Sql Sever性能优化之指定索引

背景：生产环境SQL语句查询过慢（数据总量在350万左右），日志中心一直报警

解决过程：分析无果后，求助于公司的DBA，DBA分析后建议在语句中指定索引

解决：在SQL语句中指定索引，效果相当明显，亲测有效

优化前SQL：

SELECT  ROW_NUMBER() OVER ( ORDER BY sc.ModifyTime DESC ) AS RowNumber ,
        sc.CommodityCode AS ChannelCommodityCode ,
        sc.OutShopCommodityCode AS OutItemId ,
        sc.IsShelf ,
        sc.OriginPrice ,
        sc.SalePrice ,
        sc.CreateBy ,
        sc.CreateTime ,
        sc.ModifyTime ,
        sc.StoreCode ,
        sc.ShopCommodityId AS ChannelStoreMappingId ,
        sc.ShopCommodityName AS O2OCommodityName ,
        sc.ShopCommodityTitle AS O2OCommodityTitle ,
        sc.Remark AS Comment ,
        sc.OutShopSPUCode ,
        sc.ModifyBy ,
        sc.ChannelCode
FROM    dbo.Channel_ShopCommodity AS sc WITH ( NOLOCK )
WHERE   sc.Disabled = 0
        AND sc.ChannelCode = '9000000009'
        AND sc.CityId = 'eabbe02f-59e0-46e6-90e7-cd8a89dbb98f';

优化后SQL（指定使用索引IX_CityId）：

SELECT  ROW_NUMBER() OVER ( ORDER BY sc.ModifyTime DESC ) AS RowNumber ,
        sc.CommodityCode AS ChannelCommodityCode ,
        sc.OutShopCommodityCode AS OutItemId ,
        sc.IsShelf ,
        sc.OriginPrice ,
        sc.SalePrice ,
        sc.CreateBy ,
        sc.CreateTime ,
        sc.ModifyTime ,
        sc.StoreCode ,
        sc.ShopCommodityId AS ChannelStoreMappingId ,
        sc.ShopCommodityName AS O2OCommodityName ,
        sc.ShopCommodityTitle AS O2OCommodityTitle ,
        sc.Remark AS Comment ,
        sc.OutShopSPUCode ,
        sc.ModifyBy ,
        sc.ChannelCode
FROM    dbo.Channel_ShopCommodity AS sc WITH ( NOLOCK, INDEX= [IX_CityId] )
WHERE   sc.Disabled = 0
        AND sc.ChannelCode = '9000000009'
        AND sc.CityId = 'eabbe02f-59e0-46e6-90e7-cd8a89dbb98f';

注意事项：使用指定索引后，必须保证Where条件中有这个筛选条件，否则索引不生效，查询语句耗时仍然会很长

SELECT  ROW_NUMBER() OVER ( ORDER BY sc.ModifyTime DESC ) AS RowNumber ,
        sc.CommodityCode AS ChannelCommodityCode ,
        sc.OutShopCommodityCode AS OutItemId ,
        sc.IsShelf ,
        sc.OriginPrice ,
        sc.SalePrice ,
        sc.CreateBy ,
        sc.CreateTime ,
        sc.ModifyTime ,
        sc.StoreCode ,
        sc.ShopCommodityId AS ChannelStoreMappingId ,
        sc.ShopCommodityName AS O2OCommodityName ,
        sc.ShopCommodityTitle AS O2OCommodityTitle ,
        sc.Remark AS Comment ,
        sc.OutShopSPUCode ,
        sc.ModifyBy ,
        sc.ChannelCode
FROM    dbo.Channel_ShopCommodity AS sc WITH ( NOLOCK, INDEX= [IX_CityId] )
WHERE   sc.Disabled = 0
        AND sc.ChannelCode = '9000000009'
        AND sc.StoreCode IN ( '10000723' );

以下截图为上面语句的耗时：

去除指定索引后再次执行

SELECT  ROW_NUMBER() OVER ( ORDER BY sc.ModifyTime DESC ) AS RowNumber ,
        sc.CommodityCode AS ChannelCommodityCode ,
        sc.OutShopCommodityCode AS OutItemId ,
        sc.IsShelf ,
        sc.OriginPrice ,
        sc.SalePrice ,
        sc.CreateBy ,
        sc.CreateTime ,
        sc.ModifyTime ,
        sc.StoreCode ,
        sc.ShopCommodityId AS ChannelStoreMappingId ,
        sc.ShopCommodityName AS O2OCommodityName ,
        sc.ShopCommodityTitle AS O2OCommodityTitle ,
        sc.Remark AS Comment ,
        sc.OutShopSPUCode ,
        sc.ModifyBy ,
        sc.ChannelCode
FROM    dbo.Channel_ShopCommodity AS sc WITH ( NOLOCK )
WHERE   sc.Disabled = 0
        AND sc.ChannelCode = '9000000009'
        AND sc.StoreCode IN ( '10000723' );

我们会发现耗时明显减少。

由此总结：如果确保某个条件一定会传并且该字段上加了索引，可以使用指定索引提升性能，但是使用指定索引要慎重，避免因为指定索引的Where条件不传又引起的查询耗时变长