I am wondering on the benefit of adding a clustered index to make sure data is sequential and then using a NonClustered index for all the queries. Since you mentioned that the data column is frequently Updated the NC will need to be rebuilt frequently as well. So why not adjust the fill factor and simply rebuild a clustered index. In fact you could improve ETL performance by partition switching so you could still get all the benefits of clustered index without the impact on write performance. I dont think rebuilding indexes by partition is an online operation in 2008 r2 but try it out anyway.