SQL Server查询需要3分40秒。如何提高执行时间？

Question

我的查询时间为3分40秒。

除了AirESet和ELT_B表之外，所有其他表都是查找表，这些表中的数据少得多。有趣的是，如果没有这些查找表，执行时间大约为30秒。我的理解是查找表包含的数据少得多，性能不会降低。

SELECT DISTINCT 
    RPGM.GroupID as RegionPerilGroupID,
    RPGM.GroupName as RPGroupName,
    LLGM.GroupID as LossLevelGroupID,
    LLG.GroupName As LLGroupName
FROM
    Acc.dbo.AIRESet ES 
JOIN 
    Acc.dbo.vw_RegionPerilGroup RPGM ON ES.RegionperilID = RPGM.RegionperilID 
                                     AND ModelLossFileID = 65 -- Line of Business
JOIN 
    AR.LA.ELT_B ELT WITH (NOLOCK) ON ES.EventNum = ELT.EventNum 
                                  AND ELT.Versionid = 215 
JOIN 
    AR.LA.LossLevelGroupMappings LLGM ON ELT.LossLevelID = LLGM. LossLevelID 
JOIN 
    AR.LA.LossLevelGroup LLG ON LLGM.GroupID = LLG.GroupID 
                             AND LLG.LosstypeId = 3 
JOIN 
    AR.LA.ReportMap RM ON RPGM.GroupId = RM.RegionPerilGroupID 
                       AND LLGM.GroupID = RM.LossLevelGroupID
JOIN 
    AR.LA.ReportMapFilterMappings RMFM ON RM.ReportMapID = RMFM.ReportMapID
                                       AND RMFM.FilterID = 15
ORDER BY 
    LossLevelGroupID, RegionPerilGroupID 

Answer 1

让我猜一下 请在查询中添加option (HASH JOIN)。

P.S。
SELECT distinct ...通常表示查询逻辑问题 请找出重复的原因。

......和 ​​-

1. 否。
   您不需要用于等式连接操作的索引。
   这是SQL Server，而不是MySQL，它支持HASH JOIN。

   此外 -
   使用索引检索大量数据会导致性能显着下降。

1. 否。
   GROUP BY / DISTINCT并不一定意味着SORT操作。
   它也可以使用复杂度为O（1）而不是O（n * log n）的HASH来实现。

统计数据的局限性

select @@version

Microsoft SQL Server 2014 - 12.0.4213.0 (X64) 
  Jun  9 2015 12:06:16 
  Copyright (c) Microsoft Corporation
  Express Edition (64-bit) on Windows NT 6.1 <X64> (Build 7601: Service Pack 1)

create table t (n int primary key)

;with   t (n) as (select 1 union all select n+1 from t where n < 1000)
insert into dbo.t select n from t option  (maxrecursion 0)

create statistics t_n on t (n)
update statistics t (t_n)

;with x as (select * from t where sqrt(n) = n)
select * from x x0,x x1,x x2,x x3,x x4,x x5

+---+---+---+---+---+---+
| n | n | n | n | n | n |
+---+---+---+---+---+---+
| 1 | 1 | 1 | 1 | 1 | 1 |
+---+---+---+---+---+---+



HASH JOIN Vs. LOOP JOIN

create table t (i int,j1 int,j2 int,j3 int,j4 int,j5 int)

;with   t (n) as (select 0 union all select n+1 from t where n < 999)
insert into dbo.t (i,j1,j2,j3,j4,j5) 
select      1+t0.n+1000*t1.n,1+t0.n,1+t0.n,1+t0.n,1+t0.n,1+t0.n
from        t t0,t t1 
option      (maxrecursion 0)

create table d1 (j int primary key,k int)
create table d2 (j int primary key,k int)
create table d3 (j int primary key,k int)
create table d4 (j int primary key,k int)
create table d5 (j int primary key,k int)

;with   t (n) as (select 0 union all select n+1 from t where n < 999)
insert into dbo.d1 (j,k) 
select      1+t0.n,1
from        t t0 
option      (maxrecursion 0)

insert into d2 select * from d1
insert into d3 select * from d1
insert into d4 select * from d1
insert into d5 select * from d1

select  sum (d1.k+d2.k+d3.k+d4.k+d5.k)

from            t 
        join    d1 on t.j1 = d1.j 
        join    d2 on t.j2 = d2.j 
        join    d3 on t.j3 = d3.j 
        join    d4 on t.j4 = d4.j 
        join    d5 on t.j5 = d5.j 

option (hash join)

select  sum (d1.k+d2.k+d3.k+d4.k+d5.k)

from            t 
        join    d1 on t.j1 = d1.j 
        join    d2 on t.j2 = d2.j 
        join    d3 on t.j3 = d3.j 
        join    d4 on t.j4 = d4.j 
        join    d5 on t.j5 = d5.j 

option (loop join)

HASH GROUP Vs.订单组

create table t (i int,j int)

;with   t (n) as (select 0 union all select n+1 from t where n < 9)
insert into dbo.t (i,j) 
select      t0.n,t0.n
from        t t0,t t1,t t2,t t3,t t4,t t5,t t6

select      distinct
            i,j
from        t
option      (HASH GROUP)
;

select      i,j
from        t
group by    i,j
option      (HASH GROUP)
;

select      distinct
            i,j
from        t
option      (ORDER GROUP)
;

select      i,j
from        t
group by    i,j
option      (ORDER GROUP)
;