为什么 SQL Server 不使用我的非聚集索引并执行聚集索引扫描？

Question

这是我的完整表格：

    CREATE TABLE [dbo].[tblCrawlUrls](
    [cl_IdUrl] [int] IDENTITY(1,1) NOT NULL,
    [cl_CrawlNormalizedUrl] [nvarchar](200) NOT NULL,
    [cl_RooSiteId] [smallint] NOT NULL,
    [cl_ExploreDate] [datetime] NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_ExploreDate]  DEFAULT (sysutcdatetime()),
    [cl_LastCrawlDate] [datetime] NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_LastCrawlDate]  DEFAULT ('2000-08-11 15:18:47.407'),
    [cl_CrawlSource] [nvarchar](max) NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_CrawlSource]  DEFAULT ('null'),
    [cl_CrawlOrgUrl] [nvarchar](200) NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_CrawlOrgUrl]  DEFAULT ('null'),
    [cl_ExploredURL] [nvarchar](200) NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_ExploredURL]  DEFAULT ('null'),
    [cl_Ignored_By_Containing_Word] [bit] NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_Ignored_By_Containing_Word]  DEFAULT ((0)),
    [cl_CrawlFailedTimes] [int] NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_CrawlFailedTimes]  DEFAULT ((0)),
    [cl_TotalCrawlTimes] [int] NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_TotalCrawlTimes]  DEFAULT ((0)),
    [cl_UpdatedTimes] [int] NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_UpdatedTimes]  DEFAULT ((0)),
    [cl_DuplicateUrl_ByCanonical] [bit] NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_DuplicateUrl_ByCanonical]  DEFAULT ((0)),
    [cl_PageProcessed] [bit] NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_PageProcessed]  DEFAULT ((0)),
    [cl_LastProcessDate] [datetime] NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_LastProcessDate]  DEFAULT ('2000-08-11 15:18:47.407'),
    [cl_PossibleProductPage] [bit] NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_PossibleProductPage]  DEFAULT ((0)),
    [cl_CertainlyNotProductPage] [bit] NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_CertainlyNotProductPage]  DEFAULT ((0)),
    [cl_IsProductPage] [bit] NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_IsProductPage]  DEFAULT ((0)),
    [cl_Determined_Not_A_Product_Page] [bit] NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_NotProduct_Page]  DEFAULT ((0)),
    [cl_FreeCargo] [bit] NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_FreeCargo]  DEFAULT ((0)),
    [cl_ProductPrice_TL] [int] NOT NULL CONSTRAINT [DF_tblCrawlUrls_cl_ProductPrice]  DEFAULT ((0)),
    [cl_ProductCode] [nvarchar](200) NULL,
    [cl_ProductImageLink] [nvarchar](200) NULL,
    [cl_ProductIdCode] [nvarchar](200) NULL,
    [cl_ProductCategoriesAsText] [nvarchar](200) NULL,
    [cl_ProductDetailedExplanation] [nvarchar](max) NULL,
    [cl_ProductFeatures_Wrapped] [nvarchar](max) NULL,
 CONSTRAINT [PK_tblCrawlUrls] PRIMARY KEY CLUSTERED 
(
    [cl_IdUrl] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON, FILLFACTOR = 85) ON [PRIMARY]
) ON [PRIMARY] TEXTIMAGE_ON [PRIMARY]


