更新时间:2026-08-06 GMT+08:00
分享

操作步骤

无索引和有索引性能对比

  1. 使用root用户登录数据库。
  2. 查看test_table表执行计划。

    gaussdb=# EXPLAIN ANALYZE SELECT * FROM test_table WHERE email = 'user_500000@example.com';
    
                        QUERY PLAN                    
    --------------------------------------------------
     Data Node Scan  (cost=0.00..0.00 rows=0 width=0)
       Node/s: All datanodes
    (2 rows)
    Time: 167.579 ms

    从执行结果来看,执行时间需要167.579ms。

  3. 创建索引。

    gaussdb=# CREATE INDEX idx_test_table_email ON test_table(email);
    CREATE INDEX

  4. 再次查看test_table表执行计划。

    gaussdb=# EXPLAIN ANALYZE SELECT * FROM test_table WHERE email = 'user_500000@example.com';
                                                            QUERY PLAN                                                         
    ---------------------------------------------------------------------------------------------------------------------------
     Data Node Scan  (cost=0.00..0.00 rows=0 width=0)
       Node/s: All datanodes
    
     Remote SQL: SELECT id, name, email, created_at FROM public.test_table WHERE email::text = 'user_500000@example.com'::text
     Datanode Name: dn_6001
       [Bypass]
       Index Scan using idx_test_table_email on test_table  (cost=0.00..2.47 rows=1 width=46)
         Index Cond: ((email)::text = 'user_500000@example.com'::text)
    
     Datanode Name: dn_6002
       [Bypass]
       Index Scan using idx_test_table_email on test_table  (cost=0.00..2.47 rows=1 width=46)
         Index Cond: ((email)::text = 'user_500000@example.com'::text)
    
    (14 rows)
    
    
    Time: 18.467 ms

    添加索引后,通过与无索引时执行计划的对比,查询时间从原来的167.579ms缩短到18.467ms。

单列索引和复合索引的性能对比

  1. 使用root用户登录数据库。
  2. 创建单列索引。

    gaussdb=# CREATE INDEX idx_region ON sales_records(region_id);
    CREATE INDEX
    gaussdb=# CREATE INDEX idx_store ON sales_records(store_id);
    CREATE INDEX

  3. 查看执行计划。

    gaussdb=# EXPLAIN ANALYZE SELECT * FROM sales_records WHERE region_id = 5 AND store_id = 42;
                                                                             QUERY PLAN                                                                          
    -------------------------------------------------------------------------------------------------------------------------------------------------------------
     Data Node Scan  (cost=0.00..0.00 rows=0 width=0)
       Node/s: All datanodes
    
     Remote SQL: SELECT record_id, region_id, store_id, product_id, sale_date, amount, is_refund FROM public.sales_records WHERE region_id = 5 AND store_id = 42
     Datanode Name: dn_6001
       Bitmap Heap Scan on sales_records  (cost=1354.75..2501.82 rows=1160 width=31)
         Recheck Cond: ((store_id = 42) AND (region_id = 5))
         ->  BitmapAnd  (cost=1354.75..1354.75 rows=1160 width=0)
               ->  Bitmap Index Scan on idx_store  (cost=0.00..118.79 rows=10526 width=0)
                     Index Cond: (store_id = 42)
               ->  Bitmap Index Scan on idx_region  (cost=0.00..1235.13 rows=110237 width=0)
                     Index Cond: (region_id = 5)
    
     Datanode Name: dn_6002
       Bitmap Heap Scan on sales_records  (cost=1325.15..2406.59 rows=1087 width=31)
         Recheck Cond: ((store_id = 42) AND (region_id = 5))
         ->  BitmapAnd  (cost=1325.15..1325.15 rows=1087 width=0)
               ->  Bitmap Index Scan on idx_store  (cost=0.00..113.05 rows=10053 width=0)
                     Index Cond: (store_id = 42)
               ->  Bitmap Index Scan on idx_region  (cost=0.00..1211.31 rows=108088 width=0)
                     Index Cond: (region_id = 5)
    
    (22 rows)
    
    
    Time: 28.455 ms

    从执行结果来看,执行时间需要28.455ms。

  4. 创建复合索引。

    gaussdb=# CREATE INDEX idx_region_store ON sales_records(region_id, store_id);
    CREATE INDEX

  5. 再次查看执行计划。

    gaussdb=# EXPLAIN ANALYZE SELECT * FROM sales_records WHERE region_id = 5 AND store_id = 42;
                                                                             QUERY PLAN                                                                          
    -------------------------------------------------------------------------------------------------------------------------------------------------------------
     Data Node Scan  (cost=0.00..0.00 rows=0 width=0)
       Node/s: All datanodes
    
     Remote SQL: SELECT record_id, region_id, store_id, product_id, sale_date, amount, is_refund FROM public.sales_records WHERE region_id = 5 AND store_id = 42
     Datanode Name: dn_6001
       Bitmap Heap Scan on sales_records  (cost=16.54..1163.61 rows=1160 width=31)
         Recheck Cond: ((region_id = 5) AND (store_id = 42))
         ->  Bitmap Index Scan on idx_region_store  (cost=0.00..16.25 rows=1160 width=0)
               Index Cond: ((region_id = 5) AND (store_id = 42))
    
     Datanode Name: dn_6002
       Bitmap Heap Scan on sales_records  (cost=15.79..1097.23 rows=1087 width=31)
         Recheck Cond: ((region_id = 5) AND (store_id = 42))
         ->  Bitmap Index Scan on idx_region_store  (cost=0.00..15.52 rows=1087 width=0)
               Index Cond: ((region_id = 5) AND (store_id = 42))
    
    (16 rows)
    
    
    Time: 6.856 ms

    通过对单列索引和复合索引执行计划的对比,查询时间从原来的28.455ms缩短到6.856ms。

相关文档