Relational Query Optimization Chapter 15 1 Query Evaluation

Relational Query Optimization Chapter 15 1

Query Evaluation SQL queries are translated into an extended form of relational algebra: § Query Plan Reasoning: • Tree of ops, • with choice of one among several algorithms for each operator § Query Plan Execution: • Each operator implemented using `pull’ interface: when operator is `pulled’ for next output tuples, it `pulls’ on its inputs and computes them. 2

Overview of Query Evaluation v Query Plan Optimization : § Ideally: Want to find best plan. Practically: Avoid worst plans! v Two main issues in query optimization: § For a given query, what plans are considered? • Algorithm to search plan space for cheapest (estimated) plan. § How is the cost of a plan estimated? • Cost Models based on I/O estimates 3

Physical Query Plan v. An extended relational algebra tree v. Annotations at each node indicating access methods to use for each table v. The implementation methods used for each relational operator. v Costs estimated per operator. sname bid=100 (On-the-fly) sname rating > 5 bid=100 rating > 5 (On-the-fly) (Simple Nested Loops) sid=sid Reserves Sailors 4

Schema for Examples Sailors (sid: integer, sname: string, rating: integer, age: real) Reserves (sid: integer, bid: integer, day: dates, rname: string) Similar to old schema; rname added for variations. v Reserves: v § v Each tuple is 40 bytes long, 100 tuples per page, 1000 pages. Sailors: § Each tuple is 50 bytes long, 80 tuples per page, 500 pages. 5

Query Blocks: Units of Optimization v An SQL query is parsed into a collection of query blocks. § An SQL query with no nesting and exactly one SELECT, FROM, WHERE, GROUP BY, HAVING clause. § WHERE is in conjunctive normal form. v v Nested blocks are usually treated as calls to a subroutine, made once per outer tuple. Optimizing one block at a time. SELECT S. sname FROM Sailors S WHERE S. age IN (SELECT MAX (S 2. age) FROM Sailors S 2 GROUP BY S 2. rating) Outer block SELECT S. sname FROM Sailors S WHERE S. age IN Reference to nested block SELECT MAX (S 2. age) FROM Sailors S 2 GROUP BY S 2. rating Nested block 6

Query Block v For each block, the plans considered are: – All available access methods, for each relation in FROM clause. – All left-deep join trees – i. e. , all ways to join the relations one-at-a-time, with the inner relation in the FROM clause, considering all relation permutations and join methods. MAX(S 2. age) (Group By S 2. rating( S 2. age(Sailors))) 7

Cost Estimation v For each plan considered, must estimate cost: § Must estimate cost of each operation in plan tree. • Depends on input cardinalities. • Depends on algorithm (sequential scan, index scan). § Must also estimate size of result for each operation. • Use information about the input relations. • For selections and joins, assume independence of predicates. • Must make assumptions about effect of predicates. v Cost of plan = sum of cost of each operator in tree. 8

Cost Estimation Result size = product of cardinalities of involved relations (FROM) * product of reduction factors (WHERE). 9

Size Estimation and Reduction Factors v Consider a query block: SELECT attribute list FROM relation list WHERE term 1 AND. . . AND termk Maximum # tuples in result is product of cardinalities of relations in FROM clause. v Reduction factor (RF) associated with each term reflects impact of term in reducing result size. v v Result cardinality = Max #tuples * product of all RF’s. § Implicit assumption that terms are independent! 10

Reduction Factors SELECT attribute list FROM relation list WHERE term 1 AND. . . AND v termk Reduction factor (RF) associated with each term reflects impact of term in reducing result size: § § § Term col=value has RF ? Term col 1=col 2 has RF? Term col>value has RF? 11

Cost of a Plan (cont') v Reduction Factor. § Column = value. - Given index I on column, assume uniform distribution. 1/Nkeys(I). - Otherwise, fixed reduction factor 1/10 § Column 1 = column 2 - Given indexes I 1 and I 2 on column 1 and column 2, assuming each key value in I 1 (smaller one) has a matching value in I 2. 1/MAX(Nkeys(I 1), Nkeys(I 2)). - One index I, 1/Nkeys(I) - otherwise, 1/10 § Column > values - Given an index I on column, arithmetic type. High(I) – value / High(I) – Low(I). - Not arithmetic type, or no index. A fraction Less than half is arbitrarily chosen. § Column IN (list of values) - reduction factor for (column = value) * number of items in the list. v Assumptions for reduction factors: § Uniform distribution of values § Independent distribution of values in different columns. 12

Cost of a Plan (cont) v Improved Statistics: Histograms § Uniform distribution is not accurate since real data is not uniformly distributed. § Histogram: data structure maintained by DBMS to approximate data distribution. • Equiwidth: divide range of column values into subranges (buckets). Assuming distribution within histogram bucket is uniform. • Equidepth: number of tuples within each bucket is equal (almost). • Compressed equidepth: maintain separate counts for small number of very frequent values, and maintain equidepth histogram to cover remaining values. Equiwidth 5. 0 Equidepth 9. 0 2 3 3 1 2 1 3 8 4 2 0 1 2 4 9 2. 67 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 2. 25 1. 33 1. 0 1. 75 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Bucket 1 Count 8 2 4 2. 5 3 15 4 3 5 15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Bucket 1 Count 9 2 10 3 10 4 7 5 9 13

