Just to make sure we are all on same page, the way I've heard the Microsoft CAT engineers explain for row and page compression that the data is compressed both on the hard drive and in the data buffer. Only when the actual row is moved to the CPU is the row (or page) uncompressed (and recompressed if updated or inserted) -- hence the additional CPU load. Reviewing my PASS2009 notes, from the DBA409 (Customer Experiences With Data Compression) briefing by SQLCAT engineers, their recommendation was that if row compression results in space savings, and 10% increase in CPU usage can be accommodated, all data should use row. They go on to urge careful consideration for Page compression.