Microsoft Technology Licensing, LLC

Hewlett-packard Development Company, L.p.

Sybase, Inc.

Coral 8, Inc.

Rajeev Motwani

Surajit J Chaudhuri

Vivek Ravindranath Narasayya

Mayur D Datar

Venkatesh Ganti

Glenn Robert Seidman

Mark Tsimelzon

Aleksey Sanin

Nina Mishra

Liadan O'Callaghan

Sudipto Guha

Gayatri Patel

Kris Kuppuswamy Ganjam

Sarajit Chaudhuri

A computer software language capable of expressing registered queries that operate on one more or more data streams continuously. The language of the present invention is based on a publish/subscribe model in that queries subscribe to data streams and publish to data streams. Also, the language of the present invention can express queries that operate directly on data streams. Since queries expressed in the language of the present invention may be executed continuously and directly on data streams, the language includes a clause for specifying time-based and/or row-based windows for the input data stream. Operations are then performed on the data within such windows. In one embodiment, the language is also SQL-like and includes a clause for defining named windows (which can be used in any number of queries); a clause for detecting a pattern, and correlated database subqueries for correlating data stream data with database tables.

Continuous processing language for real-time data streams

An outlier index for a database and a given workload is generated by identifying sub-relations of tuples in the database induced by selection and group by conditions in queries in the workload. A variance is then generated for values in each sub-relation. Sub-relations having higher variances are selected, and outliers from such sub-relations having higher variances are generated.

Sampling for queries

A database server supports weighted and unweighted sampling of records or tuples in accordance with desired sampling semantics such as with replacement (WR), without replacement (WoR), or independent coin flips (CF) semantics, for example. The database server may perform such sampling sequentially not only to sample non-materialized records, such as those produced as a stream by a pipeline in a query tree for example, but also to sample records, whether materialized or not, in a single pass. The database server also supports sampling over a join of two relations of records or tuples without requiring the computation of the full join and without requiring the materialization of both relations and/or indexes on the join attribute values of both relations.

Sampling for database systems

At least one implementation, described herein, detects fuzzy duplicates and eliminates such duplicates. Fuzzy duplicates are multiple, seemingly distinct tuples (i.e., records) in a database that represent the same real-world entity or phenomenon.

Robust detector of fuzzy duplicates

A method of estimating results of a database query, the results are estimated by performing a sampling of weighted tuples in a database based on a probability of usage of tuples required in executing a workload. A probability is associated with each tuple sampled. An aggregate is computed over values in each sampled tuple while multiplying by the inverses of the probabilities associated with each tuple sampled.

A Continuous Query Processor processes queries on continuously updating data sources or data streams and includes a Publication Manager for accepting published structured elements of data from data stream Publishers, a Subscription Manager for giving structured elements of data to one or more data stream Subscribers, a Query Module Manager for processing queries represented by Query Modules, a Query Module Store for maintaining deployed Query Modules, a Query Primitive Manager performing processing for various primitives that comprise a Query Module, and a Schedule Manager for coordinating when a primitive within a Query Module gets processed in order to maintain that each continuous query is continuously updated immediately upon the arrival of structured element data affecting any part of a continuous query.

Publish and subscribe capable continuous query processor for real-time data streams

Aggregation queries are performed by first identifying outlier values, aggregating the outlier values, and sampling the remaining data after pruning the outlier values. The sampled data is extrapolated and added to the aggregated outlier values to provide an estimate for each aggregation query. Outlier values are identified by selecting values outside of a selected sliding window of data having the lowest variance. An index is created for the outlier values. The outlier data is removed from the window of data, and separately aggregated. The remaining data without the outliers is then sampled to provide a statistically relevant sample that is then aggregated and extrapolated to provide an estimate for the remaining data. This sampled estimate is combined with the outlier aggregate to form an estimate for the entire set of data.

Database aggregation query result estimator

To help ensure high data quality, data warehouses validate and clean, if needed incoming data tuples from external sources. In many situations, input tuples or portions of input tuples must match acceptable tuples in a reference table. For example, product name and description fields in a sales record from a distributor must match the pre-recorded name and description fields in a product reference relation. A disclosed system implements an efficient and accurate approximate or fuzzy match operation that can effectively clean an incoming tuple if it fails to match exactly with any of the multiple tuples in the reference relation. A disclosed similarity function that utilizes token substrings referred to as q-grams overcomes limitations of prior art similarity functions while efficiently performing a fuzzy match process.

Efficient fuzzy match for evaluating data records

This disclosure describes leveraging workload information associated with executed database queries for estimating the result of a current database query. The workload information is analyzed to determine the usage of tuples in a database during query execution, such as how often a tuple is accessed and the number of different queries that accessed the tuple. A tuple is assigned a weight value that is based on the analyzed workload information. The particular tuples sampled for estimating a result for the current query is based on each tuple's weight value. The workload information may also be leveraged to generate an outlier index that identifies outlier tuples associated with the executed queries or that identifies outlier tuples associated with particular queries that are executed more frequently than other queries. The result for the current query can also be estimated using the sampled values along with the outlier tuples from the outlier index.

Aggregation queries are performed by first identifying outlier values, aggregating the outlier values, and sampling the remaining data after pruning the outlier values. The sampled data is extrapolated and added to the aggregated outlier values to provide an estimate for each aggregation query. Outlier values are identified by selecting values outside of a selected sliding window of data having the lowest variance. An index is created for the outlier values. The outlier data is removed from the window of data, and separately aggregated. The remaining data without the outliers is then sampled in one of many known ways to provide a statistically relevant sample that is then aggregated and extrapolated to provide an estimate for the remaining data. This sampled estimate is combined with the outlier aggregate to form an estimate for the entire set of data. Further methods involve the use of weighted sampling and weighted selection of outlier values for low selectivity queries, or queries having group by.

A technique that uses a weighted divide and conquer approach for clustering a set S of n data points to find k final centers. The technique comprises 1) partitioning the set S into P disjoint pieces S, . . . , S; 2) for each piece S, determining a set Dof k intermediate centers; 3) assigning each data point in each piece Sto the nearest one of the k intermediate centers; 4) weighting each of the k intermediate centers in each set Dby the number of points in the corresponding piece Sassigned to that center; and 5) clustering the weighted intermediate centers together to find said k final centers, the clustering performed using a specific error metric and a clustering method A.

Computer implemented scalable, incremental and parallel clustering based on weighted divide and conquer

Sampling for aggregation queries

A technique that uses a weighted divide and conquer approach for clustering a set S of n data points to find k final centers. The technique comprises 1) partitioning the set S into P disjoint pieces S , . . . , S ; 2) for each piece S , determining a set D of k intermediate centers; 3) assigning each data point in each piece S to the nearest one of the k intermediate centers; 4) weighting each of the k intermediate centers in each set D by the number of points in the corresponding piece S assigned to that center; and 5) clustering the weighted intermediate centers together to find said k final centers, the clustering performed using a specific error metric and a clustering method A.

A database server supports weighted and unweighted sampling of records or tuples in accordance with desired sampling semantics such as with replacement (WR), without replacement (WoR), or independent coin flips (CF) semantics, for example. The database server may perform such sampling sequentially not only to sample non-materialized records such as those produced as a stream by a pipeline in a query tree for example, but also to sample records, whether materialized or not, in a single pass. The database server also supports sampling over a join of two relations of records or tuples without requiring the computation of the full join and without requiring the materialization of both relations and/or indexes on the join attribute values of both relations.

Growing community of inventors

Palo Alto, CA, United States of America

Rajeev Motwani