I think reate snapshot is not a bad choice. Create index on major columns which you want to select from. And then use this index select data and create a snapshot. So the snapshot performs a temp table for you to handle these data. If the the data is as your estimate about 1 million,it's a better choice to tackle these data in a snapshot rather than in the base table.
统计参考物化视图
http://blog.csdn.net/suncrafted/archive/2009/06/26/4300358.aspx
http://topic.csdn.net/u/20091225/14/e225f4f6-03f6-4684-b82b-45d58e045ad5.html?seed=2012731555&r=62218282#r_62218282
Create index on major columns which you want to select from.
And then use this index select data and create a snapshot.
So the snapshot performs a temp table for you to handle these data.
If the the data is as your estimate about 1 million,it's a better choice
to tackle these data in a snapshot rather than in the base table.
1.用集合。
2.函数传参用NOCOPY bint。