thought you guys might want to be aware - we've been experimenting with compressed indexes on our server with some good preliminary results. our study is quite large, and we have added several custom indexes on all datasets (mostly targets of FKs). these indexes make a noticeable difference.

space concerns prompted the investigation into compression, but apparently it can also be a perf boost (disk IO trumps the overhead of compression). after compression we saved about 10gb on the disk, and test queries are actually quicker. this isnt exactly an exhaustive test yet, but it's looking quite good. may be something to keep in mind for other big installs. this is on SQLserver 2008 r2.

at the moment we're not doing a lot of writes (except background jobs), so we will need to evaluate further on that. posts seem to suggest the impact is minimal though.