尽管索引MSSQL 2005非常缓慢的更新

编辑：我们正在移动服务器，我刚刚在新的服务器上进行了测试。这里没有性能问题。这似乎是一个动力不足，组织严密的服务器。尽管索引MSSQL 2005非常缓慢的更新

昨晚我们的一个流程突然跑得很慢。缓慢的一步被追查到一张表格上的更新声明，这个表格当然不是太聪明的索引。

所以今天我已经为所有涉及的表格添加了索引，但我仍然表现糟糕。我真的不明白 - 可能我还在做一些比聪明更少的事情。欢迎任何建议。

更新如下：

update test_HDM_RTT 
set patient_district_no = b.legacy_number 
from test_HDM_RTT a 
inner join PHD.migration.PatScope b 
on a.patient_pas_no = b.TrustNumber

patscope是2474147行，test_hdm_rtt是表815278

定义：

CREATE TABLE [dbo].[test_HDM_RTT](
    [pk_episode_id] [int] NULL, 
    [pk_event_id] [int] NOT NULL, 
    [activity_date] [datetime] NULL, 
    [activity_datetime] [datetime] NULL, 
    [activity_subtype1] [nvarchar](50) NULL, 
    [activity_subtype1_code] [nvarchar](50) NULL, 
    [activity_subtype2] [nvarchar](50) NULL, 
    [activity_subtype2_code] [nvarchar](50) NULL, 
    [activity_type] [nvarchar](50) NULL, 
    [activity_type_code] [nvarchar](50) NULL, 
    [clock_start_date] [datetime] NULL, 
    [clock_stop_date] [datetime] NULL, 
    [dir_code] [nvarchar](10) NULL, 
    [div_code] [nvarchar](10) NULL, 
    [episode_id_ext] [nvarchar](50) NULL, 
    [episode_id_appt] [nvarchar](50) NULL, 
    [episode_id_ref] [nvarchar](50) NULL, 
    [episode_id_ref_medway] [nvarchar](50) NULL, 
    [episode_id_wl] [nvarchar](50) NULL, 
    [erod] [datetime] NULL, 
    [nhs_number] [nvarchar](20) NULL, 
    [patient_id] [int] NULL, 
    [patient_district_no] [nvarchar](20) NULL, 
    [patient_pas_no] [nvarchar](50) NULL, 
    [pathway_id] [nvarchar](50) NULL, 
    [pct_code] [nvarchar](10) NULL, 
    [ref_source_code] [nvarchar](10) NULL, 
    [rtt_episode_id] [nvarchar](50) NULL, 
    [rtt_outcome_code] [nvarchar](50) NULL, 
    [rtt_outcome_desc] [nvarchar](50) NULL, 
    [rtt_start_date] [datetime] NULL, 
    [rtt_start_ind] [nvarchar](10) NULL, 
    [rtt_stop_date] [datetime] NULL, 
    [site_code] [nvarchar](10) NULL, 
    [spec_natcode] [nvarchar](10) NULL, 
    [spec_pascode] [nvarchar](10) NULL, 
    [transfer_text] [nvarchar](100) NULL, 
    [op_rtt_count] [int] NULL, 
    [app_rec_date] [datetime] NULL, 
    [cons_code] [varchar](10) NULL 
) ON [PRIMARY] 

GO 

SET ANSI_PADDING OFF 
GO 


USE [PIP] 
/****** Object: Index [pk_event_id_clustered] Script Date: 03/06/2013 14:46:52 ******/ 
CREATE CLUSTERED INDEX [pk_event_id_clustered] ON [dbo].[test_HDM_RTT] 
(
    [pk_event_id] ASC 
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY] 
GO 


USE [PIP] 
/****** Object: Index [idx_episode_id_appt] Script Date: 03/06/2013 14:46:52 ******/ 
CREATE NONCLUSTERED INDEX [idx_episode_id_appt] ON [dbo].[test_HDM_RTT] 
(
    [episode_id_appt] ASC 
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY] 
GO 


USE [PIP] 
/****** Object: Index [idx_episode_id_ref] Script Date: 03/06/2013 14:46:52 ******/ 
CREATE NONCLUSTERED INDEX [idx_episode_id_ref] ON [dbo].[test_HDM_RTT] 
(
    [episode_id_ref] ASC 
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY] 
GO 


