Patent application serial number 846476 has not been assigned to a company or institution.
The following quote was obtained by the news editors from the background information supplied by the inventors: "In order to handle an ever growing amount of information, database management systems have recently been implemented on what are known as distributed or parallel architectures. One form of distributed architecture is the shared-nothing architecture. This architecture is called shared-nothing because the architecture does not include a single point of failure, such as a shared disk or shared memory.
"A typical shared nothing system includes a plurality of database partitions coupled together by a network. The database partitions generally include a set of storage devices, such as disk drives, and one or more processors. The network may comprise various types of networks, such as a local area network, and physically diverse paths. Data in a shared nothing system is partitioned using a partitioning scheme such as a hash partitioning scheme and stored across the various database partitions that are connected together through the network. Thus, each database partition keeps a part of the overall database and various operations on parts of the database in different database partitions may be performed in parallel.
"Data in the database is kept as tuples in relations that comprises of logically grouped attributes. Tuples and attributes in relations are also referred to as rows and columns in tables. An important feature of any database system, including shared-nothing systems, is the ability to process operations between rows in tables, such as relating data in one table to data in another table. These types of operations are particularly useful when querying related data from multiple tables such that the data needs to be retrieved in a meaningful way. For example, structured query language ('SQL') supports what are known as join operations for querying data from multiple tables. Indeed, many consider the join operation as the cornerstone of database management systems.
"A join operation may be expensive to perform in that it may require a large amount of processing and communication resources. One form of join operation in a distributed database system is the semi-join operation. A semi-join operation involves two tables R and S that are stored on different sites within a distributed system. A semi-join from R to S is performed by first projecting R on the join attributes at the site of R. The join attributes typically include one or more keys for selecting various tuples from tables R and S. Site R sends the resulting projection to the site of S, where a join operation is performed. The projection of R essentially acts as a filter of S, and thus, the result of the semi-join is usually smaller than S. Site S then sends to site R the semi-join result to complete the join operation. Hence, semi-joins can require creating and sending projections of one of the tables involved in the join across a network when the tables involved in the join are stored at different sites. This could be expensive and hence techniques to reduce semi-join processing are needed.
Most Popular Stories
- American Airlines, US Airways Complete Merger
- ACA Delay Stresses Small Businesses
- Questions Remain in Jenni Rivera's Death
- Harley Issues Motorcycle Recall
- Unemployed Wait as Lawmakers Debate
- General Dynamics Plans 200 New Jobs in N.M.
- Auto Dealer Builds Big Solar Project
- Entrepreneurs' Next Creation May Be New Laws
- Saab Gets Back into the Game; U.S. Auto Sales Soar
- Dell Offers Undisclosed Number of Employee Buyouts