执行时间要7205ms，语句如下:
select top 30 this_.id as id10_0_, this_.objid as objid10_0_, this_.objname
as objname10_0_, this_.mid as mid10_0_, this_.logtype as logtype10_0_, this_.logdesc as logdesc10_0_, this_.submitor as submitor10_0_, this_.submitdate as submitdate10_0_,
this_.submittime as submittime10_0_, this_.submitip as submitip10_0_, this_.col1 as col11_10_0_,this_.col2 as col12_10_0_, this_.col3 as col13_10_0_ from log this_ where not exists (select 'X' from Delobj del where del.objid=this_.id and del.objtable='Log') order by this_.submitdate desc, this_.submittime desc执行sql语句的执行计划:
46万行数据。排序的成本很高，我们发现是通过：
[dbo].[log].submitdate 降序, [dbo].[log].submittime 降序
重构聚集索引。
是必须在（submitdate，submittime）建立聚集索引，由于前期在设计表时的不合理，造成了时间字段教长，有18个字节，大大超过了8个字节的datetime类型，不太适合建立聚集索引。
适当提高速度。
总结：
1，字段类型和长短的是很重要，不然在后续修改和优化是很困难。
2，对与聚集索引选择一般在需要排序，进行范围选择和group by等字段上建立