Either way, page compression is going to shuffle data around a LOT. I would estimate doing the compression first would result in less overall I/O since the clustering operation will be reading compressed pages.

I would not be surprised if some of the guys at http://dba.stackexchange.com might have far more to say about it!

Either way, page compression is going to shuffle data around a LOT. I would estimate doing the compression first would result in less overall I/O since the clustering operation will be reading compressed pages.