260 likes | 785 Views
Column-Oriented Database. Yiqun Xie (Ian) & Yingbin Liang (Ben) Group 9. Row-based RDBMS vs. Column-oriented DBMS. Row-based, the best solution?. Big Data OLAP-Online Analytical Processing … …. Traditional row-based RDBMS.
E N D
Column-Oriented Database YiqunXie (Ian) & Yingbin Liang (Ben) Group 9
Row-based, the best solution? • Big Data • OLAP-Online Analytical Processing • … …
Traditional row-based RDBMS • Optimal write time and abundant reading overhead for retrieval of subset queries varchar Traversing all the headers! Going through all attributes.
From row to column… … SELECT c1 FROM bigtable; … ideal
Strengths Column-oriented DBMS • Optimal read time for subset retrieval queries SELECT c1 FROM bigtable; 100GB RDBMS col4 20GB col3 19GB col2 10GB col1 1GB col6 30GB col5 20GB
SQL Example: Only read columns needed ! (7 columns) • SELECT month, storeFROM sales, productsWHERE productType = ‘technology’AND products.id = sales.productIDGROUP BY month, store • Easier to implement data compression algorithms (e.g. Run-length encoding) Fruit (3 Apples, 2 Pears, 1 Banana, 2 Oranges, 1 Apple)
Weakness Column-oriented DBMS • Bad write performance • Advanced column-oriented storage system: HBase, MongoDB, MapReduce INSERT INTO bigtable VALUES (Rn1, Rn2, Rn3, Rn4)
References [1] http://www.cs.yale.edu/homes/dna/talks/Column_Store_Tutorial _VLDB09.pdf [2] http://www.spatial.cs.umn.edu/Courses/Fall11/5707/class_no tes/G5.ppt [3] http://www.slideshare.net/arangodb/introduction-to-column- oriented-databases [4] Column-oriented database speeds predictive criminal monitoring program