Microsoft Technology Licensing, LLC

Edgars Sedols

Maninderjit Singh Parmar

Babatunde Micheal Okutubo

Parag Nandan Paul

Sudin Bhat

Gjorgji Gjeorgjievski

Methods and systems are provided for improved access to rows of data in a distributed data system. Each data row is associated with a partition. Data rows are distributed in one or more files and an impure file includes data rows associated multiple partitions. A clustering set is generated from a plurality of impure files by selecting a candidate impure file based on file access activity metrics and one or more neighbor impure files. Data rows of the impure files included in the clustering set are sorted according to their respective associated partitions. A set of disjoint partition range files are generated based on the sorted data rows of the impure files included in the clustering set. Each file of the set of disjoint partition range files is transferred to a respective target partition.

Incrementally improving clustering of cross partition data in a distributed data system

Bias correcting system for small number estimators. A computer system includes a distinct value estimator configured to estimate a number of distinct values in a data set. The computer system includes a bias table for the estimator. The bias table includes entries with values corresponding to biases caused by the distinct value estimator correlated to values corresponding to numbers estimated. The entries in the table are optimized by having a set of entries with an optimized number of biases in the entries. The biases in the entries are associated with predetermined confidence intervals. The system includes a bias corrector configured to correct the number of distinct values in the multiset data estimated by the distinct value estimator set using values from the bias table to produce a corrected value. The system includes a user interface coupled to the bias corrector configured to output the corrected value to a user.

Low cardinality bias correction system

A method for optimizing memory usage for a hyperloglog (HLL) estimator is disclosed. When a multiset is applied to an HLL estimator, a sparse representation of a sketch may be created. The sparse representation may be sorted whenever a total number of buckets in the sparse representation is evenly divisible by a sorting threshold and is also less than a transition threshold. The sparse representation may be converted to a dense representation when the number of buckets in the sparse representation reaches the transition threshold.

Growing community of inventors

Redmond, WA, United States of America

Edgars Sedols