A table has 200000 rows.When I update one row with primary key, it cost 0.2s. I checked the excute plan, and found something can not be understood. In the excute plan. the clusted index update cost 1%, but before clusted index update there is a nested loops(inner join).seems constant scan->compute scalar then clusted index seek.The clusted index seek cost 99%. It reads 2159 pages. I found the order of where clause in the clusted index seek is different from the order of primary key. Why???Maybe that is the reason that result of full scan. Because I use the primary key to update, I think the clusted index update is enough. What is the use of constant scan and compute scalar? Hope get some help from you. Thanks.