This presentation focuses on two of the new features in MySQL 8.0.18: hash joins and EXPLAIN ANALYZE. It covers how these features work, both on the surface and on the inside, and how you can use them to improve your queries and make them go faster. Both features are the result of major refactoring of how the MySQL executor works. In addition to explaining and demonstrating the features themselves, the presentation looks at how the investment in a new iterator based executor prepares MySQL for a future with faster queries, greater plan flexibility and even more SQL features.

1. Copyright © 2019 Oracle and/or its affiliates.1
Hash join and
EXPLAIN ANALYZE
Norvald H. Ryeng
Software Development Senior Manager
MySQL Optimizer Team
October 1, 2019
MySQL 8.0.18 latest updates

2. Copyright © 2019 Oracle and/or its affiliates.2
Safe harbor statement
The following is intended to outline our general product direction. It is intended for information
purposes only, and may not be incorporated into any contract. It is not a commitment to deliver
any material, code, or functionality, and should not be relied upon in making purchasing
decisions.
The development, release, timing, and pricing of any features or functionality described for
Oracle’s products may change and remains at the sole discretion of Oracle Corporation.

3. Copyright © 2019 Oracle and/or its affiliates.3
Program agenda
1 Quick demo
2 How did we get here?
Phase separation
Volcano iterator model
3 Hash join
4 EXPLAIN ANALYZE

4. Copyright © 2019 Oracle and/or its affiliates.4
Quick demo
MySQL 8.0.18

5. Copyright © 2019 Oracle and/or its affiliates.5

6. Copyright © 2019 Oracle and/or its affiliates.6
How did we get here?
A long term investment that finally pays off

7. Copyright © 2019 Oracle and/or its affiliates.7
https://twitter.com/vlad_mihalcea/status/1173842312398614528

8. Copyright © 2019 Oracle and/or its affiliates.8
MySQL refactoring: Separating phases
 Started ~10 years ago
Considered finished now
 A clear separation between query processing phases
 Fixed a large number of bugs
Improved stability
 Faster feature development
Fewer surprises and complications during development
Parse
Prepare
Optimize
Execute
SQL
Resolve
Transform
Abstract syntax tree
Logical plan
Physical plan

9. Copyright © 2019 Oracle and/or its affiliates.9
MySQL refactoring: Parsing and preparing
 Still ongoing
Implemented piece by piece
 Separating parsing and resolving phases
Eliminate semantic actions that do too much
Get a true bottom-up parser
 Makes it easier to extend with new SQL syntax
Parsing doesn't have unintended side effects
 Consistent name and type resolving
Names resolved top-down
Types resolved bottom-up
 Transformations done in the prepare phase
Bottom-up
Parse
Prepare
Optimize
Execute
SQL
Resolve
Transform
Abstract syntax tree
Logical plan
Physical plan

10. Copyright © 2019 Oracle and/or its affiliates.10
MySQL features made possible because we invested in refactoring
CTEs
 Recursive and non-recursive CTEs
 Traverse hierarchies
 Write more readable SQL
LATERAL
 "For-each loops"
… and many, many more!
Window functions
 Aggregation, ranking, analytics
 Sliding windows
JSON
 JSON_TABLE
 JSON window functions

11. Copyright © 2019 Oracle and/or its affiliates.11
MySQL refactoring: Iterator executor
 Volcano iterator model
 Possible because phases were separated
 Ongoing for ~1,5 year
 Much more modular exeuctor
Common iterator interface for all operations
Each operation is contained within an iterator
 Able to put together plans in new ways
Immediate benefit: Removes temporary tables in some cases
 Join is just an iterator
Nested loop join is just an iterator
Hash join is just an iterator
Your favorite join method is just an iterator
Parse
Prepare
Optimize
Execute
SQL
Resolve
Transform
Abstract syntax tree
Logical plan
Physical plan

12. Copyright © 2019 Oracle and/or its affiliates.12
Old MySQL executor vs. iterator executor
Old executor
 Nested loop focused
 Hard to extend
 Code for one operation spread out
 Different interfaces for each operation
 Combination of operations hard coded
Iterator executor
 Modular
 Easy to extend
 Each iterator encapsulates one operation
 Same interface for all iterators
 All operations can be connected