CREATE TABLE [dbo].[tblRootSites](
    [cl_RootSiteId] [smallint] NOT NULL,
    [cl_SiteRootUrl] [nvarchar](200) NOT NULL,
    [cl_Disabled] [bit] NOT NULL CONSTRAINT [DF_tblRootSites_Disabled]  DEFAULT ((0)),
    [cl_AlexaRank_TR] [int] NOT NULL CONSTRAINT [DF_tblRootSites_cl_AlexaRank]  DEFAULT ((0)),
    [cl_RegisterTime] [datetime] NOT NULL CONSTRAINT [DF_tblRootSites_cl_RegisterTime]  DEFAULT (sysutcdatetime()),
    [cl_PriceDelimeter] [varchar](1) NOT NULL CONSTRAINT [DF_tblRootSites_cl_PriceDelimeter]  DEFAULT ('.'),
    [cl_PriceIgnoreDelimeter] [varchar](1) NOT NULL CONSTRAINT [DF_tblRootSites_cl_PriceIgnoreDelimeter]  DEFAULT (','),
    [cl_CertainProductPageDefiner] [nvarchar](200) NOT NULL CONSTRAINT [DF_tblRootSites_cl_CertainProductPageDefiner]  DEFAULT ('null'),
    [cl_ProductCategoryListing_Priority] [smallint] NOT NULL CONSTRAINT [DF_tblRootSites_cl_ProductCategoryListing_Priority]  DEFAULT ((0)),
    [cl_Ignore_Words_From_Category] [nvarchar](200) NOT NULL CONSTRAINT [DF_tblRootSites_cl_Ignore_From_Category]  DEFAULT ('null'),
    [cl_Ignore_Words_From_Urls] [nvarchar](200) NOT NULL CONSTRAINT [DF_tblRootSites_cl_Ignore_Urls_Parameters]  DEFAULT ('null'),
    [cl_Use_Last_Category_As_A_Product_Code] [bit] NOT NULL CONSTRAINT [DF_tblRootSites_cl_Use_Last_Category_As_A_Product_Code]  DEFAULT ((0)),
    [cl_Use_Custom_Character_Encoding] [int] NOT NULL CONSTRAINT [DF_tblRootSites_cl_Use_Custom_Character_Encoding]  DEFAULT ((0)),
    [cl_IgnorePages_Words] [nvarchar](200) NOT NULL CONSTRAINT [DF_tblRootSites_cl_IgnorePages_Words]  DEFAULT ('null'),
    [cl_IgnoreLastCategory] [bit] NOT NULL CONSTRAINT [DF_tblRootSites_cl_IgnoreLastCategory]  DEFAULT ((0)),
 CONSTRAINT [PK_tblRootSites] PRIMARY KEY CLUSTERED 
(
    [cl_RootSiteId] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON, FILLFACTOR = 85) ON [PRIMARY]
) ON [PRIMARY]

这是我的 SQL 查询：

 SELECT TOP 1000 cl_idurl,
                cl_roositeid,
                cl_crawlsource,
                cl_crawlorgurl
FROM   tblcrawlurls
WHERE  cl_pageprocessed = 0
       AND cl_totalcrawltimes > 0
       AND cl_certainlynotproductpage = 0
       AND cl_duplicateurl_bycanonical = 0
       AND cl_roositeid IN (SELECT cl_rootsiteid
                            FROM   tblrootsites
                            WHERE  cl_disabled = 0)
ORDER  BY cl_lastprocessdate ASC 

这是我假设会使用的索引，但它没有被使用。

    CREATE NONCLUSTERED INDEX [TCT-PP-LPD-CNPP-DUBC] ON [dbo].[tblCrawlUrls]
(
    [cl_TotalCrawlTimes] ASC,
    [cl_PageProcessed] ASC,
    [cl_LastProcessDate] ASC,
    [cl_CertainlyNotProductPage] ASC,
    [cl_DuplicateUrl_ByCanonical] ASC
)
INCLUDE (   [cl_IdUrl],
    [cl_RooSiteId],
    [cl_CrawlOrgUrl]) WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON, FILLFACTOR = 85)

在我的查询中，我将这些用作限制（在 where 子句中）

cl_pageprocessed
cl_totalcrawltimes
cl_certainlynotproductpage
cl_duplicateurl_bycanonical

我假设使用包含这些列的索引来确定可能的主键 ID 应该比总聚集索引扫描更快。但是 SQL Server 并没有这样做。

所以我的逻辑不正确，这就是我想要学习的。为什么 SQL Server 选择不使用我的索引？不使用我的索引意味着我的数据库和硬盘上有额外的不必要的负载。

下面是执行计划：

Answer 1

索引中的第一列是 cl_totalcrawltimes。在 WHERE 子句中，您使用此列来定义较低的范围 (cl_totalcrawltimes > 0)。

根据数据分布，这很可能会导致索引扫描而不是索引查找。此外， cl_crawlsource 不是索引的一部分（最好是包含部分），因此无论如何这都需要聚集索引键查找。

因此，SQL Server 会忽略索引，因为执行聚集索引扫描似乎更有效。

为获得最佳结果，使您查询的那些列与非聚集索引的第一列完全匹配。（下一列应该是不等式列还是按列排序取决于优化器喜欢什么，如果有的话。）