No assignee for patent application serial number 821971 has been made.
News editors obtained the following quote from the background information supplied by the inventors: "Traditional relational database systems (RDBMS) use a row-oriented storage system (row store). Examples include
"A huge drawback of row stores can be seen for read-only queries. Assume a read-only query reading two out of ten attributes of any given table. In that case the RDBMS would ideally only need to access two of the attributes on disk in order to compute the result to the query. However, as the storage system of the database is strictly page-oriented, the RDBMS has to read all pages, even though those pages contain data for all ten attributes, i.e. the data belonging to the other eight attributes that are not required to compute the result to the query are also fetched from disk. This means, a row store needs to read 100% of the data pages.
"As a consequence, over the past ten years several companies have proposed RDBMSs using a different type of storage system. It is based on a column store. Examples include Sybase IQ, SAP BI Accelerator, Vertica, and VectorWise. In a column store the values for all rows are stored contiguously on disk per column, i.e. these systems store each column in a separate physical unit. Consider again a read-only query requiring only two out of ten attributes. In order to compute the result to this query in a column store it suffices to read the two files containing the attribute values for these two columns only. This means that a column store does not need to read data belonging to any of the eight remaining columns. As a consequence, if we assume that each attribute has the same width, a column store needs to read only 20% of the data pages.
"Their high update costs are a huge drawback of column stores. Assume a new tuple is to be inserted into a table. Then all column files have to be changed individually, i.e. in our example ten random I/O-operations will have to be performed. This is extremely costly when compared to a row store. A second drawback of column stores consists in additional costs to merge (join) attributes from the different files to form a contiguous tuple. These costs may be lowered by applying late materialization (
Most Popular Stories
- 'Beyonce' Tops the U.S. Album Chart
- Singer Ian Watkins Sentenced to 35 Years Prison for Child Sex Abuse
- Universal CityWalk Rings in 2014 with Southern California's Biggest New Year's Eve Party, Featuring Dazzling Fireworks Displays, Three Live Concert Stages, DJ Performances and a Midnight Cascade of 1,000 Pounds of Confetti
- The Illuminati Don't Exist, Just Ask Them
- SAC Capital Employee Guilty of Insider Trading
- Brothel Write-offs Cost Taxpayers Millions, Coburn Says
- Joaquin Phoenix Reboots in 'Her'
- Fed Signals Strong Confidence in U.S. Economy
- GLAAD Denounces 'Duck Dynasty' Star for Remarks About Gays
- Senate Sends Bipartisan Budget Plan to President