SQL Server将不同于特定值的连续行分组，并计算间隔

我有这张桌子

我想构建一个查询以获得以下结果：

StartID   EndID   BeltBit   StartStamp                EndStamp                Interval (in sec)
-----------------------------------------------------------------------------------------------
4561      4564       0      2014 03 10 07:56:03.953   2014 03 10 13:27:12.113    19869
4565      4566       1      2014 03 10 13:27:12.113   2014 03 10 13:54:08.727     1616
4567      4567       0      2014 03 10 13:54:08.727   2014 03 10 13:56:16.523      128
4568      4570       1      2014 03 10 13:56:16.523   2014 03 10 13:56:39.580       23

我想分组的贝尔特比特（所有连续与相同的贝尔特比特状态）和计算他们的间隔

请注意：此表包含数百万条记录，因此多个连接将非常缓慢

我发现这样的查询使用sql-2012滞后和超前函数请告诉我是否有任何增强功能

二级

---

如果没有正确的索引，则多个联接的速度会很慢，因此需要扫描表。调优包括创建正确的索引。我认为可以使用分区而不使用连接来解决这个问题

StartID   EndID   BeltBit   StartStamp                EndStamp                Interval (in sec)
-----------------------------------------------------------------------------------------------
4561      4564       0      2014 03 10 07:56:03.953   2014 03 10 13:27:12.113    19869
4565      4566       1      2014 03 10 13:27:12.113   2014 03 10 13:54:08.727     1616
4567      4567       0      2014 03 10 13:54:08.727   2014 03 10 13:56:16.523      128
4568      4570       1      2014 03 10 13:56:16.523   2014 03 10 13:56:39.580       23

  ;   WITH Level1 
   AS ( SELECT ID , 
               BeltBit , 
               timeStamp AS EventDateTime , 
               LAG(BeltBit, 1, null) 
                  OVER ( PARTITION BY 1 
                     ORDER BY timestamp ) AS LastEvent , 
               LEAD(BeltBit, 1, null) 
                  OVER ( PARTITION BY 1 
                     ORDER BY timestamp) AS NextEvent 
                         FROM MachineLog ),

   AS ( SELECT ID , 
               BeltBit , 
               EventDateTime , 
               LastEvent , 
               NextEvent 
        FROM   Level1 
        WHERE  NOT ( BeltBit = 0 
                     AND LastEvent = 0
                   ) 
           AND NOT ( BeltBit = 1 
                     AND NextEvent = 1 
                   ) 

      ),
 Level3 
   AS ( SELECT ID , 
               BeltBit , 
               EventDateTime , 
               DATEDIFF(second, EventDateTime, 
                        LEAD(EventDateTime) 
                           OVER ( PARTITION BY 1 
                              ORDER BY EventDateTime )) AS Seconds 
        FROM Level2 
       ) 

       select * from Level3