Teradata Us, Inc.

Mcdata Corporation

Jeremy L Branscome

Joseph I Chamdani

Michael Paul Corwin

Liuxi Yang

Krishnan Meiyyappan

Ravi Krishnamurthy

Kapil Laxmikant Surlaker

James Shau

Hung Viet Tran

Alan Lee Beck

Rajasekhar Cherabuddi

Raj Cherabuddi

Ravindran Krishnamurthy

Hareesh Boinepelli

Robert Hou

Jim Crowley

Norm Wayne Smith

Liuxi Lang

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.

Transitioning between code-based and data-based execution forms in computing systems and environments

Embodiments of the present invention provide a hardware accelerator that assists a host database system in processing its queries. The hardware accelerator comprises special purpose processing elements that are capable of receiving database query/operation tasks in the form of machine code database instructions, execute them in hardware without software, and return the query/operation result back to the host system.

Methods and systems for hardware acceleration of database operations and queries for a versioned database based on multiple hardware accelerators

Methods and an apparatus for data sorting is provided. Keys are derived from a data set and a mapping function is obtained for sorting the data set in accordance with the mapping function. A wide key sort on the keys is performed over a plurality of distributed nodes using the mapping function, resulting in sorted lists of rows from the data set produced in parallel from the nodes with each row associated with a unique one of the keys pushed to a stack machine. The sort process is an ordered row traversal from the stack machine.

Method and a system for efficient data sorting

Storage devices and components, including memory components (e.g., non-volatile memory) can be trained by executable code that facilitates and/or performs reads and/or write requests to one or more storage sub-modules of a storage component (e.g., memory configured on a memory channel) made up of multiple storage components (e.g., DIMMs). The executable code can also train multiple storage components at the same time and/or in parallel.

Training of storage devices in computing systems and environments

A source data set is processed to produce a symbol table and a distribution without using a tree construct or any tree-related processing. The symbol table and the distribution outputted for encoding the data set and decoding encoded versions of the data set.

Efficient data encoding

Embodiments of the present invention provide hardware-friendly indexing of databases. In particular, forward and reverse indexing are utilized to allow for easy traversal of primary key to foreign key relationships. A novel structure known as a hit list also allows for easy scanning of various indexes in hardware. Group indexing is provided for flexible support of complex group key definition, such as for date range indexing and text indexing. A Replicated Reordered Column (RRC) may also be added to the group index to convert random I/O pattern into sequential I/O of only needed column elements.

Accessing data in a column store database based on hardware compatible indexing and replicated reordered columns

Embodiments of the present invention provide a run-time scheduler that schedules tasks for database queries on one or more execution resources in a dataflow fashion. In some embodiments, the run-time scheduler may comprise a task manager, a memory manager, and hardware resource manager. When a query is received by a host database management system, a query plan is created for that query. The query plan splits a query into various fragments. These fragments are further compiled into a directed acyclic graph of tasks. Unlike conventional scheduling, the dependency arc in the directed acyclic graph is based on page resources. Tasks may comprise machine code that may be executed by hardware to perform portions of the query. These tasks may also be performed in software or relate to I/O.

Methods and systems for run-time scheduling database operations that are executed in hardware

Embodiments of the present invention provide one or more hardware-friendly data structures that enable efficient hardware acceleration of database operations. In particular, the present invention employs a column-store format for the database. In the database, column-groups are stored with implicit row ids (RIDs) and a RID-to-primary key column having both column-store and row-store benefits via column hopping and a heap structure for adding new data. Fixed-width column compression allow for easy hardware database processing directly on the compressed data. A global database virtual address space is utilized that allows for arithmetic derivation of any physical address of the data regardless of its location. A word compression dictionary with token compare and sort index is also provided to allow for efficient hardware-based searching of text. A tuple reconstruction process is provided as well that allows hardware to reconstruct a row by stitching together data from multiple column groups.

Accessing data in column store database based on hardware compatible data structures

Divisions by numbers that are not divisible by two (2) can be performed in a computing system based on a summation that estimates and/or approximates the reciprocal of the dividing number or denominator value. By way of example, dividing by three (3) can be calculated based on a summation that approximates or estimates one third (&#x2153;) represented as the sum of a selected group of the inverses of the powers of two (2) in a pattern, namely the sum of: &#xbc;, 1/16, 1/64, 1/256, . . . ). Applications of the division techniques are virtually unlimited and include memory mapping of global memory addresses to memory channel addresses by dividing a global memory address into the number of memory channels, allowing memory mapping to be performed in an efficient manner even for large memory spaces using a number of memory channels that are not divisible by two, including prime numbers.

Division of numerical values based on summations and memory mapping in computing systems

A memory channel can be divided into two or more memory sub-channels, wherein each one of the memory sub-channels includes two or more memory components configured to store data made accessible on that memory sub-channel, and wherein the two or more memory components in each one of the memory sub-channels are respectively connected via at least one transmission line and can be individually accessed (addressed) on their associated sub-channel.

Designated memory sub-channels for computing systems and environments

Methods and systems for hardware acceleration of streamed database operations and queries based on multiple hardware accelerators

Methods and systems for hardware acceleration of database operations and queries based on multiple hardware accelerators

Embodiments of the present invention provide fine grain concurrency control for transactions in the presence of database updates. During operations, each transaction is assigned a snapshot version number or SVN. A SVN refers to a historical snapshot of the database that can be created periodically or on demand. Transactions are thus tied to a particular SVN, such as, when the transaction was created. Queries belonging to the transactions can access data that is consistent as of a point in time, for example, corresponding to the latest SVN when the transaction was created. At various times, data from the database stored in a memory can be updated using the snapshot data corresponding to a SVN. When a transaction is committed, a snapshot of the database with a new SVN is created based on the data modified by the transaction and the snapshot is synchronized to the memory. When a transaction query requires data from a version of the database corresponding to a SVN, the data in the memory may be synchronized with the snapshot data corresponding to that SVN.

Growing community of inventors

Santa Clara, CA, United States of America

Jeremy L Branscome