Decision SupportIntroductionThree Complementary TrendsData WarehousingWarehousing IssuesMultidimensional Data ModelMOLAP vs ROLAPDimension HierarchiesOLAP QueriesSlide 10Comparison with SQL QueriesThe CUBE OperatorDesign IssuesImplementation IssuesJoin IndexesBitmapped Join IndexViews and Decision SupportView Modification (Evaluate On Demand)View Materialization (Precomputation)Issues in View MaterializationInteractive Queries: Beyond MaterializationTop N QueriesSlide 23SummaryOLAP & Data Warehousing. R. Ramakrishnan and J. Gehrke 1Decision SupportChapter 23OLAP & Data Warehousing. R. Ramakrishnan and J. Gehrke 2IntroductionIncreasingly, organizations are analyzing current and historical data to identify useful patterns and support business strategies.Emphasis is on complex, interactive, exploratory analysis of very large datasets created by integrating data from across all parts of an enterprise; data is fairly static.Contrast such On-Line Analytic Processing (OLAP) with traditional On-line Transaction Processing (OLTP): mostly long queries, instead of short update Xacts.OLAP & Data Warehousing. R. Ramakrishnan and J. Gehrke 3Three Complementary TrendsData Warehousing: Consolidate data from many sources in one large repository.Loading, periodic synchronization of replicas.Semantic integration.OLAP: Complex SQL queries and views. Queries based on spreadsheet-style operations and “multidimensional” view of data.Interactive and “online” queries.Data Mining: Exploratory search for interesting trends and anomalies. (Another lecture!)OLAP & Data Warehousing. R. Ramakrishnan and J. Gehrke 4Data WarehousingIntegrated data spanning long time periods, often augmented with summary information. Several gigabytes to terabytes common.Interactive response times expected for complex queries; ad-hoc updates uncommon.EXTERNAL DATA SOURCES EXTRACTTRANSFORM LOAD REFRESH DATAWAREHOUSE MetadataRepositorySUPPORTSOLAPDATAMININGOLAP & Data Warehousing. R. Ramakrishnan and J. Gehrke 5Warehousing IssuesSemantic Integration: When getting data from multiple sources, must eliminate mismatches, e.g., different currencies, schemas.Heterogeneous Sources: Must access data from a variety of source formats and repositories.Replication capabilities can be exploited here.Load, Refresh, Purge: Must load data, periodically refresh it, and purge too-old data.Metadata Management: Must keep track of source, loading time, and other information for all data in the warehouse.OLAP & Data Warehousing. R. Ramakrishnan and J. Gehrke 6Multidimensional Data ModelCollection of numeric measures, which depend on a set of dimensions. E.g., measure Sales, dimensions Product (key: pid), Location (locid), and Time (timeid).8 10 1030 20 5025 8 15 1 2 3 timeid pid11 12 1311 1 1 2511 2 1 811 3 1 1512 1 1 3012 2 1 2012 3 1 5013 1 1 813 2 1 1013 3 1 1011 1 2 35pidtimeidlocidsaleslocidSlice locid=1is shown:OLAP & Data Warehousing. R. Ramakrishnan and J. Gehrke 7MOLAP vs ROLAPMultidimensional data can be stored physically in a (disk-resident, persistent) array; called MOLAP systems. Alternatively, can store as a relation; called ROLAP systems.The main relation, which relates dimensions to a measure, is called the fact table. Each dimension can have additional attributes and an associated dimension table.E.g., Products(pid, pname, category, price)Fact tables are much larger than dimensional tables.OLAP & Data Warehousing. R. Ramakrishnan and J. Gehrke 8Dimension HierarchiesFor each dimension, the set of values can be organized in a hierarchy:PRODUCT TIME LOCATIONcategory week month statepname date city yearquarter countryOLAP & Data Warehousing. R. Ramakrishnan and J. Gehrke 9OLAP QueriesInfluenced by SQL and by spreadsheets.A common operation is to aggregate a measure over one or more dimensions.Find total sales.Find total sales for each city, or for each state.Find top five products ranked by total sales.Roll-up: Aggregating at different levels of a dimension hierarchy. E.g., Given total sales by city, we can roll-up to get sales by state.OLAP & Data Warehousing. R. Ramakrishnan and J. Gehrke 10OLAP QueriesDrill-down: The inverse of roll-up. E.g., Given total sales by state, can drill-down to get total sales by city.E.g., Can also drill-down on different dimension to get total sales by product for each state.Pivoting: Aggregation on selected dimensions.E.g., Pivoting on Location and Time yields this cross-tabulation:63 81 14438 107 14575 35 110 WI CA Total199519961997176 223 339Total Slicing and Dicing: Equality and range selections on one or more dimensions.OLAP & Data Warehousing. R. Ramakrishnan and J. Gehrke 11Comparison with SQL QueriesThe cross-tabulation obtained by pivoting can also be computed using a collection of SQLqueries:SELECT SUM(S.sales)FROM Sales S, Times T, Locations LWHERE S.timeid=T.timeid AND S.timeid=L.timeidGROUP BY T.year, L.stateSELECT SUM(S.sales)FROM Sales S, Times TWHERE S.timeid=T.timeidGROUP BY T.yearSELECT SUM(S.sales)FROM Sales S, Location LWHERE S.timeid=L.timeidGROUP BY L.stateOLAP & Data Warehousing. R. Ramakrishnan and J. Gehrke 12The CUBE OperatorGeneralizing the previous example, if there are k dimensions, we have 2^k possible SQL GROUP BY queries that can be generated through pivoting on a subset of dimensions.CUBE pid, locid, timeid BY SUM SalesEquivalent to rolling up Sales on all eight subsets of the set {pid, locid, timeid}; each roll-up corresponds to an SQL query of the form:SELECT SUM(S.sales)FROM Sales SGROUP BY grouping-listLots of recent work onoptimizing the CUBE operator!OLAP & Data Warehousing. R. Ramakrishnan and J. Gehrke 13Design IssuesFact table in BCNF; dimension tables not normalized.Dimension tables are small; updates/inserts/deletes are rare. So, anomalies less important than good query performance.This kind of schema is very common in OLAP applications, and is called a star schema; computing the join of all these relations is called a star join. pricecategorypnamepid countrystatecitylocidsaleslocidtimeidpidholiday_flagweekdatetimeidmonthquarteryear(Fact table)SALESTIMESPRODUCTS LOCATIONSOLAP & Data