system or environment. The techniques also allow switching back to the previous execution model when that previous model is more suitable than the execution model currently being used. In other words, the techniques allow transitioning (or switching) back and forth between a data-based and code-based execution (or information processing) models.

Abstract: A database system analyzes a query to determine the resources required to execute the query by identifying the department that issued the query (the “source department”) and identifying query-recited tables and determining the department by which each of the query-recited tables is controlled (the “accessed departments”). The database system determines that the resources required to execute the query are controlled by more than one department: the source department and one or more accessed departments. For each accessed department, a cost of a first strategy involving a SQL rewrite and a cost of a second strategy not involving a SQL rewrite are determined by preparing a query execution plan, determining the cost of the resources, and considering whether data controlled by the accessed department is not accessible except to the accessed department. The first strategy is chosen for at least one of the accessed departments because the first-strategy cost is lower.

Abstract: Improved techniques for management of access in computing environments and systems are disclosed. An object-level data access mechanism can be provided. to effectively provide an object-level locking mechanism for locking data objects of database tables, individually, as individual data objects. Furthermore, the object-level data access mechanism can be provided as a safe and efficient filtering mechanism (e.g., cuckoo filter) that effectively provide an object-level locking mechanisms for locking data objects of a database table, individually (i.e., as individual locks placed on individual data objects). For example, a set of filters (e.g., write cuckoo and read cuckoo) can be provided for a database table to facilitate concurrent database operations in a safe but efficient manner.