1. CPS216: Data-Intensive Computing Systems Query Processing (contd.)
Shivnath Babu

2. Physical plan generation
Overview of
Query
Processing
SQL query
parse
parse tree
Query rewriting
Query
Optimization
statistics
logical query plan
Physical plan generation
physical query plan
execute
Query
Execution
result

3. Physical plan generation
SQL query
Initial logical plan
parse
Rewrite rules
parse tree
Query rewriting
Logical plan
statistics
logical query plan
Physical plan generation
“Best” logical plan
physical query plan
execute
result

4. B,D B,D R.C = S.C R.A = “c” Λ R.C = S.C R.A = “c” Query RewritingX X R S R S
We will revisit it towards the end of this lecture

5. Physical plan generation
SQL query
parse
parse tree
Query rewriting
statistics
Best logical query plan
Physical plan generation
Best physical query plan
execute
result

6. Physical Plan Generation
B,D
Project
Natural join
Hash join
R.A = “c” S
Index scan
Table scan R R S
Best logical plan

7. Physical plan generation
SQL query
parse
parse tree
Enumerate possible
physical plans
Query rewriting
statistics
Best logical query plan
Find the cost of
each plan
Physical plan generation
Best physical query plan
Pick plan with
minimum cost
execute
result

8. Physical Plan Generation
Logical Query Plan
P1 P2 …. Pn
C1 C2 …. Cn
Pick minimum cost one
Physical
plans
Costs

9. Plans for Query Execution
Roadmap
Path of a SQL query
Operator trees
Physical Vs Logical plans
Plumbing: Materialization Vs pipelining

10. Logical Plans Vs. Physical Plans
B,D
Project
Natural join
Hash join
R.A = “c” S
Index scan
Table scan R R S
Best logical plan

11. Operator Plumbing B,D R.A = “c” S R
Materialization: output of one operator written to disk, next operator reads from the disk
Pipelining: output of one operator directly fed to next operator

12. Materialization
B,D
Materialized here
R.A = “c” S R

13. Iterators: Pipelining
B,D
 Each operator supports:
Open()
GetNext()
Close()
R.A = “c” S R

14. Iterator for Table Scan (R)
Open() {
/** initialize variables */
b = first block of R;
t = first tuple in block b; }
GetNext() {
IF (t is past last tuple in block b) {
set b to next block;
IF (there is no next block)
/** no more tuples */
RETURN EOT;
ELSE t = first tuple in b; }
/** return current tuple */
oldt = t;
set t to next tuple in block b;
RETURN oldt;
Close() {
/** nothing to be done */ }

15. Iterator for Select R.A = “c” GetNext() { LOOP: t = Child.GetNext();
IF (t == EOT) {
/** no more tuples */
RETURN EOT; }
ELSE IF (t.A == “c”)
RETURN t;
ENDLOOP:
Open() {
/** initialize child */
Child.Open(); }
Close() {
/** inform child */
Child.Close(); }

16. Iterator for Sort R.A GetNext() { IF (more tuples)
RETURN next tuple in order;
ELSE RETURN EOT; }
Open() {
/** Bulk of the work is here */
Child.Open();
Read all tuples from Child
and sort them }
Close() {
/** inform child */
Child.Close(); }

17. Iterator for Tuple Nested Loop Join
Lexp
Rexp
TNLJ (conceptually)
for each r  Lexp do
for each s  Rexp do
if Lexp.C = Rexp.C, output r,s

18. Example 1: Left-Deep Plan
TNLJ
TableScan
TNLJ
R3(C,D)
TableScan
TableScan
R1(A,B)
R2(B,C)
Question: What is the sequence of getNext() calls?

19. Example 2: Right-Deep Plan
TNLJ
R3(C,D)
TableScan
R1(A,B)
TableScan
R2(B,C)
TNLJ
Question: What is the sequence of getNext() calls?

20. Cost Measure for a Physical Plan
There are many cost measures
Time to completion
Number of I/Os (we will see a lot of this)
Number of getNext() calls
Tradeoff: Simplicity of estimation Vs. Accurate estimation of performance as seen by user

21. Why do we need Query Rewriting?
Pruning the HUGE space of physical plans
Eliminating redundant conditions/operators
Rules that will improve performance with very high probability
Preprocessing
Getting queries into a form that we know how to handle best
Reduces optimization time drastically without noticeably affecting quality

22. Some Query Rewrite Rules
Transform one logical plan into another
Do not use statistics
Equivalences in relational algebra
Push-down predicates
Do projects early
Avoid cross-products if possible

23. Equivalences in Relational Algebra
R S = S R Commutativity
(R S) T = R (S T) Associativity
Also holds for: Cross Products, Union, Intersection
R x S = S x R
(R x S) x T = R x (S x T)
R U S = S U R
R U (S U T) = (R U S) U T

24. B,D B,D R.C = S.C R.A = “c” Λ R.C = S.C R.A = “c”
Apply Rewrite Rule (1)
B,D
B,D
R.C = S.C
R.A = “c” Λ R.C = S.C
R.A = “c” X X R S R S
B,D [ sR.C=S.C [R.A=“c”(R X S)]]

25. pxy (R) = px [py (R)] Rules: Project Let: X = set of attributes
Y = set of attributes
XY = X U Y
pxy (R) =
px [py (R)]

26. sp (R S) = sq (R S) = Rules: s + combined
Let p = predicate with only R attribs
q = predicate with only S attribs
m = predicate with only R,S attribs
sp (R S) =
sq (R S) =
[sp (R)] S
R [sq (S)]

27. B,D B,D R.C = S.C R.C = S.C R.A = “c” R.A = “c”
Apply Rewrite Rule (2)
B,D
B,D
R.C = S.C
R.C = S.C
R.A = “c” X
R.A = “c” S X R S R
B,D [ sR.C=S.C [R.A=“c”(R)] X S]

28. B,D B,D R.C = S.C R.A = “c” R.A = “c” Apply Rewrite Rule (3) X S
Natural join
R.A = “c” X S
R.A = “c” S R R
B,D [[R.A=“c”(R)] S]

29. Rules: s + combined (continued)
spq (R S) = [sp (R)] [sq (S)]
spqm (R S) =
sm [(sp R) (sq S)]
spvq (R S) =
[(sp R) S] U [R (sq S)]

30. Which are “good” transformations?
sp1p2 (R)  sp1 [sp2 (R)]
sp (R S)  [sp (R)] S
R S  S R
px [sp (R)]  px {sp [pxz (R)]}

31. Conventional wisdom: do projects early
Example: R(A,B,C,D,E) P: (A=3)  (B=“cat”)
pE {sp (R)} vs. pE {sp{pABE(R)}}

32. But: What if we have A, B indexes?
B = “cat” A=3
Intersect pointers to get
pointers to matching tuples

33. Bottom line: No transformation is always good Some are usually good:
Push selections down
Avoid cross-products if possible
Subqueries  Joins

34. Avoid Cross Products (if possible)
Select B,D
From R,S,T,U
Where R.A = S.B  R.C=T.C  R.D = U.D
Which join trees avoid cross-products?
If you can't avoid cross products, perform them as late as possible

35. More Query Rewrite Rules
Transform one logical plan into another
Do not use statistics
Equivalences in relational algebra
Push-down predicates
Do projects early
Avoid cross-products if possible
Use left-deep trees
Subqueries  Joins
Use of constraints, e.g., uniqueness