SQL Server should create clustered index before non-clustered index in exists table

版权声明：複製請註明來源 https://blog.csdn.net/weixin_39392627/article/details/86654433

env:Windows Server 2016

       SQL Server 2016 SP2

 

有一張表格要補上clustered index與non-clustered index。這是一張既有表格且有4億筆資料，裡面也有一支non-clustered index。

統計資訊的更新程度是80000000筆。

 

1.執行create non-clustered index後再執行clustered index

command:

CREATE NONCLUSTERED INDEX [IX_TESTId_CreatOn]
ON [dbo].[TESTTB] ([TESTId],[CreatOn])
INCLUDE ([Id],[LISTId],[Amount],[Revenue],[Session])
WITH(ONLINE=ON, DATA_COMPRESSION=ON)
GO

CREATE CLUSTERED INDEX [IX_CreatOn] ON [dbo].[TESTTB]
(
[CreatOn] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, DROP_EXISTING = OFF, ONLINE = ON, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON, DATA_COMPRESSION=PAGE) ON [PRIMARY]
GO

在建立clustered index花費許多時間，主要是掃描既有索引造成。利用SSMS建立cluster index會提示其他所引也會重建。

執行次序	servername	db_date	table_name	idx_id	index_name	row_count	data_used_mb	Exec(Mins)
1	TESTDB01	TESTDB01	dbo.TESTTB	7	IX_TESTId_CreatOn	404502153	17368	51
2	TESTDB01	TESTDB01	dbo.TESTTB	1	IX_CreatOn	404502153	47358	94

 

2.這次讓clustered index優先建立

command:

CREATE CLUSTERED INDEX [IX_CreatOn] ON [dbo].[TESTTB]
(
[CreatOn] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, DROP_EXISTING = OFF, ONLINE = ON, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON, DATA_COMPRESSION=PAGE) ON [PRIMARY]
GO

CREATE NONCLUSTERED INDEX [IX_TESTId_CreatOn]
ON [dbo].[TESTTB] ([TESTId],[CreatOn])
INCLUDE ([Id],[LISTId],[Amount],[Revenue],[Session])
WITH(ONLINE=ON, DATA_COMPRESSION=ON)
GO

這次可以發現建立clustered index時間大幅縮短。

執行次序	servername	db_date	table_name	idx_id	index_name	row_count	data_used_mb	Exec(Mins)
1	TESTDB01	TESTDB01	dbo.TESTTB	1	IX_CreatOn	404502153	47358	50
2	TESTDB01	TESTDB01	dbo.TESTTB	7	IX_TESTId_CreatOn	404502153	17368	43

結論:

若要在既有表格建立索引，clustered index要優先處理。

若表格內已經有許多non-clustered index，是否要建立空表後再進行資料複製移轉?這是值得考慮的問題。