13. Copyright © 2019 Oracle and/or its affiliates.13
MySQL 8.0 features based on the iterator executor
 EXPLAIN FORMAT=TREE
Print the iterator tree
 EXPLAIN ANALYZE
1. Insert intstrumentation nodes in the tree
2. Execute the query
3. Print the iterator tree
 Hash join
Just another iterator type
Parse
Prepare
Optimize
Execute
SQL
Resolve
Transform
Abstract syntax tree
Logical plan
Physical plan

14. Copyright © 2019 Oracle and/or its affiliates.14
Hash join
MySQL internals
New in
8.0.18

15. Copyright © 2019 Oracle and/or its affiliates.15
MySQL hash join
 Hybrid hash join
 Three execution modes
Everything fits in memory
Spill to disk (GRACE hash join)
Looping
 Equi-join
 xxHash64
Fast
Good distribution
HashJoinIterator
TableScanIterator TableScanIterator
SELECT * FROM t1 JOIN t2 ON t1.a = t2.a;

16. Copyright © 2019 Oracle and/or its affiliates.16
MySQL hash join — everything fits in memory
1) Hash one table into memory
Smallest table
The whole table fits in memory (if not, use next method)
2) Scan the other table and match with rows in memory
2
1
A
=
B

17. Copyright © 2019 Oracle and/or its affiliates.17
MySQL hash join — spill to disk
1
2
3
4
1) Hash the smallest table into memory until the buffer is full
When the buffer is full, dump the rest as chunks on disk
Produce up to 128 hash buckets (chunks) on disk
2) Scan the other table and match with rows in memory
Write to chunks on disk at the same time
Produce up to 128 hash buckets (chunks) on disk
3) Hash one chunk of the smallest table into memory
Hash using a different seed than used to create disk buckets
(If the bucket doesn't fit, see next method)
4) Scan the corresponding chunk and match with rows in memory
5) Repeat 3-4 until all chunks have been processed
=
=
A
B

18. Copyright © 2019 Oracle and/or its affiliates.18
MySQL hash join — looping
1) Hash a batch of the chunk into memory
When the buffer is full, pause reading this operand
2) Scan the corresponding chunk and match with rows in memory
3) Hash the next batch into memory
Resume from where it was paused
4) Scan the corresponding chunk and match with rows in memory
5) Repeat 3-4 until the whole chunk has been processed
1 & 3
2 & 4
=

19. Copyright © 2019 Oracle and/or its affiliates.19
MySQL hash join optimization
 Replaces BNL
 Currently optimized as BNL
Replace with hash join after optimization
A conservative (safe) choice
 Optimizer switch to turn on/off
SET optimizer_switch='hash_join=on';
 Hints
/*+ HASH_JOIN(tables or query blocks) */
/*+ NO_HASH_JOIN(tables or query blocks) */
 Buffer size
SET join_buffer_size=number;
Parse
Prepare
Optimize
Execute
SQL
Resolve
Transform
Abstract syntax tree
Logical plan
Physical plan

20. Copyright © 2019 Oracle and/or its affiliates.20
467x
1520x>1400x
1332x
968x
MySQL hash join performance
 BNL compared to hash join
 Force BNL/hash join in DBT-3/TPC-H
DBT-3/TPC-H without indexes
Optimizer selects BNL
Automatic conversion to hash join
 Hash join is much faster than BNL
 Can't expect same improvement when
indexes are available
Lowernumberisbetter

21. Copyright © 2019 Oracle and/or its affiliates.21
New in
8.0.18
EXPLAIN ANALYZE
MySQL internals

22. Copyright © 2019 Oracle and/or its affiliates.22
TimingIteratorTimingIterator
TimingIterator
MySQL EXPLAIN ANALYZE
 Wrap iterators in instrumentation nodes
 Measurements
Time (in ms) to first row
Time (in ms) to last row
Number of rows
Number of loops
 Execute the query and dump the stats
HashJoinIterator
TableScanIterator TableScanIterator

23. Copyright © 2019 Oracle and/or its affiliates.23
467x
1520x>1400x
1332x
968x
What's wrong with Q2?
SELECT s_acctbal, s_name, n_name, p_partkey, p_mfgr, s_address,
s_phone, s_comment
FROM part, supplier, partsupp, nation, region
WHERE
p_partkey = ps_partkey AND s_suppkey = ps_suppkey AND p_size = 4
AND p_type LIKE '%TIN' AND s_nationkey = n_nationkey
AND n_regionkey = r_regionkey AND r_name = 'AMERICA'
AND ps_supplycost = (
SELECT min(ps_supplycost)
FROM partsupp, supplier, nation, region
WHERE
p_partkey = ps_partkey AND s_suppkey = ps_suppkey
AND s_nationkey = n_nationkey AND n_regionkey = r_regionkey
AND r_name = 'AMERICA'
)
ORDER BY s_acctbal DESC, n_name, s_name, p_partkey
LIMIT 100;
MySQL EXPLAIN ANALYZE to the rescue!