Relational Algebra Equivalences Allow to choose different join orders and to `push’ selections and projections ahead of joins. v Selections: (Cascade) v Combine several selections into one selection. Evaluate the conjunctive condition to each of the tuple. Separate conjunctions into smaller selection operations, Able to combine as Join. (Commute) v Projections: (Cascade) Where ai ai+1 for i =1…n-1. 14

Relational Algebra Equivalences (S T) (R (R S) (S R) + Show that: R (S T) v Joins: R S) (Associative) T (Commute) (T R) S • When joining several relations, we are free to join the relations in any order we choose. 15

Selects, Projects and Joins v v v Case 1: a ( c(R)) c ( a(R)) A projection commutes with a selection that only uses attributes retained by the projection. XCase 2: R c S c (R S) Combine a selection with a cross-product to form a join. Case 3: i. e. , (R S) (R) S A selection on just attributes of R commutes with Join. 16

Selects, Projects and Joins v In general, part of the selection condition can be pushed ahead of the Join. c 1 c 2 c 3 (R S) c 1( c 2 (R) c 3 (S)) v Project a (R S) a 1 (R) a 2 (S) a (R c S) a 1 (R) c a 2 (S) (Where c appears in a) 18

Enumeration of Alternative Plans v There are two main cases: § § Single-relation plans Multiple-relation plans 19

Single Relation Plans v Queries over a single relation: combination of selects, projects, and aggregate operations: § Main decision: which access path to use in retrieving tuples. • § Most selective access path (file scan / index) if only single operator considered. Different operations are essentially carried out together. • e. g. , if an index is used for a selection, projection is done for each retrieved tuple, and the resulting tuples are pipelined into the aggregate computation. 20

Single Relation Plans without Index SELECT S. rating, COUNT(*) FROM Sailors S WHERE S. rating > 5 AND S. age = 20 GROUP BY S. rating HAVING COUNT DISTINCT (S. sname) > 2 s. rating, COUNT(*)( HAVING COUNT DISTINCT(S. sname) > 2( GROUP Bys. rating( s. rating, S. sname ( s. rating>5 S. age=20 ( Sailors))))) • A file scan to retrieve tuples and apply selections and projections. • Writing out tuples after the selections and projections. • Sorting these tuples to implement GROUP BY clause. * Note: GROUP BY and HAVING are done on-the-fly. e. g. , Cost = Cost 1(scan) + cost 2 (writing <S. rating, S. sname> pairs) + cost 3 (sorting as per the GROUP BY clause). Sailors = 500 pages cost 1 = 500, cost 3 = 3*Npages = 60 (assuming two passes) cost 2 = 500* ratio of tuple size * RFs = 20 ratio of tuple size = pair size / tuple size = 0. 8 RF(s. rating>5) = 0. 5 RF(S. age=20) = 0. 1 21

Single-Relation Plans with Index v Single-index access path • When several indexes match the selection conditions, choose the most selective access path. v Multiple-index access path • Several indexes using Alternatives (2) or (3) for data entries match the selection condition. • Retrieve rids using them individually. • Intersect result set. Sort by page id. v Sorted-index access path • Grouping attributes is a prefix of a tree index. • Using index to retrieve tuples in the order required by GROUP BY clause. v Index-only access path • All attributes mentioned in the query are included in the search key for some dense index on the relation in the FROM clause. • Index-only scan to compute the answer. 22

Single-Relation Plans with Index v Index I on primary key matches selection: § v Clustered index I matching one or more selects: § v (NPages(I)+NTuples(R)) * product of RF’s of matching selects. Sequential scan of file: § + (NPages(I)+NPages(R)) * product of RF’s of matching selects. Non-clustered index I matching one or more selects: § v Cost is Height(I)+1 for a B+ tree, about 1. 2 for hash index. NPages(R). Note: Typically, no duplicate elimination on projections! (Exception: Done on answers if user says DISTINCT. ) 23

Example v v Sailors: 500 pages, 80 tuples/page Case 2: index on rating. § v Clustered index: (1/NKeys(I)) * (NPages(I)+NPages(S)) = (1/10) * (50+500) pages Unclustered index: (1/NKeys(I)) * (NPages(I)+NTuples(S)) = (1/10) * (50+40000) pages Case 3: index on sid. § v SELECT S. sid FROM Sailors S WHERE S. rating=8 Would have to retrieve all tuples/pages. With a clustered index, the cost is 50+500, with unclustered index, 50+40000. Case 4: file scan: § We retrieve all file pages (500). 24

Queries Over Multiple Relations v Fundamental decision in System R: only left-deep join trees are considered. § § As the number of joins increases, the number of alternative plans grows rapidly; we need to restrict the search space. Left-deep trees allow us to generate all fully pipelined plans. • Intermediate results not written to temporary files. • Not all left-deep trees are fully pipelined (e. g. , SM join). D D C A B C D A B C A B 25

