Query Processing and Optimization

More documents

Recommendations

Info

Selection Strategies for Conjunctive Queries • Use any available indices for attributes involved in simple conditions. • If several are available, use the most selective index. Then check each record with respect to the remaining conditions. • Attempt to use composite indices. • This can be very efficient. • Do intersection of record pointers. • If several indices with record pointers are applicable to the selection predicate, retrieve and intersect the pointers. Then retrieve (and check) the qualifying records. • Disjunctive queries provide little opportunity for smart processing. 8
Joins • Join Strategies • Nested loop join • Index-based join • Sort-merge join • Hash join • Strategies work on a per block (not per record) basis. • Need to estimate #I/Os (block retrievals) • Relation sizes and join selectivities impact join cost. • Query selectivity = #tuples in result / #candidates ‘More selective’ means smaller ‘selectivity value’ • For join, #candidates is the size of Cartesian product 9
Page 1 and 2: Query Processing + Optimization: Ou
Page 3 and 4: Architectural Context DDL Statement
Page 5 and 6: Overview of Query Processing 1. Par
Page 7: • Linear search Selection Strateg
Page 11 and 12: Nested Loop • Basically, for each
Page 13 and 14: Index-based Join • Requires (at l
Page 15 and 16: Sort-Merge Join • Sort each relat
Page 17 and 18: External or Disk-based Sorting, Con
Page 19 and 20: Example of Sort-Merge Join • Cost
Page 21 and 22: Partitioning Phase • Partitioning
Page 23 and 24: Hash Join Cost • Cost Total = Cos
Page 25 and 26: Cost and Applicability of Join Stra
Page 27 and 28: Query Optimization • Aim: Transfo
Page 29 and 30: Apply Selections Early π Name σ F
Page 31 and 32: Form Joins π Name n CU.CustomerID
Page 33 and 34: Some Transformation Rules • Casca
Page 35 and 36: Heuristic Query Optimization Summar
Page 37 and 38: • Per relation • Tuple size Rel
Page 39 and 40: • Per Index • Base relation Rel
Page 41 and 42: • Statistics Operation 1: σ foll
Page 43 and 44: • Statistics Operation 3: σ foll
Page 45 and 46: Comparison • Heuristic query opti
Page 47 and 48: Query Tuning • Query optimization
Page 49 and 50: Join on Clustering Index, and Integ
Page 51 and 52: Unnesting Nested Queries • Uncorr
Page 53: Summary • Query processing & opti

Query Processing and Optimization

Create successful ePaper yourself

Delete template?

Save as template?