USE [PIP] 
/****** Object: Index [idx_episode_id_wl] Script Date: 03/06/2013 14:46:52 ******/ 
CREATE NONCLUSTERED INDEX [idx_episode_id_wl] ON [dbo].[test_HDM_RTT] 
(
    [episode_id_wl] ASC 
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY] 
GO 


USE [PIP] 
/****** Object: Index [patient_pas_no] Script Date: 03/06/2013 14:46:52 ******/ 
CREATE NONCLUSTERED INDEX [patient_pas_no] ON [dbo].[test_HDM_RTT] 
(
    [patient_pas_no] ASC 
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY] 
GO

和

USE [PHD] 
GO 

/****** Object: Table [migration].[PatScope] Script Date: 03/06/2013 14:47:57 ******/ 
SET ANSI_NULLS ON 
GO 

SET QUOTED_IDENTIFIER ON 
GO 

SET ANSI_PADDING ON 
GO 

CREATE TABLE [migration].[PatScope](
    [RID] [varchar](7) NOT NULL, 
    [Number] [varchar](17) NOT NULL, 
    [TrustNumber] [varchar](10) NULL, 
    [NumberType] [nvarchar](10) NULL, 
    [legacy_number] [varchar](10) NULL 
) ON [PRIMARY] 

GO 

SET ANSI_PADDING OFF 
GO 


USE [PHD] 
/****** Object: Index [TrustNoClustered] Script Date: 03/06/2013 14:47:57 ******/ 
CREATE CLUSTERED INDEX [TrustNoClustered] ON [migration].[PatScope] 
(
    [TrustNumber] ASC 
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY] 
GO 


USE [PHD] 
/****** Object: Index [TrustNo] Script Date: 03/06/2013 14:47:57 ******/ 
CREATE NONCLUSTERED INDEX [TrustNo] ON [migration].[PatScope] 
(
    [TrustNumber] ASC, 
    [Number] ASC 
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY] 
GO 


USE [PHD] 
/****** Object: Index [TrustNumber_legacy_lookup] Script Date: 03/06/2013 14:47:57 ******/ 
CREATE UNIQUE NONCLUSTERED INDEX [TrustNumber_legacy_lookup] ON [migration].[PatScope] 
(
    [TrustNumber] ASC, 
    [legacy_number] ASC 
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY] 
GO

来源

2013-03-06 DanBennett

查看查询计划。它使用你期望的索引吗？检查碎片。即使你确定价值不一样，我仍然会按照马特的建议去做。我在a.patient_pas_no上看不到索引。 – Paparazzi 2013-03-06 15:09:24

@Blam a.patient_pas_no上的索引位于第一个代码块的最底部。我最初也错过了。 – 2013-03-06 15:11:36

@MattBusche你是对的 – Paparazzi 2013-03-06 15:13:06

显示执行计划&看看计划是否使用索引。如果没有，你可以强制使用特定的索引：

SET ANSI_NULLS OFF 
GO 

update test_HDM_RTT 
set patient_district_no = b.legacy_number 
from test_HDM_RTT a 
inner join PHD.migration.PatScope b WITH (INDEX(TrustNumber_legacy_lookup)) 
on a.patient_pas_no = b.TrustNumber 
where a.patient_district_no <> b.legacy_number

你会想看看你是否需要强制上test_HDM_RTT的索引，而不是因为它看起来像它已经打算做一个索引扫描TrustNumber_legacy_lookup以获取其数据。

来源

2013-03-06 17:32:42

个人而言，我会如果值与现有的值不一致，则更新。这应该加快查询

update test_HDM_RTT 
set patient_district_no = b.legacy_number 
from test_HDM_RTT a 
inner join PHD.migration.PatScope b 
on a.patient_pas_no = b.TrustNumber 
where a.patient_district_no <> b.legacy_number

我也检查了EXPLAIN结果（ctrl + l）您的查询可能使用了错误的指标。

来源

2013-03-06 14:56:04

谢谢，我同意 - 我通常也会。在这种情况下，虽然我们确定知道如果在patscope中找到一个值，它将不会与表中的值相同。 – DanBennett 2013-03-06 14:59:59

谢谢解释提示 - 我会试试看。 – DanBennett 2013-03-06 15:00:27

'where a.patient_district_no <> b.legacy_number'就是你错过了'NULL'值的问题 - 涉及'NULL'的比较返回NULL，除非你在执行更新之前'SET ANSI_NULLS OFF' 。 – 2013-03-06 17:18:44

尽管索引MSSQL 2005非常缓慢的更新

回答

相关问题