如何诊断高扫描计数

Question

我很好奇如何诊断永久表和临时表中的高扫描计数。我在运行存储过程时打开了统计信息，我试图弄清楚为什么扫描计数如此之高。我有实际的查询计划，SSMS 建议添加一些缺少的索引，但此时没有别的了。我可以考虑哪些策略来诊断这些问题？

https://www.brentozar.com/pastetheplan/?id=HksFRMLw-

SQL Server Execution Times:
CPU time = 0 ms,  elapsed time = 0 ms.
Table 'disbursement_check'. Scan count 9, logical reads 2281, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'cif'. Scan count 9, logical reads 490, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'voucher'. Scan count 9, logical reads 3067, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'cif_loan_relationship'. Scan count 9, logical reads 890, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'loanacct'. Scan count 45, logical reads 44124, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'Workfile'. Scan count 0, logical reads 0, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'Worktable'. Scan count 0, logical reads 0, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'loanacct_trans_history'. Scan count 1166870, logical reads 3654129, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table '#B17AB456'. Scan count 8, logical reads 1180568, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table '#BAD74620'. Scan count 2423794, logical reads 2486442, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table '#BBCB6A59'. Scan count 62661, logical reads 125306, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'loan_class'. Scan count 2, logical reads 4, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'lookup_table'. Scan count 2, logical reads 6, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'loanacct_detail'. Scan count 9, logical reads 1348, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'Worktable'. Scan count 0, logical reads 0, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

SQL Server Execution Times:
CPU time = 59160 ms,  elapsed time = 9877 ms.
Table '#TargetTransactions_________________________________________________________________________________________________0000000019E6'. Scan count 1, logical reads 1, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

Answer 1

好问题。

根据一些统计数据，我倾向于认为存在一些索引或缺少索引，这是有问题的。查询计划可能会有所启发，尽管截至发帖时我还没有。

我认为可以安全地假设loanacct_trans_history 表是建议索引的表。此外，这 3 个临时表似乎也做了很多工作......

下面的链接将帮助您充分了解索引以回答您的问题。

请观看这个由世界级 DBA Brent Ozar 主持的 90 分钟免费网络研讨会：https://www.brentozar.com/archive/2016/10/think-like-engine-class-now-free-open-source/

Answer 2

我正试图弄清楚为什么扫描计数如此之高。

和

表“loanacct_trans_history”。扫描计数 1166870，逻辑读取 3654129，物理读取 0，预读读取 0，lob 逻辑读取 0， lob 物理读取 0，lob 预读读取 0。表 '#B17AB456'。扫描 计数 8，逻辑读取 1180568，物理读取 0，预读读取 0， lob 逻辑读取 0，lob 物理读取 0，lob 预读读取 0。

因为我们（还没有）实际的执行计划，最好的假设是

[ 1 ] 一些Index Seek 产生大量范围扫描（很可能），其中Execution Count 数量很多（可能是FORCESEEK 表提示强制）

[ 2 ] 和/或由以下运算符组合生成的范围扫描或全扫描

... - Nested Loops (Join) - Any Data Access Operator
           |
           ---------------- Index Seek / *Scan loanacct_trans_history 

在这种情况下，我假设 Estimated execution count 是 1 由实际/真实 Execution Count 高。

[ 3 ] 从演示点 [1] 和 [2] 的执行计划中裁剪，这与表 loanacct_trans_history 上的高扫描计数有关：

表“loanacct_trans_history”。扫描计数 1166870，...

似乎以下语句会产生如此高的扫描计数：

INSERT INTO #TargetTransactions (Id, CreditServiceFeeTransactionId, CustomerCifNo, VoucherId)
 SELECT mt.transrefno, rt.CreditServiceFeeTransactionId, contact.cifno, rt.VoucherId
 FROM dbo.getMyTransactions
 (
 @AssignedDealerIds,
 @DealerIds
 ) mt
 JOIN loanacct_trans_history th ON th.transrefno = mt.transrefno
...

几个简单的解决方案：在此语句的末尾，我将尝试添加以下 查询 提示：

INSERT #T...
SELECT ...
...
OPTION(HASH JOIN)
-- Or OPTION(MERGE JOIN)
-- Or OPTION(HASH JOIN, MERGE JOIN)