以下是我的SQL和执行计划
explain SELECT
pos_uuid,
pos_bus_uuid,
pos_gather_time,
pos_real_time_status,
pos_odometer,
pos_drv_ic_card
FROM t_bus_position_20160306_20160308 WHERE pos_gather_time >= '2016-03-06 00:08:00' AND pos_gather_time <='2016-03-06 23:59:59' and pos_is_offset = 't' and pos_bus_uuid ='b478193370f24b8f9fe1' ORDER BY pos_gather_time ;
QUERY PLAN
Sort (cost=94141.44..94176.57 rows=14052 width=145)
Sort Key: pos_gather_time
-> Bitmap Heap Scan on t_bus_position_20160306_20160308 (cost=1039.36..93173.36 rows=14052 width=145)
Recheck Cond: (((pos_bus_uuid)::text = 'b478193370f24b8f9fe1'::text) AND (pos_gather_time >= '2016-03-06 00:08:00'::timestamp withou
t time zone) AND (pos_gather_time <= '2016-03-06 23:59:59'::timestamp without time zone))
Filter: pos_is_offset
-> Bitmap Index Scan on idx_check_index (cost=0.00..1035.85 rows=28103 width=0)
Index Cond: (((pos_bus_uuid)::text = 'b478193370f24b8f9fe1'::text) AND (pos_gather_time >= '2016-03-06 00:08:00'::timestamp wi
thout time zone) AND (pos_gather_time <= '2016-03-06 23:59:59'::timestamp without time zone))
It is several important parameters of the configuration file
shared_buffers =2GB
huge_pages = try
temp_buffers = 16MB
work_mem =128MB
maintenance_work_mem =512MB
max_stack_depth = 8MB
dynamic_shared_memory_type = posix
checkpoint_segments = 32
max_connections = 100
seq_page_cost = 1.0
random_page_cost = 4
cpu_tuple_cost = 0.01
cpu_index_tuple_cost = 0.005
cpu_operator_cost = 0.0025
effective_cache_size = 2GB
\d t_bus_position_20160303_20160305
Table "public.t_bus_position_20160303_20160305"
Column | Type | Modifiers
------------------------+--------------------------------+-----------
pos_uuid | character varying(20) | not null
pos_line_uuid | character varying(20) |
pos_line_type | character varying(20) |
pos_bus_uuid | character varying(20) |
pos_dev_uuid | character varying(20) |
pos_sta_uuid | character varying(20) |
pos_drv_ic_card | character varying(30) |
pos_lng | character varying(30) |
pos_lat | character varying(30) |
pos_bus_speed | character varying(20) |
pos_real_time_status | character varying(20) |
pos_gather_time | timestamp(6) without time zone |
pos_storage_time | timestamp(6) without time zone |
pos_is_offset | character varying(20) |
pos_is_overspeed | character varying(1) |
pos_cursor_over_ground | character varying(20) |
pos_all_alarms | character varying(30) |
pos_is_in_station | character varying(1) |
pos_closed_alarms | character varying(30) |
pos_dis_to_pre_i | integer |
pos_odometer_i | bigint |
pos_relative_location | real |
pos_dis_to_pre | real |
pos_odometer | double precision |
pos_gather_date | date |
Indexes:
"tp_20160303_20160305_pos_uuid_idx" UNIQUE, btree (pos_uuid)
"tp_20160303_20160305_pos_bus_gather_date_idx" btree ((pos_bus_uuid::text || extract_date(pos_gather_time)))
"tp_20160303_20160305_pos_bus_gather_idx" btree (pos_bus_uuid, extract_date(pos_gather_time))
"tp_20160306_20160308_pos_bus_gather_time_idx" btree (pos_gather_time, pos_gather_time DESC)
Check constraints:
"t_bus_position_20160303_20160305_pos_gather_time_check" CHECK (pos_gather_time >= '2016-03-03 00:00:00'::timestamp without time zone AND
pos_gather_time < '2016-03-06 00:00:00'::timestamp without time zone)
Inherits: t_bus_position
**This is the is Boolean pos_is_offset implementation plan**
explain SELECT
pos_uuid,
pos_bus_uuid,
pos_gather_time,
pos_real_time_status,
pos_odometer,
pos_drv_ic_card
FROM t_bus_position_20160306_20160308 WHERE pos_gather_time >= '2016-03-06 00:08:00' AND pos_gather_time <='2016-03-06 23:59:59' and pos_is_offset = 't' and pos_bus_uuid ='b478193370f24b8f9fe1' ORDER BY pos_gather_time ;
QUERY PLAN
---------------------------------------------------------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------
Sort (cost=94141.44..94176.57 rows=14052 width=145)
Sort Key: pos_gather_time
-> Bitmap Heap Scan on t_bus_position_20160306_20160308 (cost=1039.36..93173.36 rows=14052 width=145)
Recheck Cond: (((pos_bus_uuid)::text = 'b478193370f24b8f9fe1'::text) AND (pos_gather_time >= '2016-03-06 00:08:00'::timestamp withou
t time zone) AND (pos_gather_time <= '2016-03-06 23:59:59'::timestamp without time zone))
Filter: pos_is_offset
-> Bitmap Index Scan on idx_check_index (cost=0.00..1035.85 rows=28103 width=0)
Index Cond: (((pos_bus_uuid)::text = 'b478193370f24b8f9fe1'::text) AND (pos_gather_time >= '2016-03-06 00:08:00'::timestamp wi
thout time zone) AND (pos_gather_time <= '2016-03-06 23:59:59'::timestamp without time zone))
(7 rows)
答案 0 :(得分：1)
两件事：
这里的扰流器看起来像是pos_is_offset列，它是唯一不属于所用索引的字段。这是一个非常重要的查询（足以修改索引并添加该布尔值），这可能有助于加快查询速度。
第一次运行缓慢，后续运行速度更快的查询清楚地表明正在从磁盘读取表，缓存会影响后续运行的速度。加快查询速度变得非常困难，特别是如果你的RAM +大数据集不足。
另外，你真的需要输出14k行吗？或者表格不经常ANALYSE'd？如果运行EXPLAIN ANALYSE时的行计数给出的数字太远，那么您可能还需要ANALYSE该表。