Microsoft Technology Licensing, LLC

Singlestore, Inc.

Adrian Birka

Cristian Diaconu

Xin Zhang

Rodrigo Toste Gomes

Georgiy Reynya

Ganapathy Subramaniam Krishnamoorthy

Vasileios Papadimos

Adam Prout

Joseph Victor

Weiyun Huang

Per-Åke Larson

Sangeetha Shekar

Tiago Vale

A database management system configured to perform a computer-implemented method is provided. The method involving configuring a first set of nodes to determine an instantaneous state of the database management system, generate system configuration data, and provide the system configuration data to a second set of nodes of the database management system, wherein the configuration data is replicated amongst the first set of nodes according to a consensus-based replication protocol so as to maintain the system configuration data. The method also involves configuring the second set of nodes to receive the system configuration data and implement a state machine that is arranged to modify a configuration of the second set of nodes according to the system configuration data, wherein the second set of nodes are arranged to replicate user data according to a non-consensus-based replication protocol. A non-transitory computer-readable storage medium comprising instructions for performing the computer-implemented method is provided.

High availability with consensus in database systems

A method for managing log files for recording operations on data stored in a database is provided, wherein a set of log files having an allocated first portion of storage is updated, the allocated first portion of storage is monitored, and a second portion of storage is allocated in dependence on a determination that an available portion of storage is below a predetermined size. A method for generating a snapshot is provided, wherein data entries are included in the snapshot in dependence on a determined relative order of log records in a set of log files. A method of replicating a binary large object is provided, wherein the binary large object is sent to a second database in response to identifying a log record comprising data indicating the binary large object. A method for deleting one or more binary large objects is provided, wherein the data indicating one or more binary large objects that have been logically deleted are stored in a snapshot and the one or more binary large objects are deleted using the data indicating the one or more binary large objects after a previous snapshot is deleted.

Database management system

The processing of transaction oriented data tends to be row-oriented, while the processing of analytical operations tends to be column-oriented. Various systems, sometimes referred to as operational data warehouses, may comprise mechanisms adapted for use in scenarios where both transactional data processing and analytical queries are to be performed efficiently. The operational data warehouse (ODW) may perform and update data efficiently by maintaining a table in structures comprising a column store, a delta store, a delete bitmap, and a delete buffer. In this environment, key values may be associated for each row such that the ODW may more efficiently seek rows. Further, rows may also be excluded from a column store based at least in part on a filter criterion. The filtering criterion may be used to filter out rows based on a created predicate set by a user or the system.

Efficient updates in non-clustered column stores

Embodiments are directed to efficiently determining which cache entries are to be evicted from memory and to incorporating a probability of reuse estimation in a cache entry eviction determination. A computer system with multiple different caches accesses a cache entry. The computer system determines an entry cost value for the accessed cache entry. The entry cost value indicates an amount of time the computer system is slowed down by to load the cache entry into cache memory. The computer system determines an opportunity cost value for the computing system caches. The opportunity cost value indicates an amount of time by which the computer system is slowed down while performing other operations that could have used the cache entry's cache memory space. Upon determining that the entry cost value is lower than the opportunity cost value, the computer system probabilistically evicts the cache entry from cache memory.

Cache eviction using memory entry value

Efficient server-side query statement parameter bucketization is provided for a query language, such as the structured query language (SQL). A process is applied for replacing the parameters of a parameterized query with new parameters, whose types are normalized to enable more query execution plan sharing. In one embodiment, in response to receiving a parameterized query expression of a query language, the bucketization process operates to bucketize parameter(s) of the parameterized expression based on the type(s) of the parameter(s) by replacing the type(s) with new broader, or more specific, type(s). An existing server-side parameterization framework that includes a mechanism for determining when parameterization is 'safe,' is extended, and is applied to bucketize parameters by using the mechanism to determine when replacing a parameter(s) of parameterized expressions with new parameter(s) of different type(s) is safe.

Server side bucketization of parameterized queries

Growing community of inventors

Seattle, WA, United States of America

Adrian Birka