Enumeration of Left-Deep Plans v Left-deep plans differ only in the order of relations, the access method for each relation, and the join method for each join. v Enumerated using N passes (if N relations joined): § § § Pass 1: Find best 1 -relation plan for each relation. Pass 2: Find best way to join result of each 1 -relation plan (as outer) to another relation. (All 2 -relation plans. ) Pass N: Find best way to join result of a (N-1)-relation plan (as outer) to the N’th relation. (All N-relation plans. ) Pass 1 v For each subset of relations, retain only: § § Cheapest plan overall, plus Cheapest plan for each interesting order of tuples. A B C D Pass 2 Pass 3 26

Enumeration of Left-Deep Plans (contd. ) v Pass 1: 1 relation plan. • Identify selection terms in the WHERE clause that mention only attributes of A. (perform first access of A, before Join) • Identify attributes of A not mentioned in SELECT or WHERE clause (Project out when first access of A, before Join) • Cheapest plan keeping order. • E. g. , a file scan (as cheapest overall plan for fetching all tuples) and a B+ tree index (as the cheapest plan for fetching all tuples in the search key order). 27

Enumeration of Left-Deep Plans (contd. ) v Pass 2: All 2 -relational Plans. • Each of the single-relation plan from Pass 1 as the outer relation, and every other relation as the inner relation. • Examine WHERE clauses: • Selections involving only attributes of inner relation (apply before Join). • Selections defining the Join. • Selections involving attributes of other relations (apply after Join). • Actually, only selections that are really applied before the join are those that match the chosen access paths for A and B. • Depending on Join algorithm chosen, the cost may include the materializing the outer relation. 28

Enumeration of Plans (Contd. ) v ORDER BY, GROUP BY, aggregates etc. handled as a final step, using either an `interestingly ordered’ plan or an additional sorting operator. v An N-1 way plan is not combined with an additional relation unless there is a join condition between them, unless all predicates in WHERE have been used up. § v i. e. , avoid Cartesian products if possible. In spite of pruning plan space, this approach is still exponential in the # of tables. 29

Example v Pass 1: § Sailors: B+ tree on rating Hash on sid Reserves: B+ tree on bid sname sid=sid Choice 1: B+ tree matches rating>5, probably cheapest. bid=100 rating > 5 • if selection is expected to retrieve a lot of tuples, and index is unclustered, file scan may be cheaper. • Decision: B+ tree plan kept (because tuples are in rating Reserves Sailors order). • § v Reserves: B+ tree on bid matches bid=500; cheapest. Pass 2: – each plan retained from Pass 1 as the outer. – how to join it with the (only) other relation. e. g. , Reserves as outer : Hash index can be used to get Sailors tuples that satisfy sid = outer tuple’s sid value. 30

Nested Queries v v Nested block is optimized independently, with the outer tuple considered as providing a selection condition. Outer block is optimized with the cost of `calling’ nested block computation taken into account. Implicit ordering of these blocks means that some good strategies are not considered. The non-nested version of the query is typically optimized better. SELECT S. sname FROM Sailors S WHERE EXISTS (SELECT * FROM Reserves R WHERE R. bid=103 AND R. sid=S. sid) Nested block to optimize: SELECT * FROM Reserves R WHERE R. bid=103 AND S. sid= outer value Equivalent non-nested query: SELECT S. sname FROM Sailors S, Reserves WHERE S. sid=R. sid AND R. bid=103 R 31

Highlights of System R Optimizer v Impact: § v Cost estimation: § § v Most widely used currently; works well for < 10 joins. Statistics, maintained in system catalogs, used to estimate cost of operations and result sizes. Considers combination of CPU and I/O costs. Plan Space: § Only the space of left-deep plans is considered. • Left-deep plans allow output of each operator to be pipelined into the next operator without storing it in a temporary relation. § Focus on optimizing SQLs without nesting. No duplication in projections. (unless DISTINCT used) § Cartesian products avoided. 32

Other Approaches to Query Optimization v Exclusive search is not suitable for large number of Joins. v Rule-based optimizers: a set of rules to guide the generation of candidate plans. v Randomized plan generation: probabilistic algorithms to explore a large space of plans. 33

Summary Query optimization is an important task in a relational DBMS. v Must understand optimization in order to understand the performance impact of a given database design (relations, indexes) on a workload (set of queries). v Two parts to optimizing a query: v § Consider a set of alternative plans. • Must prune search space; typically, left-deep plans only. § Must estimate cost of each plan that is considered. • Must estimate size of result and cost for each plan node. • Key issues: Statistics, indexes, operator implementations. 34

Summary v Single-relation queries: § § v All access paths considered, cheapest is chosen. Issues: Selections that match index, whether index key has all needed fields and/or provides tuples in a desired order. Multiple-relation queries: § § All single-relation plans are first enumerated. • Selections/projections considered as early as possible. Next, for each 1 -relation plan, all ways of joining another relation (as inner) are considered. Next, for each 2 -relation plan that is `retained’, all ways of joining another relation (as inner) are considered, etc. At each level, for each subset of relations, only best plan for each interesting order of tuples is `retained’. 35