我对PostgreSQL 中 index scan 与 seq scan 的对比学习

开始

数据量很小的时候，我们可以看到，seq scan 比 index scan 更加有效。那是因为 index scan 至少要发生两次I/O，一次是 读取索引块， 一次是读取数据块。当index 很大的时候，情况可能会更加复杂。 

postgres=# select a.relpages, a.reltuples, a.relfilenode,a.reltype,b.typname from pg_class a, pg_type b where a.relname like 'gaotab%' and a.reltype=b.oid;    
 relpages | reltuples | relfilenode | reltype | typname     
----------+-----------+-------------+---------+---------    
        1 |      100 |       16396 |   16386 | gaotab    

数据量为 100条记录。

预估成本：

postgres=# set session enable_seqscan=false;                    
SET                    
postgres=# explain select name from gaotab where id=50;                    
                             QUERY PLAN                                                  
---------------------------------------------------------------------                    
 Index Scan using idx_id on gaotab  (cost=0.00..8.27 rows=1 width=5)                    
   Index Cond: (id = 50)                    
(2 rows)                    

postgres=# set session enable_seqscan=true;        
SET        
        
postgres=# explain select name from gaotab where id=50;        
                      QUERY PLAN                              
------------------------------------------------------        
 Seq Scan on gaotab  (cost=0.00..2.25 rows=1 width=5)        
   Filter: (id = 50)        
(2 rows)        

实际执行：

postgres=# set session enable_seqscan=false;        
SET        
postgres=# explain analyze select name from gaotab where id=50;        
                                                  QUERY PLAN                            
                                       
--------------------------------------------------------------------------------        
-------------------------------        
 Index Scan using idx_id on gaotab  (cost=0.00..8.27 rows=1 width=5) (actual tim        
e=0.112..0.113 rows=1 loops=1)        
   Index Cond: (id = 50)        
 Total runtime: 0.133 ms        
(3 rows)        

postgres=# set session enable_seqscan=true;        
SET        
postgres=# explain analyze select name from gaotab where id=50;        
                                           QUERY PLAN                                   
                        
--------------------------------------------------------------------------------        
----------------        
 Seq Scan on gaotab  (cost=0.00..2.25 rows=1 width=5) (actual time=0.014..0.018         
rows=1 loops=1)        
   Filter: (id = 50)        
   Rows Removed by Filter: 99        
 Total runtime: 0.034 ms        
(4 rows)        

等到数据量大的时候，就是截然不同了。

数据为1000条记录时，通过查询可以看到，已经跨越了7个page：

postgres=# analyze;    
ANALYZE    
postgres=# select a.relpages, a.reltuples, a.relfilenode,a.reltype,b.typname from pg_class a, pg_type b where a.relname like 'gaotab%' and a.reltype=b.oid;    
 relpages | reltuples | relfilenode | reltype | typname     
----------+-----------+-------------+---------+---------    
        7 |      1000 |       16396 |   16386 | gaotab    
(1 row)    
    
postgres=#     

再次预估成本，此时seq scan 已经开始变得不划算了：

postgres=# set session enable_seqscan=false;
SET
postgres=# explain select name from gaotab where id=50;
                             QUERY PLAN                              
---------------------------------------------------------------------
 Index Scan using idx_id on gaotab  (cost=0.00..8.27 rows=1 width=6)
   Index Cond: (id = 50)
(2 rows)

postgres=# set session enable_seqscan=true;
SET
postgres=# explain select name from gaotab where id=50;
                             QUERY PLAN                              
---------------------------------------------------------------------
 Index Scan using idx_id on gaotab  (cost=0.00..8.27 rows=1 width=6)
   Index Cond: (id = 50)
(2 rows)

postgres=# set session enable_indexscan=false;
SET
postgres=# explain select name from gaotab where id=50;
                             QUERY PLAN                              
---------------------------------------------------------------------
 Bitmap Heap Scan on gaotab  (cost=4.26..8.27 rows=1 width=6)
   Recheck Cond: (id = 50)
   ->  Bitmap Index Scan on idx_id  (cost=0.00..4.26 rows=1 width=0)
         Index Cond: (id = 50)
(4 rows)

postgres=# set session enable_bitmapscan=false;
SET
postgres=# explain select name from gaotab where id=50;
                      QUERY PLAN                       
-------------------------------------------------------
 Seq Scan on gaotab  (cost=0.00..19.50 rows=1 width=6)
   Filter: (id = 50)
(2 rows)

postgres=# 

实际执行

postgres=# set session enable_seqscan=false;
SET
postgres=# explain analyze select name from gaotab where id=50;
                                                            QUERY PLAN                                      
                       
------------------------------------------------------------------------------------------------------------
-----------------------
 Index Scan using idx_id on gaotab  (cost=10000000000.00..10000000008.27 rows=1 width=6) (actual time=0.020.
.0.022 rows=1 loops=1)
   Index Cond: (id = 50)
 Total runtime: 0.051 ms
(3 rows)

postgres=# set session enable_seqscan=true;
SET
postgres=# set session enable_indexscan=false;
SET
postgres=# set session enable_bitmapscan=false;
SET
postgres=# explain analyze select name from gaotab where id=50;
                                           QUERY PLAN                                            
-------------------------------------------------------------------------------------------------
 Seq Scan on gaotab  (cost=0.00..19.50 rows=1 width=6) (actual time=0.015..0.095 rows=1 loops=1)
   Filter: (id = 50)
   Rows Removed by Filter: 999
 Total runtime: 0.109 ms
(4 rows)

postgres=# 

[作者：技术者高健@博客园  mail: luckyjackgao@gmail.com ]

结束