24. -> Limit: 100 row(s) (actual time=591739.000..591739.018 rows=100 loops=1)
-> Sort: <temporary>.s_acctbal DESC, <temporary>.n_name, <temporary>.s_name, <temporary>.p_partkey, limit input to 100 row(s) per chunk (actual time
-> Stream results (actual time=591738.599..591738.772 rows=462 loops=1)
-> Inner hash join (nation.n_regionkey = region.r_regionkey), (nation.n_nationkey = supplier.s_nationkey) (cost=2074295.37 rows=98) (actual
-> Table scan on nation (cost=0.00 rows=25) (actual time=0.024..0.026 rows=25 loops=1)
-> Hash
-> Inner hash join (supplier.s_suppkey = partsupp.ps_suppkey) (cost=2074041.10 rows=98) (actual time=591735.554..591738.311 rows=46
-> Table scan on supplier (cost=0.06 rows=9760) (actual time=0.068..2.024 rows=10000 loops=1)
-> Hash
-> Filter: (partsupp.ps_supplycost = (select #2)) (cost=1977898.52 rows=98) (actual time=1282.855..591733.987 rows=462 loop
-> Inner hash join (partsupp.ps_partkey = part.p_partkey) (cost=1977898.52 rows=98) (actual time=84.827..271.307 rows=3
-> Table scan on partsupp (cost=3.54 rows=796168) (actual time=0.034..108.684 rows=800000 loops=1)
-> Hash
-> Inner hash join (cost=20353.91 rows=24) (actual time=0.274..83.964 rows=780 loops=1)
-> Filter: ((part.p_size = 4) and (part.p_type like '%TIN')) (cost=20353.16 rows=2204) (actual time=0.212..
-> Table scan on part (cost=20353.16 rows=198401) (actual time=0.160..63.957 rows=200000 loops=1)
-> Hash
-> Filter: (region.r_name = 'AMERICA') (cost=0.75 rows=1) (actual time=0.029..0.036 rows=1 loops=1)
-> Table scan on region (cost=0.75 rows=5) (actual time=0.021..0.030 rows=5 loops=1)
-> Select #2 (subquery in condition; dependent)
-> Aggregate: min(partsupp.ps_supplycost) (actual time=189.566..189.566 rows=1 loops=3120)
-> Inner hash join (nation.n_regionkey = region.r_regionkey), (nation.n_nationkey = supplier.s_nationkey) (cost
-> Table scan on nation (cost=0.00 rows=25) (actual time=0.005..0.007 rows=25 loops=3120)
-> Hash
-> Inner hash join (supplier.s_suppkey = partsupp.ps_suppkey) (cost=77789257.28 rows=79617) (actual tim
-> Table scan on supplier (cost=0.02 rows=9760) (actual time=0.014..1.237 rows=10000 loops=3120)
-> Hash
-> Filter: (part.p_partkey = partsupp.ps_partkey) (cost=81757.41 rows=79617) (actual time=92.08
-> Inner hash join (cost=81757.41 rows=79617) (actual time=0.018..155.118 rows=800000 loops
-> Table scan on partsupp (cost=10101.54 rows=796168) (actual time=0.011..97.353 rows=8
-> Hash
-> Filter: (region.r_name = 'AMERICA') (cost=0.75 rows=1) (actual time=0.004..0.005
-> Table scan on region (cost=0.75 rows=5) (actual time=0.003..0.003 rows=5 loo
Time is in milliseconds

25. Copyright © 2019 Oracle and/or its affiliates.25
Make your queries run faster with MySQL 8.0
Automatic
 Temporary table elimination
Recursive CTEs
Some derived tables
UNION into derived tables
Input to sorting
 Faster duplicate removal
 Hash join instead of BNL
 Analyze queries to find out where time is spent
EXPLAIN FORMAT=TREE
EXPLAIN ANALYZE
Optimizer trace

26. Copyright © 2019 Oracle and/or its affiliates.26
Feature descriptions and design details
directly from the source
https://mysqlserverteam.com/

27. Copyright © 2019 Oracle and/or its affiliates.27
Thank you!
Norvald H. Ryeng
Software Development Senior Manager
MySQL Optimizer Team